{"role": "session_meta", "tools": [{"type": "function", "function": {"name": "browser_back", "description": "Navigate back to the previous page in browser history. Requires browser_navigate to be called first.", "parameters": {"type": "object", "properties": {}}}}, {"type": "function", "function": {"name": "browser_click", "description": "Click on an element identified by its ref ID from the snapshot (e.g., '@e5'). The ref IDs are shown in square brackets in the snapshot output. Requires browser_navigate and browser_snapshot to be called first.", "parameters": {"type": "object", "properties": {"ref": {"type": "string", "description": "The element reference from the snapshot (e.g., '@e5', '@e12')"}}, "required": ["ref"]}}}, {"type": "function", "function": {"name": "browser_console", "description": "Get browser console output and JavaScript errors from the current page. Returns console.log/warn/error/info messages and uncaught JS exceptions. Use this to detect silent JavaScript errors, failed API calls, and application warnings. Requires browser_navigate to be called first. When 'expression' is provided, evaluates JavaScript in the page context and returns the result — use this for DOM inspection, reading page state, or extracting data programmatically.", "parameters": {"type": "object", "properties": {"clear": {"type": "boolean", "default": false, "description": "If true, clear the message buffers after reading"}, "expression": {"type": "string", "description": "JavaScript expression to evaluate in the page context. Runs in the browser like DevTools console — full access to DOM, window, document. Return values are serialized to JSON. Example: 'document.title' or 'document.querySelectorAll(\"a\").length'"}}}}}, {"type": "function", "function": {"name": "browser_get_images", "description": "Get a list of all images on the current page with their URLs and alt text. Useful for finding images to analyze with the vision tool. Requires browser_navigate to be called first.", "parameters": {"type": "object", "properties": {}}}}, {"type": "function", "function": {"name": "browser_navigate", "description": "Navigate to a URL in the browser. Initializes the session and loads the page. Must be called before other browser tools. For plain-text endpoints — URLs ending in .md, .txt, .json, .yaml, .yml, .csv, .xml, raw.githubusercontent.com, or any documented API endpoint — prefer curl via the terminal tool or web_extract; the browser stack is overkill and much slower for these. Use browser tools when you need to interact with a page (click, fill forms, dynamic content). Returns a compact page snapshot with interactive elements and ref IDs — no need to call browser_snapshot separately after navigating.", "parameters": {"type": "object", "properties": {"url": {"type": "string", "description": "The URL to navigate to (e.g., 'https://example.com')"}}, "required": ["url"]}}}, {"type": "function", "function": {"name": "browser_press", "description": "Press a keyboard key. Useful for submitting forms (Enter), navigating (Tab), or keyboard shortcuts. Requires browser_navigate to be called first.", "parameters": {"type": "object", "properties": {"key": {"type": "string", "description": "Key to press (e.g., 'Enter', 'Tab', 'Escape', 'ArrowDown')"}}, "required": ["key"]}}}, {"type": "function", "function": {"name": "browser_scroll", "description": "Scroll the page in a direction. Use this to reveal more content that may be below or above the current viewport. Requires browser_navigate to be called first.", "parameters": {"type": "object", "properties": {"direction": {"type": "string", "enum": ["up", "down"], "description": "Direction to scroll"}}, "required": ["direction"]}}}, {"type": "function", "function": {"name": "browser_snapshot", "description": "Get a text-based snapshot of the current page's accessibility tree. Returns interactive elements with ref IDs (like @e1, @e2) for browser_click and browser_type. full=false (default): compact view with interactive elements. full=true: complete page content. Snapshots over 8000 chars are truncated or LLM-summarized. Requires browser_navigate first. Note: browser_navigate already returns a compact snapshot — use this to refresh after interactions that change the page, or with full=true for complete content.", "parameters": {"type": "object", "properties": {"full": {"type": "boolean", "description": "If true, returns complete page content. If false (default), returns compact view with interactive elements only.", "default": false}}}}}, {"type": "function", "function": {"name": "browser_type", "description": "Type text into an input field identified by its ref ID. Clears the field first, then types the new text. Requires browser_navigate and browser_snapshot to be called first.", "parameters": {"type": "object", "properties": {"ref": {"type": "string", "description": "The element reference from the snapshot (e.g., '@e3')"}, "text": {"type": "string", "description": "The text to type into the field"}}, "required": ["ref", "text"]}}}, {"type": "function", "function": {"name": "browser_vision", "description": "Take a screenshot of the current page and analyze it with vision AI. Use this when you need to visually understand what's on the page - especially useful for CAPTCHAs, visual verification challenges, complex layouts, or when the text snapshot doesn't capture important visual information. Returns both the AI analysis and a screenshot_path that you can share with the user by including MEDIA: in your response. Requires browser_navigate to be called first.", "parameters": {"type": "object", "properties": {"question": {"type": "string", "description": "What you want to know about the page visually. Be specific about what you're looking for."}, "annotate": {"type": "boolean", "default": false, "description": "If true, overlay numbered [N] labels on interactive elements. Each [N] maps to ref @eN for subsequent browser commands. Useful for QA and spatial reasoning about page layout."}}, "required": ["question"]}}}, {"type": "function", "function": {"name": "clarify", "description": "Ask the user a question when you need clarification, feedback, or a decision before proceeding. Supports two modes:\n\n1. **Multiple choice** — provide up to 4 choices. The user picks one or types their own answer via a 5th 'Other' option.\n2. **Open-ended** — omit choices entirely. The user types a free-form response.\n\nUse this tool when:\n- The task is ambiguous and you need the user to choose an approach\n- You want post-task feedback ('How did that work out?')\n- You want to offer to save a skill or update memory\n- A decision has meaningful trade-offs the user should weigh in on\n\nDo NOT use this tool for simple yes/no confirmation of dangerous commands (the terminal tool handles that). Prefer making a reasonable default choice yourself when the decision is low-stakes.", "parameters": {"type": "object", "properties": {"question": {"type": "string", "description": "The question to present to the user."}, "choices": {"type": "array", "items": {"type": "string"}, "maxItems": 4, "description": "Up to 4 answer choices. Omit this parameter entirely to ask an open-ended question. When provided, the UI automatically appends an 'Other (type your answer)' option."}}, "required": ["question"]}}}, {"type": "function", "function": {"name": "cronjob", "description": "Manage scheduled cron jobs with a single compressed tool.\n\nUse action='create' to schedule a new job from a prompt or one or more skills.\nUse action='list' to inspect jobs.\nUse action='update', 'pause', 'resume', 'remove', or 'run' to manage an existing job.\n\nTo stop a job the user no longer wants: first action='list' to find the job_id, then action='remove' with that job_id. Never guess job IDs — always list first.\n\nJobs run in a fresh session with no current-chat context, so prompts must be self-contained.\nIf skills are provided on create, the future cron run loads those skills in order, then follows the prompt as the task instruction.\nOn update, passing skills=[] clears attached skills.\n\nNOTE: The agent's final response is auto-delivered to the target. Put the primary\nuser-facing content in the final response. Cron jobs run autonomously with no user\npresent — they cannot ask questions or request clarification.\n\nImportant safety rule: cron-run sessions should not recursively schedule more cron jobs.", "parameters": {"type": "object", "properties": {"action": {"type": "string", "description": "One of: create, list, update, pause, resume, remove, run"}, "job_id": {"type": "string", "description": "Required for update/pause/resume/remove/run"}, "prompt": {"type": "string", "description": "For create: the full self-contained prompt. If skills are also provided, this becomes the task instruction paired with those skills."}, "schedule": {"type": "string", "description": "For create/update: '30m', 'every 2h', '0 9 * * *', or ISO timestamp"}, "name": {"type": "string", "description": "Optional human-friendly name"}, "repeat": {"type": "integer", "description": "Optional repeat count. Omit for defaults (once for one-shot, forever for recurring)."}, "deliver": {"type": "string", "description": "Omit this parameter to auto-deliver back to the current chat and topic (recommended). Auto-detection preserves thread/topic context. Only set explicitly when the user asks to deliver somewhere OTHER than the current conversation. Values: 'origin' (same as omitting), 'local' (no delivery, save only), 'all' (fan out to every connected home channel), or platform:chat_id:thread_id for a specific destination. Combine with comma: 'origin,all' delivers to the origin plus every other connected channel. Examples: 'telegram:-1001234567890:17585', 'discord:#engineering', 'sms:+15551234567', 'all'. WARNING: 'platform:chat_id' without :thread_id loses topic targeting. 'all' resolves at fire time, so a job created before a channel was wired up will pick it up automatically once connected."}, "skills": {"type": "array", "items": {"type": "string"}, "description": "Optional ordered list of skill names to load before executing the cron prompt. On update, pass an empty array to clear attached skills."}, "model": {"type": "object", "description": "Optional per-job model override. If provider is omitted, the current main provider is pinned at creation time so the job stays stable.", "properties": {"provider": {"type": "string", "description": "Provider name (e.g. 'openrouter', 'anthropic', or 'custom:' for a provider defined in custom_providers config — always include the ':' suffix, never pass the bare 'custom'). Omit to use and pin the current provider."}, "model": {"type": "string", "description": "Model name (e.g. 'anthropic/claude-sonnet-4', 'claude-sonnet-4')"}}, "required": ["model"]}, "script": {"type": "string", "description": "Optional path to a script that runs each tick. In the default mode its stdout is injected into the agent's prompt as context (data-collection / change-detection pattern). With no_agent=True, the script IS the job and its stdout is delivered verbatim (classic watchdog pattern). Relative paths resolve under ~/.hermes/scripts/. ``.sh``/``.bash`` extensions run via bash, everything else via Python. On update, pass empty string to clear."}, "no_agent": {"type": "boolean", "default": false, "description": "Default: False (LLM-driven job — the agent runs the prompt each tick). Set True to skip the LLM entirely: the scheduler just runs ``script`` on schedule and delivers its stdout verbatim. No tokens, no agent loop, no model override honoured. \n\nREQUIREMENTS when True: ``script`` MUST be set (``prompt`` and ``skills`` are ignored). \n\nDELIVERY SEMANTICS when True: (a) non-empty stdout is sent verbatim as the message; (b) EMPTY stdout means SILENT — nothing is sent to the user and they won't see anything happened, so design your script to stay quiet when there's nothing to report (the watchdog pattern); (c) non-zero exit / timeout sends an error alert so a broken watchdog can't fail silently. \n\nWHEN TO USE True: recurring script-only pings where the script itself produces the exact message text (memory/disk/GPU watchdogs, threshold alerts, heartbeats, CI notifications, API pollers with a fixed output shape). WHEN TO USE False (default): anything that needs reasoning — summarize a feed, draft a daily briefing, pick interesting items, rephrase data for a human, follow conditional logic based on content."}, "context_from": {"type": "array", "items": {"type": "string"}, "description": "Optional job ID or list of job IDs whose most recent completed output is injected into the prompt as context before each run. Use this to chain cron jobs: job A collects data, job B processes it. Each entry must be a valid job ID (from cronjob action='list'). Note: injects the most recent completed output — does not wait for upstream jobs running in the same tick. On update, pass an empty array to clear."}, "enabled_toolsets": {"type": "array", "items": {"type": "string"}, "description": "Optional list of toolset names to restrict the job's agent to (e.g. [\"web\", \"terminal\", \"file\", \"delegation\"]). When set, only tools from these toolsets are loaded, significantly reducing input token overhead. When omitted, all default tools are loaded. Infer from the job's prompt — e.g. use \"web\" if it calls web_search, \"terminal\" if it runs scripts, \"file\" if it reads files, \"delegation\" if it calls delegate_task. On update, pass an empty array to clear."}, "workdir": {"type": "string", "description": "Optional absolute path to run the job from. When set, AGENTS.md / CLAUDE.md / .cursorrules from that directory are injected into the system prompt, and the terminal/file/code_exec tools use it as their working directory — useful for running a job inside a specific project repo. Must be an absolute path that exists. When unset (default), preserves the original behaviour: no project context files, tools use the scheduler's cwd. On update, pass an empty string to clear. Jobs with workdir run sequentially (not parallel) to keep per-job directories isolated."}}, "required": ["action"]}}}, {"type": "function", "function": {"name": "delegate_task", "description": "Spawn one or more subagents to work on tasks in isolated contexts. Each subagent gets its own conversation, terminal session, and toolset. Only the final summary is returned -- intermediate tool results never enter your context window.\n\nTWO MODES (one of 'goal' or 'tasks' is required):\n1. Single task: provide 'goal' (+ optional context, toolsets)\n2. Batch (parallel): provide 'tasks' array with up to delegation.max_concurrent_children items (default 3, configurable via config.yaml, no hard ceiling). All run concurrently and results are returned together. Nested delegation requires role='orchestrator' and delegation.max_spawn_depth >= 2.\n\nWHEN TO USE delegate_task:\n- Reasoning-heavy subtasks (debugging, code review, research synthesis)\n- Tasks that would flood your context with intermediate data\n- Parallel independent workstreams (research A and B simultaneously)\n\nWHEN NOT TO USE (use these instead):\n- Mechanical multi-step work with no reasoning needed -> use execute_code\n- Single tool call -> just call the tool directly\n- Tasks needing user interaction -> subagents cannot use clarify\n- Durable long-running work that must outlive the current turn -> use cronjob (action='create') or terminal(background=True, notify_on_complete=True) instead. delegate_task runs SYNCHRONOUSLY inside the parent turn: if the parent is interrupted (user sends a new message, /stop, /new) the child is cancelled with status='interrupted' and its work is discarded. Children cannot continue in the background.\n\nIMPORTANT:\n- Subagents have NO memory of your conversation. Pass all relevant info (file paths, error messages, constraints) via the 'context' field.\n- If the user is writing in a non-English language, or asked for output in a specific language / tone / style, say so in 'context' (e.g. \"respond in Chinese\", \"return output in Japanese\"). Otherwise subagents default to English and their summaries will contaminate your final reply with the wrong language.\n- Subagent summaries are SELF-REPORTS, not verified facts. A subagent that claims \"uploaded successfully\" or \"file written\" may be wrong. For operations with external side-effects (HTTP POST/PUT, remote writes, file creation at shared paths, publishing), require the subagent to return a verifiable handle (URL, ID, absolute path, HTTP status) and verify it yourself — fetch the URL, stat the file, read back the content — before telling the user the operation succeeded.\n- Leaf subagents (role='leaf', the default) CANNOT call: delegate_task, clarify, memory, send_message, execute_code.\n- Orchestrator subagents (role='orchestrator') retain delegate_task so they can spawn their own workers, but still cannot use clarify, memory, send_message, or execute_code. Orchestrators are bounded by delegation.max_spawn_depth (default 2) and can be disabled globally via delegation.orchestrator_enabled=false.\n- Each subagent gets its own terminal session (separate working directory and state).\n- Results are always returned as an array, one entry per task.", "parameters": {"type": "object", "properties": {"goal": {"type": "string", "description": "What the subagent should accomplish. Be specific and self-contained -- the subagent knows nothing about your conversation history."}, "context": {"type": "string", "description": "Background information the subagent needs: file paths, error messages, project structure, constraints. The more specific you are, the better the subagent performs."}, "toolsets": {"type": "array", "items": {"type": "string"}, "description": "Toolsets to enable for this subagent. Default: inherits your enabled toolsets. Available toolsets: 'browser', 'cronjob', 'discord', 'discord_admin', 'feishu_doc', 'feishu_drive', 'file', 'homeassistant', 'image_gen', 'kanban', 'search', 'session_search', 'skills', 'spotify', 'terminal', 'todo', 'tts', 'video', 'vision', 'web', 'yuanbao'. Common patterns: ['terminal', 'file'] for code work, ['web'] for research, ['browser'] for web interaction, ['terminal', 'file', 'web'] for full-stack tasks."}, "tasks": {"type": "array", "items": {"type": "object", "properties": {"goal": {"type": "string", "description": "Task goal"}, "context": {"type": "string", "description": "Task-specific context"}, "toolsets": {"type": "array", "items": {"type": "string"}, "description": "Toolsets for this specific task. Available: 'browser', 'cronjob', 'discord', 'discord_admin', 'feishu_doc', 'feishu_drive', 'file', 'homeassistant', 'image_gen', 'kanban', 'search', 'session_search', 'skills', 'spotify', 'terminal', 'todo', 'tts', 'video', 'vision', 'web', 'yuanbao'. Use 'web' for network access, 'terminal' for shell, 'browser' for web interaction."}, "acp_command": {"type": "string", "description": "Per-task ACP command override (e.g. 'copilot'). Overrides the top-level acp_command for this task only."}, "acp_args": {"type": "array", "items": {"type": "string"}, "description": "Per-task ACP args override."}, "role": {"type": "string", "enum": ["leaf", "orchestrator"], "description": "Per-task role override. See top-level 'role' for semantics."}}, "required": ["goal"]}, "description": "Batch mode: tasks to run in parallel (limit configurable via delegation.max_concurrent_children, default 3). Each gets its own subagent with isolated context and terminal session. When provided, top-level goal/context/toolsets are ignored."}, "role": {"type": "string", "enum": ["leaf", "orchestrator"], "description": "Role of the child agent. 'leaf' (default) = focused worker, cannot delegate further. 'orchestrator' = can use delegate_task to spawn its own workers. Requires delegation.max_spawn_depth >= 2 in config; ignored (treated as 'leaf') when the child would exceed max_spawn_depth or when delegation.orchestrator_enabled=false."}, "acp_command": {"type": "string", "description": "Override ACP command for child agents (e.g. 'copilot'). When set, children use ACP subprocess transport instead of inheriting the parent's transport. Requires an ACP-compatible CLI (currently GitHub Copilot CLI via 'copilot --acp --stdio'). See agent/copilot_acp_client.py for the implementation."}, "acp_args": {"type": "array", "items": {"type": "string"}, "description": "Arguments for the ACP command (default: ['--acp', '--stdio']). Only used when acp_command is set."}}}}}, {"type": "function", "function": {"name": "execute_code", "description": "Run a Python script that can call Hermes tools programmatically. Use this when you need 3+ tool calls with processing logic between them, need to filter/reduce large tool outputs before they enter your context, need conditional branching (if X then Y else Z), or need to loop (fetch N pages, process N files, retry on failure).\n\nUse normal tool calls instead when: single tool call with no processing, you need to see the full result and apply complex reasoning, or the task requires interactive user input.\n\nAvailable via `from hermes_tools import ...`:\n\n read_file(path: str, offset: int = 1, limit: int = 500) -> dict\n Lines are 1-indexed. Returns {\"content\": \"...\", \"total_lines\": N}\n write_file(path: str, content: str) -> dict\n Always overwrites the entire file.\n search_files(pattern: str, target=\"content\", path=\".\", file_glob=None, limit=50) -> dict\n target: \"content\" (search inside files) or \"files\" (find files by name). Returns {\"matches\": [...]}\n patch(path: str, old_string: str, new_string: str, replace_all: bool = False) -> dict\n Replaces old_string with new_string in the file.\n terminal(command: str, timeout=None, workdir=None) -> dict\n Foreground only (no background/pty). Returns {\"output\": \"...\", \"exit_code\": N}\n\nLimits: 5-minute timeout, 50KB stdout cap, max 50 tool calls per script. terminal() is foreground-only (no background or pty).\n\nScripts run in the session's working directory with the active venv's python, so project deps (pandas, etc.) and relative paths work like in terminal().\n\nPrint your final result to stdout. Use Python stdlib (json, re, math, csv, datetime, collections, etc.) for processing between tool calls.\n\nAlso available (no import needed — built into hermes_tools):\n json_parse(text: str) — json.loads with strict=False; use for terminal() output with control chars\n shell_quote(s: str) — shlex.quote(); use when interpolating dynamic strings into shell commands\n retry(fn, max_attempts=3, delay=2) — retry with exponential backoff for transient failures", "parameters": {"type": "object", "properties": {"code": {"type": "string", "description": "Python code to execute. Import tools with `from hermes_tools import terminal, ...` and print your final result to stdout."}}, "required": ["code"]}}}, {"type": "function", "function": {"name": "mcp_openclaw_attachments_fetch", "description": "List non-text attachments for a message in one OpenClaw conversation.", "parameters": {"$schema": "http://json-schema.org/draft-07/schema#", "type": "object", "properties": {"session_key": {"type": "string", "minLength": 1}, "message_id": {"type": "string", "minLength": 1}, "limit": {"type": "integer", "minimum": 1, "maximum": 200}}, "required": ["session_key", "message_id"]}}}, {"type": "function", "function": {"name": "mcp_openclaw_conversation_get", "description": "Get one OpenClaw conversation by session key.", "parameters": {"$schema": "http://json-schema.org/draft-07/schema#", "type": "object", "properties": {"session_key": {"type": "string", "minLength": 1}}, "required": ["session_key"]}}}, {"type": "function", "function": {"name": "mcp_openclaw_conversations_list", "description": "List OpenClaw channel-backed conversations available through session routes.", "parameters": {"$schema": "http://json-schema.org/draft-07/schema#", "type": "object", "properties": {"limit": {"type": "integer", "minimum": 1, "maximum": 500}, "search": {"type": "string"}, "channel": {"type": "string"}, "includeDerivedTitles": {"type": "boolean"}, "includeLastMessage": {"type": "boolean"}}}}}, {"type": "function", "function": {"name": "mcp_openclaw_events_poll", "description": "Poll queued OpenClaw conversation events since a cursor.", "parameters": {"$schema": "http://json-schema.org/draft-07/schema#", "type": "object", "properties": {"after_cursor": {"type": "integer", "minimum": 0, "maximum": 9007199254740991}, "session_key": {"type": "string"}, "limit": {"type": "integer", "minimum": 1, "maximum": 200}}}}}, {"type": "function", "function": {"name": "mcp_openclaw_events_wait", "description": "Wait for the next queued OpenClaw conversation event.", "parameters": {"$schema": "http://json-schema.org/draft-07/schema#", "type": "object", "properties": {"after_cursor": {"type": "integer", "minimum": 0, "maximum": 9007199254740991}, "session_key": {"type": "string"}, "timeout_ms": {"type": "integer", "minimum": 1, "maximum": 300000}}}}}, {"type": "function", "function": {"name": "mcp_openclaw_messages_read", "description": "Read recent messages for one OpenClaw conversation.", "parameters": {"$schema": "http://json-schema.org/draft-07/schema#", "type": "object", "properties": {"session_key": {"type": "string", "minLength": 1}, "limit": {"type": "integer", "minimum": 1, "maximum": 200}}, "required": ["session_key"]}}}, {"type": "function", "function": {"name": "mcp_openclaw_messages_send", "description": "Send a message back through the same OpenClaw conversation route.", "parameters": {"$schema": "http://json-schema.org/draft-07/schema#", "type": "object", "properties": {"session_key": {"type": "string", "minLength": 1}, "text": {"type": "string", "minLength": 1}}, "required": ["session_key", "text"]}}}, {"type": "function", "function": {"name": "mcp_openclaw_permissions_list_open", "description": "List open OpenClaw exec or plugin approval requests visible through the Gateway.", "parameters": {"$schema": "http://json-schema.org/draft-07/schema#", "type": "object", "properties": {}}}}, {"type": "function", "function": {"name": "mcp_openclaw_permissions_respond", "description": "Allow or deny one pending OpenClaw exec or plugin approval request.", "parameters": {"$schema": "http://json-schema.org/draft-07/schema#", "type": "object", "properties": {"kind": {"type": "string", "enum": ["exec", "plugin"]}, "id": {"type": "string", "minLength": 1}, "decision": {"type": "string", "enum": ["allow-once", "allow-always", "deny"]}}, "required": ["kind", "id", "decision"]}}}, {"type": "function", "function": {"name": "memory", "description": "Save durable information to persistent memory that survives across sessions. Memory is injected into future turns, so keep it compact and focused on facts that will still matter later.\n\nWHEN TO SAVE (do this proactively, don't wait to be asked):\n- User corrects you or says 'remember this' / 'don't do that again'\n- User shares a preference, habit, or personal detail (name, role, timezone, coding style)\n- You discover something about the environment (OS, installed tools, project structure)\n- You learn a convention, API quirk, or workflow specific to this user's setup\n- You identify a stable fact that will be useful again in future sessions\n\nPRIORITY: User preferences and corrections > environment facts > procedural knowledge. The most valuable memory prevents the user from having to repeat themselves.\n\nDo NOT save task progress, session outcomes, completed-work logs, or temporary TODO state to memory; use session_search to recall those from past transcripts.\nIf you've discovered a new way to do something, solved a problem that could be necessary later, save it as a skill with the skill tool.\n\nTWO TARGETS:\n- 'user': who the user is -- name, role, preferences, communication style, pet peeves\n- 'memory': your notes -- environment facts, project conventions, tool quirks, lessons learned\n\nACTIONS: add (new entry), replace (update existing -- old_text identifies it), remove (delete -- old_text identifies it).\n\nSKIP: trivial/obvious info, things easily re-discovered, raw data dumps, and temporary task state.", "parameters": {"type": "object", "properties": {"action": {"type": "string", "enum": ["add", "replace", "remove"], "description": "The action to perform."}, "target": {"type": "string", "enum": ["memory", "user"], "description": "Which memory store: 'memory' for personal notes, 'user' for user profile."}, "content": {"type": "string", "description": "The entry content. Required for 'add' and 'replace'."}, "old_text": {"type": "string", "description": "Short unique substring identifying the entry to replace or remove."}}, "required": ["action", "target"]}}}, {"type": "function", "function": {"name": "patch", "description": "Targeted find-and-replace edits in files. Use this instead of sed/awk in terminal. Uses fuzzy matching (9 strategies) so minor whitespace/indentation differences won't break it. Returns a unified diff. Auto-runs syntax checks after editing.\n\nReplace mode (default): find a unique string and replace it.\nPatch mode: apply V4A multi-file patches for bulk changes.", "parameters": {"type": "object", "properties": {"mode": {"type": "string", "enum": ["replace", "patch"], "description": "Edit mode: 'replace' for targeted find-and-replace, 'patch' for V4A multi-file patches", "default": "replace"}, "path": {"type": "string", "description": "File path to edit (required for 'replace' mode)"}, "old_string": {"type": "string", "description": "Text to find in the file (required for 'replace' mode). Must be unique in the file unless replace_all=true. Include enough surrounding context to ensure uniqueness."}, "new_string": {"type": "string", "description": "Replacement text (required for 'replace' mode). Can be empty string to delete the matched text."}, "replace_all": {"type": "boolean", "description": "Replace all occurrences instead of requiring a unique match (default: false)", "default": false}, "patch": {"type": "string", "description": "V4A format patch content (required for 'patch' mode). Format:\n*** Begin Patch\n*** Update File: path/to/file\n@@ context hint @@\n context line\n-removed line\n+added line\n*** End Patch"}}, "required": ["mode"]}}}, {"type": "function", "function": {"name": "process", "description": "Manage background processes started with terminal(background=true). Actions: 'list' (show all), 'poll' (check status + new output), 'log' (full output with pagination), 'wait' (block until done or timeout), 'kill' (terminate), 'write' (send raw stdin data without newline), 'submit' (send data + Enter, for answering prompts), 'close' (close stdin/send EOF).", "parameters": {"type": "object", "properties": {"action": {"type": "string", "enum": ["list", "poll", "log", "wait", "kill", "write", "submit", "close"], "description": "Action to perform on background processes"}, "session_id": {"type": "string", "description": "Process session ID (from terminal background output). Required for all actions except 'list'."}, "data": {"type": "string", "description": "Text to send to process stdin (for 'write' and 'submit' actions)"}, "timeout": {"type": "integer", "description": "Max seconds to block for 'wait' action. Returns partial output on timeout.", "minimum": 1}, "offset": {"type": "integer", "description": "Line offset for 'log' action (default: last 200 lines)"}, "limit": {"type": "integer", "description": "Max lines to return for 'log' action", "minimum": 1}}, "required": ["action"]}}}, {"type": "function", "function": {"name": "read_file", "description": "Read a text file with line numbers and pagination. Use this instead of cat/head/tail in terminal. Output format: 'LINE_NUM|CONTENT'. Suggests similar filenames if not found. Use offset and limit for large files. Reads exceeding ~100K characters are rejected; use offset and limit to read specific sections of large files. NOTE: Cannot read images or binary files — use vision_analyze for images.", "parameters": {"type": "object", "properties": {"path": {"type": "string", "description": "Path to the file to read (absolute, relative, or ~/path)"}, "offset": {"type": "integer", "description": "Line number to start reading from (1-indexed, default: 1)", "default": 1, "minimum": 1}, "limit": {"type": "integer", "description": "Maximum number of lines to read (default: 500, max: 2000)", "default": 500, "maximum": 2000}}, "required": ["path"]}}}, {"type": "function", "function": {"name": "search_files", "description": "Search file contents or find files by name. Use this instead of grep/rg/find/ls in terminal. Ripgrep-backed, faster than shell equivalents.\n\nContent search (target='content'): Regex search inside files. Output modes: full matches with line numbers, file paths only, or match counts.\n\nFile search (target='files'): Find files by glob pattern (e.g., '*.py', '*config*'). Also use this instead of ls — results sorted by modification time.", "parameters": {"type": "object", "properties": {"pattern": {"type": "string", "description": "Regex pattern for content search, or glob pattern (e.g., '*.py') for file search"}, "target": {"type": "string", "enum": ["content", "files"], "description": "'content' searches inside file contents, 'files' searches for files by name", "default": "content"}, "path": {"type": "string", "description": "Directory or file to search in (default: current working directory)", "default": "."}, "file_glob": {"type": "string", "description": "Filter files by pattern in grep mode (e.g., '*.py' to only search Python files)"}, "limit": {"type": "integer", "description": "Maximum number of results to return (default: 50)", "default": 50}, "offset": {"type": "integer", "description": "Skip first N results for pagination (default: 0)", "default": 0}, "output_mode": {"type": "string", "enum": ["content", "files_only", "count"], "description": "Output format for grep mode: 'content' shows matching lines with line numbers, 'files_only' lists file paths, 'count' shows match counts per file", "default": "content"}, "context": {"type": "integer", "description": "Number of context lines before and after each match (grep mode only)", "default": 0}}, "required": ["pattern"]}}}, {"type": "function", "function": {"name": "send_message", "description": "Send a message to a connected messaging platform, or list available targets.\n\nIMPORTANT: When the user asks to send to a specific channel or person (not just a bare platform name), call send_message(action='list') FIRST to see available targets, then send to the correct one.\nIf the user just says a platform name like 'send to telegram', send directly to the home channel without listing first.", "parameters": {"type": "object", "properties": {"action": {"type": "string", "enum": ["send", "list"], "description": "Action to perform. 'send' (default) sends a message. 'list' returns all available channels/contacts across connected platforms."}, "target": {"type": "string", "description": "Delivery target. Format: 'platform' (uses home channel), 'platform:#channel-name', 'platform:chat_id', or 'platform:chat_id:thread_id' for Telegram topics and Discord threads. Examples: 'telegram', 'telegram:-1001234567890:17585', 'discord:999888777:555444333', 'discord:#bot-home', 'slack:#engineering', 'signal:+155****4567', 'matrix:!roomid:server.org', 'matrix:@user:server.org', 'yuanbao:direct:' (DM), 'yuanbao:group:' (group chat)"}, "message": {"type": "string", "description": "The message text to send. To send an image or file, include MEDIA: (e.g. 'MEDIA:/tmp/hermes/cache/img_xxx.jpg') in the message — the platform will deliver it as a native media attachment."}}}}}, {"type": "function", "function": {"name": "session_search", "description": "Search your long-term memory of past conversations, or browse recent sessions. This is your recall -- every past session is searchable, and this tool summarizes what happened.\n\nTWO MODES:\n1. Recent sessions (no query): Call with no arguments to see what was worked on recently. Returns titles, previews, and timestamps. Zero LLM cost, instant. Start here when the user asks what were we working on or what did we do recently.\n2. Keyword search (with query): Search for specific topics across all past sessions. Returns LLM-generated summaries of matching sessions.\n\nUSE THIS PROACTIVELY when:\n- The user says 'we did this before', 'remember when', 'last time', 'as I mentioned'\n- The user asks about a topic you worked on before but don't have in current context\n- The user references a project, person, or concept that seems familiar but isn't in memory\n- You want to check if you've solved a similar problem before\n- The user asks 'what did we do about X?' or 'how did we fix Y?'\n\nDon't hesitate to search when it is actually cross-session -- it's fast and cheap. Better to search and confirm than to guess or ask the user to repeat themselves.\n\nSearch syntax: keywords joined with OR for broad recall (elevenlabs OR baseten OR funding), phrases for exact match (\"docker networking\"), boolean (python NOT java), prefix (deploy*). IMPORTANT: Use OR between keywords for best results — FTS5 defaults to AND which misses sessions that only mention some terms. If a broad OR query returns nothing, try individual keyword searches in parallel. Returns summaries of the top matching sessions.", "parameters": {"type": "object", "properties": {"query": {"type": "string", "description": "Search query — keywords, phrases, or boolean expressions to find in past sessions. Omit this parameter entirely to browse recent sessions instead (returns titles, previews, timestamps with no LLM cost)."}, "role_filter": {"type": "string", "description": "Optional: only search messages from specific roles (comma-separated). E.g. 'user,assistant' to skip tool outputs."}, "limit": {"type": "integer", "description": "Max sessions to summarize (default: 3, max: 5).", "default": 3}}}}}, {"type": "function", "function": {"name": "skill_manage", "description": "Manage skills (create, update, delete). Skills are your procedural memory — reusable approaches for recurring task types. New skills go to ~/.hermes/skills/; existing skills can be modified wherever they live.\n\nActions: create (full SKILL.md + optional category), patch (old_string/new_string — preferred for fixes), edit (full SKILL.md rewrite — major overhauls only), delete, write_file, remove_file.\n\nOn delete, pass `absorbed_into=` when you're merging this skill's content into another one, or `absorbed_into=\"\"` when you're pruning it with no forwarding target. This lets the curator tell consolidation from pruning without guessing, so downstream consumers (cron jobs that reference the old skill name, etc.) get updated correctly. The target you name in `absorbed_into` must already exist — create/patch the umbrella first, then delete.\n\nCreate when: complex task succeeded (5+ calls), errors overcome, user-corrected approach worked, non-trivial workflow discovered, or user asks you to remember a procedure.\nUpdate when: instructions stale/wrong, OS-specific failures, missing steps or pitfalls found during use. If you used a skill and hit issues not covered by it, patch it immediately.\n\nAfter difficult/iterative tasks, offer to save as a skill. Skip for simple one-offs. Confirm with user before creating/deleting.\n\nGood skills: trigger conditions, numbered steps with exact commands, pitfalls section, verification steps. Use skill_view() to see format examples.\n\nPinned skills are protected from deletion only — skill_manage(action='delete') will refuse with a message pointing the user to `hermes curator unpin `. Patches and edits go through on pinned skills so you can still improve them as pitfalls come up; pin only guards against irrecoverable loss.", "parameters": {"type": "object", "properties": {"action": {"type": "string", "enum": ["create", "patch", "edit", "delete", "write_file", "remove_file"], "description": "The action to perform."}, "name": {"type": "string", "description": "Skill name (lowercase, hyphens/underscores, max 64 chars). Must match an existing skill for patch/edit/delete/write_file/remove_file."}, "content": {"type": "string", "description": "Full SKILL.md content (YAML frontmatter + markdown body). Required for 'create' and 'edit'. For 'edit', read the skill first with skill_view() and provide the complete updated text."}, "old_string": {"type": "string", "description": "Text to find in the file (required for 'patch'). Must be unique unless replace_all=true. Include enough surrounding context to ensure uniqueness."}, "new_string": {"type": "string", "description": "Replacement text (required for 'patch'). Can be empty string to delete the matched text."}, "replace_all": {"type": "boolean", "description": "For 'patch': replace all occurrences instead of requiring a unique match (default: false)."}, "category": {"type": "string", "description": "Optional category/domain for organizing the skill (e.g., 'devops', 'data-science', 'mlops'). Creates a subdirectory grouping. Only used with 'create'."}, "file_path": {"type": "string", "description": "Path to a supporting file within the skill directory. For 'write_file'/'remove_file': required, must be under references/, templates/, scripts/, or assets/. For 'patch': optional, defaults to SKILL.md if omitted."}, "file_content": {"type": "string", "description": "Content for the file. Required for 'write_file'."}, "absorbed_into": {"type": "string", "description": "For 'delete' only — declares intent so the curator can tell consolidation from pruning without guessing. Pass the umbrella skill name when this skill's content was merged into another (the target must already exist). Pass an empty string when the skill is truly stale and being pruned with no forwarding target. Omitting the arg on delete is supported for backward compatibility but downstream tooling (e.g. cron-job skill reference rewriting) will have to guess at intent."}}, "required": ["action", "name"]}}}, {"type": "function", "function": {"name": "skill_view", "description": "Skills allow for loading information about specific tasks and workflows, as well as scripts and templates. Load a skill's full content or access its linked files (references, templates, scripts). First call returns SKILL.md content plus a 'linked_files' dict showing available references/templates/scripts. To access those, call again with file_path parameter.", "parameters": {"type": "object", "properties": {"name": {"type": "string", "description": "The skill name (use skills_list to see available skills). For plugin-provided skills, use the qualified form 'plugin:skill' (e.g. 'superpowers:writing-plans')."}, "file_path": {"type": "string", "description": "OPTIONAL: Path to a linked file within the skill (e.g., 'references/api.md', 'templates/config.yaml', 'scripts/validate.py'). Omit to get the main SKILL.md content."}}, "required": ["name"]}}}, {"type": "function", "function": {"name": "skills_list", "description": "List available skills (name + description). Use skill_view(name) to load full content.", "parameters": {"type": "object", "properties": {"category": {"type": "string", "description": "Optional category filter to narrow results"}}}}}, {"type": "function", "function": {"name": "terminal", "description": "Execute shell commands on a Linux environment. Filesystem usually persists between calls.\n\nDo NOT use cat/head/tail to read files — use read_file instead.\nDo NOT use grep/rg/find to search — use search_files instead.\nDo NOT use ls to list directories — use search_files(target='files') instead.\nDo NOT use sed/awk to edit files — use patch instead.\nDo NOT use echo/cat heredoc to create files — use write_file instead.\nReserve terminal for: builds, installs, git, processes, scripts, network, package managers, and anything that needs a shell.\n\nForeground (default): Commands return INSTANTLY when done, even if the timeout is high. Set timeout=300 for long builds/scripts — you'll still get the result in seconds if it's fast. Prefer foreground for short commands.\nBackground: Set background=true to get a session_id. Two patterns:\n (1) Long-lived processes that never exit (servers, watchers).\n (2) Long-running tasks with notify_on_complete=true — you can keep working on other things and the system auto-notifies you when the task finishes. Great for test suites, builds, deployments, or anything that takes more than a minute.\nFor servers/watchers, do NOT use shell-level background wrappers (nohup/disown/setsid/trailing '&') in foreground mode. Use background=true so Hermes can track lifecycle and output.\nAfter starting a server, verify readiness with a health check or log signal, then run tests in a separate terminal() call. Avoid blind sleep loops.\nUse process(action=\"poll\") for progress checks, process(action=\"wait\") to block until done.\nWorking directory: Use 'workdir' for per-command cwd.\nPTY mode: Set pty=true for interactive CLI tools (Codex, Claude Code, Python REPL).\n\nDo NOT use vim/nano/interactive tools without pty=true — they hang without a pseudo-terminal. Pipe git output to cat if it might page.\n", "parameters": {"type": "object", "properties": {"command": {"type": "string", "description": "The command to execute on the VM"}, "background": {"type": "boolean", "description": "Run the command in the background. Two patterns: (1) Long-lived processes that never exit (servers, watchers). (2) Long-running tasks paired with notify_on_complete=true — you can keep working and get notified when the task finishes. For short commands, prefer foreground with a generous timeout instead.", "default": false}, "timeout": {"type": "integer", "description": "Max seconds to wait (default: 180, foreground max: 600). Returns INSTANTLY when command finishes — set high for long tasks, you won't wait unnecessarily. Foreground timeout above 600s is rejected; use background=true for longer commands.", "minimum": 1}, "workdir": {"type": "string", "description": "Working directory for this command (absolute path). Defaults to the session working directory."}, "pty": {"type": "boolean", "description": "Run in pseudo-terminal (PTY) mode for interactive CLI tools like Codex, Claude Code, or Python REPL. Only works with local and SSH backends. Default: false.", "default": false}, "notify_on_complete": {"type": "boolean", "description": "When true (and background=true), you'll be automatically notified exactly once when the process finishes. **This is the right choice for almost every long-running task** — tests, builds, deployments, multi-item batch jobs, anything that takes over a minute and has a defined end. Use this and keep working on other things; the system notifies you on exit. MUTUALLY EXCLUSIVE with watch_patterns — when both are set, watch_patterns is dropped.", "default": false}, "watch_patterns": {"type": "array", "items": {"type": "string"}, "description": "Strings to watch for in background process output. HARD RATE LIMIT: at most 1 notification per 15 seconds per process — matches arriving inside the cooldown are dropped. After 3 consecutive 15-second windows with dropped matches, watch_patterns is automatically disabled for that process and promoted to notify_on_complete behavior (one notification on exit, no more mid-process spam). USE ONLY for truly rare, one-shot mid-process signals on LONG-LIVED processes that will never exit on their own — e.g. ['Application startup complete'] on a server so you know when to hit its endpoint, or ['migration done'] on a daemon. DO NOT use for: (1) end-of-run markers like 'DONE'/'PASS' — use notify_on_complete instead; (2) error patterns like 'ERROR'/'Traceback' in loops or multi-item batch jobs — they fire on every iteration and you'll hit the strike limit fast; (3) anything you'd ever combine with notify_on_complete. When in doubt, choose notify_on_complete. MUTUALLY EXCLUSIVE with notify_on_complete — set one, not both."}}, "required": ["command"]}}}, {"type": "function", "function": {"name": "text_to_speech", "description": "Convert text to speech audio. Returns a MEDIA: path that the platform delivers as native audio. Compatible providers render as a voice bubble on Telegram; otherwise audio is sent as a regular attachment. In CLI mode, saves to ~/voice-memos/. Voice and provider are user-configured (built-in providers like edge/openai or custom command providers under tts.providers.), not model-selected.", "parameters": {"type": "object", "properties": {"text": {"type": "string", "description": "The text to convert to speech. Provider-specific character caps apply and are enforced automatically (OpenAI 4096, xAI 15000, MiniMax 10000, ElevenLabs 5k-40k depending on model); over-long input is truncated."}, "output_path": {"type": "string", "description": "Optional custom file path to save the audio. Defaults to ~/.hermes/audio_cache/.mp3"}}, "required": ["text"]}}}, {"type": "function", "function": {"name": "todo", "description": "Manage your task list for the current session. Use for complex tasks with 3+ steps or when the user provides multiple tasks. Call with no parameters to read the current list.\n\nWriting:\n- Provide 'todos' array to create/update items\n- merge=false (default): replace the entire list with a fresh plan\n- merge=true: update existing items by id, add any new ones\n\nEach item: {id: string, content: string, status: pending|in_progress|completed|cancelled}\nList order is priority. Only ONE item in_progress at a time.\nMark items completed immediately when done. If something fails, cancel it and add a revised item.\n\nAlways returns the full current list.", "parameters": {"type": "object", "properties": {"todos": {"type": "array", "description": "Task items to write. Omit to read current list.", "items": {"type": "object", "properties": {"id": {"type": "string", "description": "Unique item identifier"}, "content": {"type": "string", "description": "Task description"}, "status": {"type": "string", "enum": ["pending", "in_progress", "completed", "cancelled"], "description": "Current status"}}, "required": ["id", "content", "status"]}}, "merge": {"type": "boolean", "description": "true: update existing items by id, add new ones. false (default): replace the entire list.", "default": false}}}}}, {"type": "function", "function": {"name": "vision_analyze", "description": "Inspect an image from a URL, file path, or tool output when you need closer detail than what's visible in the conversation. If the user's image is already attached to the conversation and you can see it, just answer directly — only call this tool for images referenced by URL/path, images returned inside other tool results (browser screenshots, search thumbnails), or when you need a deeper look at a specific region the main model's vision may have missed.", "parameters": {"type": "object", "properties": {"image_url": {"type": "string", "description": "Image URL (http/https) or local file path to analyze."}, "question": {"type": "string", "description": "Your specific question or request about the image to resolve. The AI will automatically provide a complete image description AND answer your specific question."}}, "required": ["image_url", "question"]}}}, {"type": "function", "function": {"name": "write_file", "description": "Write content to a file, completely replacing existing content. Use this instead of echo/cat heredoc in terminal. Creates parent directories automatically. OVERWRITES the entire file — use 'patch' for targeted edits. Auto-runs syntax checks on .py/.json/.yaml/.toml and other linted languages; only NEW errors introduced by this write are surfaced (pre-existing errors are filtered out).", "parameters": {"type": "object", "properties": {"path": {"type": "string", "description": "Path to the file to write (will be created if it doesn't exist, overwritten if it does)"}, "content": {"type": "string", "description": "Complete content to write to the file"}}, "required": ["path", "content"]}}}], "model": "qwen-3.6-agent", "platform": "telegram", "timestamp": "2026-05-10T08:31:28.853389"} {"role": "user", "content": "Erstelle einen ,Archiv‘ ordner im Gmail account. Verschiebe diese beiden emails ins Archiv.", "timestamp": "2026-05-10T08:31:28.853389"} {"role": "assistant", "content": "Ich werde zuerst den Archiv-Ordner erstellen und dann die E-Mails suchen und verschieben.", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "taPUsubXwDLMAZQtyglJZJFT2fHOqHYA", "call_id": "taPUsubXwDLMAZQtyglJZJFT2fHOqHYA", "response_item_id": "fc_taPUsubXwDLMAZQtyglJZJFT2fHOqHYA", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"openclaw gog label create --name \\\"Archiv\\\" --action label\",\"timeout\":30}"}}], "timestamp": "2026-05-10T08:31:28.853389"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"[plugins] [lcm] OpenClaw runtime.modelAuth is unavailable; using legacy auth-profiles fallback. Stock lossless-claw 0.2.7 expects OpenClaw plugin runtime support from PR #41090 (https://github.com/openclaw/openclaw/pull/41090). OpenClaw 2026.3.8 and 2026.3.8-beta.1 do not include merge commit 4790e40; the first OpenClaw release after 2026.3.8 is required for stock lossless-claw 0.2.7 without this fallback patch.\\n[plugins] [lcm] Transcript GC enabled (default false)\\n[plugins] [lcm] Proactive threshold compaction mode: deferred (default deferred)\\n[plugins] [lcm] migration step complete: step=ensureSummaryDepthColumn durationMs=0\\n[plugins] [lcm] migration step complete: step=ensureSummaryMetadataColumns durationMs=0\\n[plugins] [lcm] migration step complete: step=ensureSummaryModelColumn durationMs=0\\n[plugins] [lcm] migration step complete: step=ensureMessageIdentityHashColumn durationMs=0\\n[plugins] [lcm] migration step complete: step=backfillMessageIdentityHashes durationMs=85\\n[plugins] [lcm] migration step complete: step=createMessagesIdentityHashIndex durationMs=0\\n[plugins] [lcm] migration step complete: step=ensureCompactionTelemetryColumns durationMs=0\\n[plugins] [lcm] migration step skipped: step=backfillSummaryDepths algorithmVersion=1 reason=already-complete\\n[plugins] [lcm] migration step complete: step=createSummariesDepthIndex durationMs=0\\n[plugins] [lcm] migration step skipped: step=backfillSummaryMetadata algorithmVersion=1 reason=already-complete\\n[plugins] [lcm] migration step skipped: step=backfillToolCallColumns algorithmVersion=1 reason=already-complete\\n[plugins] [lcm] migration step complete: step=ensureMessagesFts durationMs=1\\n[plugins] [lcm] migration step complete: step=ensureSummariesFts durationMs=0\\n[plugins] [lcm] migration step complete: step=ensureSummariesFtsCjk durationMs=1\\n[plugins] [lcm] Migration run completed during engine init: duration=108ms fts5=undefined\\n[plugins] [lcm] Engine initialized for db=/home/agent/.openclaw/lcm.db duration=111ms\\n[plugins] [lcm] Plugin loaded (enabled=true, db=/home/agent/.openclaw/lcm.db, threshold=0.8, proactiveThresholdCompactionMode=deferred)\\n[plugins] [lcm] State dir: /home/agent/.openclaw\\n[plugins] [lcm] Compaction summarization model: qwen3-6-80k-v1/qwen-3.6-agent (default)\\n[openclaw] Failed to start CLI: Error: Unknown command: openclaw gog. No built-in command or plugin CLI metadata owns \\\"gog\\\".\\n at runCli (file:///home/agent/.npm-global/lib/node_modules/openclaw/dist/cli/run-main.js:310:29)\\n at async runMainOrRootHelp (file:///home/agent/.npm-global/lib/node_modules/openclaw/dist/entry.js:468:3)\\n at async file:///home/agent/.npm-global/lib/node_modules/openclaw/dist/entry.js:438:55\", \"exit_code\": 1, \"error\": null}", "tool_call_id": "taPUsubXwDLMAZQtyglJZJFT2fHOqHYA", "timestamp": "2026-05-10T08:31:28.853389"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "JNDvzy6dQxt0vb27VQgGOjXw2BGQR4jB", "call_id": "JNDvzy6dQxt0vb27VQgGOjXw2BGQR4jB", "response_item_id": "fc_JNDvzy6dQxt0vb27VQgGOjXw2BGQR4jB", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"openclaw skills list\",\"timeout\":15}"}}], "timestamp": "2026-05-10T08:31:28.853389"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"Skills (20/63 ready)\\n┌─────────────┬──────────────────────────┬────────────────────────────────────────────────────────┬────────────────────┐\\n│ Status │ Skill │ Description │ Source │\\n├─────────────┼──────────────────────────┼────────────────────────────────────────────────────────┼────────────────────┤\\n│ ⏸ disabled │ 🔐 1password │ Set up and use 1Password CLI for sign-in, desktop │ openclaw-bundled │\\n│ │ │ integration, and reading or injecting secrets. │ │\\n│ ⏸ disabled │ 📝 apple-notes │ Create, view, edit, delete, search, move, or export │ openclaw-bundled │\\n│ │ │ Apple Notes via the memo CLI on macOS. │ │\\n│ ⏸ disabled │ ⏰ apple-reminders │ List, add, edit, complete, or delete Apple Reminders │ openclaw-bundled │\\n│ │ │ and reminder lists via remindctl. │ │\\n│ ⏸ disabled │ 🐻 bear-notes │ Create, search, and manage Bear notes via grizzly CLI. │ openclaw-bundled │\\n│ ⏸ disabled │ 📰 blogwatcher │ Monitor blogs and RSS/Atom feeds for updates using │ openclaw-bundled │\\n│ │ │ the blogwatcher CLI. │ │\\n│ ⏸ disabled │ 🫐 blucli │ BluOS CLI (blu) for discovery, playback, grouping, │ openclaw-bundled │\\n│ │ │ and volume. │ │\\n│ ⏸ disabled │ 🫧 bluebubbles │ Send and manage iMessages via BlueBubbles, including │ openclaw-bundled │\\n│ │ │ attachments, tapbacks, edits, replies, and groups. │ │\\n│ ✓ ready │ 📦 browser-act │ Browser automation CLI for AI agents with anti- │ openclaw-workspace │\\n│ │ │ detection stealth browsing, captcha solving, and │ │\\n│ │ │ parallel multi-browser support. Use when the user │ │\\n│ │ │ needs to interact with websites, including navigating │ │\\n│ │ │ pages, filling forms, clicking buttons, taking │ │\\n│ │ │ screenshots, extracting data, scraping sites with bot │ │\\n│ │ │ detection, or automating any browser task. Also use │ │\\n│ │ │ when the user needs to connect to their existing │ │\\n│ │ │ Chrome session, configure proxy-based stealth │ │\\n│ │ │ browsing, or run parallel browser sessions. Triggers │ │\\n│ │ │ on requests to open a website, fill out a form, click │ │\\n│ │ │ a button, take a screenshot, scrape data from a page, │ │\\n│ │ │ login to a site, automate browser actions, handle │ │\\n│ │ │ captcha challenges, or any task requiring │ │\\n│ │ │ programmatic web interaction. │ │\\n│ ✓ ready │ 📦 browser-automation │ Use when controlling web pages with the OpenClaw │ openclaw-extra │\\n│ │ │ browser tool, especially multi-step flows, login │ │\\n│ │ │ checks, tab management, or recovery from stale refs/ │ │\\n│ │ │ timeouts. │ │\\n│ ⏸ disabled │ 📸 camsnap │ Capture frames or clips from RTSP/ONVIF cameras. │ openclaw-bundled │\\n│ ⏸ disabled │ 📦 clawhub │ Search, install, update, sync, or publish agent │ openclaw-bundled │\\n│ │ │ skills with the ClawHub CLI and registry. │ │\\n│ ⏸ disabled │ 🧩 coding-agent │ Delegate coding tasks to Codex, Claude Code, │ openclaw-bundled │\\n│ │ │ OpenCode, or Pi agents via immediate background │ │\\n│ │ │ processes. Use when: (1) building or creating │ │\\n│ │ │ features/apps, (2) reviewing PRs in a temp clone/ │ │\\n│ │ │ worktree, (3) refactoring large codebases, (4) │ │\\n│ │ │ iterative coding that needs file exploration. NOT │ │\\n│ │ │ for: simple one-line fixes (just edit), reading code │ │\\n│ │ │ (use read tool), thread-bound ACP harness requests in │ │\\n│ │ │ chat (use sessions_spawn with runtime:\\\"acp\\\"), or any │ │\\n│ │ │ work in ~/clawd workspace (never spawn agents here). │ │\\n│ │ │ All coding-agent runs start with background:true │ │\\n│ │ │ immediately. Claude Code: use --print --permission- │ │\\n│ │ │ mode bypassPermissions (no PTY). Codex/Pi/OpenCode: │ │\\n│ │ │ pty:true required. Completion notification must use │ │\\n│ │ │ openclaw message send, not system event/heartbeat. │ │\\n│ ⏸ disabled │ 🎮 discord │ Discord ops via the message tool (channel=discord). │ openclaw-bundled │\\n│ ⏸ disabled │ 🛌 eightctl │ Control Eight Sleep pods (status, temperature, │ openclaw-bundled │\\n│ │ │ alarms, schedules). │ │\\n│ ⏸ disabled │ ✨ gemini │ Gemini CLI for one-shot Q&A, summaries, and │ openclaw-bundled │\\n│ │ │ generation. │ │\\n│ ⏸ disabled │ 📦 gh-issues │ Fetch GitHub issues, delegate fixes to subagents, │ openclaw-bundled │\\n│ │ │ open PRs, watch reviews, or run /gh-issues workflows. │ │\\n│ ⏸ disabled │ 🧲 gifgrep │ Search GIF providers with CLI/TUI, download results, │ openclaw-bundled │\\n│ │ │ and extract stills/sheets. │ │\\n│ ⏸ disabled │ 🐙 github │ Use gh for GitHub issues, PR status, CI/logs, │ openclaw-bundled │\\n│ │ │ comments, reviews, releases, and API queries. │ │\\n│ ✓ ready │ 🎮 gog │ Google Workspace CLI for Gmail, Calendar, Drive, │ openclaw-workspace │\\n│ │ │ Contacts, Sheets, and Docs. │ │\\n│ ⏸ disabled │ 📍 goplaces │ Query Google Places for text search, place details, │ openclaw-bundled │\\n│ │ │ resolve, reviews, or scriptable JSON via goplaces. │ │\\n│ ✓ ready │ 📦 healthcheck │ Audit and harden hosts running OpenClaw for SSH, │ openclaw-bundled │\\n│ │ │ firewall, updates, exposure, cron checks, and risk │ │\\n│ │ │ posture. │ │\\n│ ⏸ disabled │ 📧 himalaya │ Use himalaya to list, read, search, compose, reply, │ openclaw-bundled │\\n│ │ │ forward, and organize IMAP/SMTP email. │ │\\n│ ⏸ disabled │ 📨 imsg │ iMessage/SMS CLI for listing chats, history, and │ openclaw-bundled │\\n│ │ │ sending messages via Messages.app. │ │\\n│ ✓ ready │ 📦 lossless-claw │ Configure, diagnose, and use lossless-claw │ openclaw-extra │\\n│ │ │ effectively in OpenClaw, with emphasis on key │ │\\n│ │ │ settings, summary health, and recall-tool usage. │ │\\n│ ⏸ disabled │ 📦 mcporter │ List, configure, authenticate, call, and inspect MCP │ openclaw-bundled │\\n│ │ │ servers/tools with mcporter over HTTP or stdio. │ │\\n│ ⏸ disabled │ 📊 model-usage │ Summarize CodexBar local cost logs by model for Codex │ openclaw-bundled │\\n│ │ │ or Claude, including current or full breakdowns. │ │\\n│ ✓ ready │ 📦 multi-search-engine │ Multi search engine integration with 17 engines (8 CN │ openclaw-workspace │\\n│ │ │ + 9 Global). Supports advanced search operators, time │ │\\n│ │ │ filters, site search, privacy engines, and │ │\\n│ │ │ WolframAlpha knowledge queries. No API keys required. │ │\\n│ ⏸ disabled │ 📄 nano-pdf │ Edit PDFs with natural-language instructions using │ openclaw-bundled │\\n│ │ │ the nano-pdf CLI. │ │\\n│ ✓ ready │ 📦 node-connect │ Diagnose OpenClaw Android, iOS, or macOS node │ openclaw-bundled │\\n│ │ │ pairing, QR/setup code, route, auth, and connection │ │\\n│ │ │ failures. │ │\\n│ ⏸ disabled │ 📝 notion │ Notion API for creating and managing pages, │ openclaw-bundled │\\n│ │ │ databases, and blocks. │ │\\n│ ⏸ disabled │ 💎 obsidian │ Work with Obsidian vaults (plain Markdown notes) and │ openclaw-bundled │\\n│ │ │ automate via obsidian-cli. │ │\\n│ ✓ ready │ 🎙️ openai-whisper │ Local speech-to-text with the Whisper CLI (no API │ openclaw-workspace │\\n│ │ │ key). │ │\\n│ ⏸ disabled │ 🌐 openai-whisper-api │ Transcribe audio via OpenAI Audio Transcriptions API │ openclaw-bundled │\\n│ │ │ (Whisper). │ │\\n│ ⏸ disabled │ 💡 openhue │ Control Philips Hue lights and scenes via the OpenHue │ openclaw-bundled │\\n│ │ │ CLI. │ │\\n│ ⏸ disabled │ 🧿 oracle │ Use oracle CLI to bundle prompts and files for second- │ openclaw-bundled │\\n│ │ │ model debugging, refactor, design, or review checks. │ │\\n│ ⏸ disabled │ 🛵 ordercli │ Foodora-only CLI for checking past orders and active │ openclaw-bundled │\\n│ │ │ order status (Deliveroo WIP). │ │\\n│ ⏸ disabled │ 👀 peekaboo │ Capture and automate macOS UI with the Peekaboo CLI. │ openclaw-bundled │\\n│ ⏸ disabled │ 🔊 sag │ ElevenLabs text-to-speech with mac-style say UX. │ openclaw-bundled │\\n│ ✓ ready │ 📦 self-improvement │ Captures learnings, errors, and corrections to enable │ openclaw-workspace │\\n│ │ │ continuous improvement. Use when: (1) A command or │ │\\n│ │ │ operation fails unexpectedly, (2) User corrects │ │\\n│ │ │ Claude ('No, that's wrong...', 'Actually...'), (3) │ │\\n│ │ │ User requests a capability that doesn't exist, (4) An │ │\\n│ │ │ external API or tool fails, (5) Claude realizes its │ │\\n│ │ │ knowledge is outdated or incorrect, (6) A better │ │\\n│ │ │ approach is discovered for a recurring task. Also │ │\\n│ │ │ review learnings before major tasks. │ │\\n│ ⏸ disabled │ 📜 session-logs │ Search and analyze your own session logs (older/ │ openclaw-bundled │\\n│ │ │ parent conversations) using jq. │ │\\n│ ⏸ disabled │ 🔉 sherpa-onnx-tts │ Local text-to-speech via sherpa-onnx (offline, no │ openclaw-bundled │\\n│ │ │ cloud) │ │\\n│ ✓ ready │ 📦 skill-creator │ Create, edit, improve, tidy, review, audit, or │ openclaw-bundled │\\n│ │ │ restructure AgentSkills and SKILL.md files. │ │\\n│ ✓ ready │ 📦 skill-vetter │ Security-first skill vetting for AI agents. Use │ openclaw-workspace │\\n│ │ │ before installing any skill from ClawdHub, GitHub, or │ │\\n│ │ │ other sources. Checks for red flags, permission │ │\\n│ │ │ scope, and suspicious patterns. │ │\\n│ ⏸ disabled │ 💬 slack │ Use the Slack tool to react, pin/unpin, send, edit, │ openclaw-bundled │\\n│ │ │ delete messages, or fetch Slack member info. │ │\\n│ ⏸ disabled │ 🌊 songsee │ Generate spectrograms and feature-panel │ openclaw-bundled │\\n│ │ │ visualizations from audio with the songsee CLI. │ │\\n│ ⏸ disabled │ 🔊 sonoscli │ Control Sonos speakers (discover/status/play/volume/ │ openclaw-bundled │\\n│ │ │ group). │ │\\n│ ⏸ disabled │ 🎵 spotify-player │ Terminal Spotify playback/search via spogo │ openclaw-bundled │\\n│ │ │ (preferred) or spotify_player. │ │\\n│ ✓ ready │ 🧾 summarize │ Summarize URLs or files with the summarize CLI (web, │ openclaw-workspace │\\n│ │ │ PDFs, images, audio, YouTube). │ │\\n│ ✓ ready │ 🪝 taskflow │ Coordinate multi-step detached tasks as one durable │ openclaw-bundled │\\n│ │ │ TaskFlow job with owner context, state, waits, and │ │\\n│ │ │ child tasks. │ │\\n│ ✓ ready │ 📥 taskflow-inbox-triage │ Example TaskFlow pattern for inbox triage, intent │ openclaw-bundled │\\n│ │ │ routing, waiting on replies, and later summaries. │ │\\n│ ⏸ disabled │ 🔍 tavily │ Tavily web search, content extraction, and research │ openclaw-workspace │\\n│ │ │ tools. │ │\\n│ ✓ ready │ 📦 tavily-extract │ Extract clean markdown or text content from specific │ openclaw-workspace │\\n│ │ │ URLs via the Tavily CLI. Use this skill when the user │ │\\n│ │ │ has one or more URLs and wants their content, says │ │\\n│ │ │ \\\"extract\\\", \\\"grab the content from\\\", \\\"pull the text │ │\\n│ │ │ from\\\", \\\"get the page at\\\", \\\"read this webpage\\\", or │ │\\n│ │ │ needs clean text from web pages. Handles JavaScript- │ │\\n│ │ │ rendered pages, returns LLM-optimized markdown, and │ │\\n│ │ │ supports query-focused chunking for targeted │ │\\n│ │ │ extraction. Can process up to 20 URLs in a single │ │\\n│ │ │ call. │ │\\n│ ✓ ready │ 📦 tavily-search │ Search the web with LLM-optimized results via the │ openclaw-workspace │\\n│ │ │ Tavily CLI. Use this skill when the user wants to │ │\\n│ │ │ search the web, find articles, look up information, │ │\\n│ │ │ get recent news, discover sources, or says \\\"search │ │\\n│ │ │ for\\\", \\\"find me\\\", \\\"look up\\\", \\\"what's the latest on\\\", │ │\\n│ │ │ \\\"find articles about\\\", or needs current information │ │\\n│ │ │ from the internet. Returns relevant results with │ │\\n│ │ │ content snippets, relevance scores, and metadata — │ │\\n│ │ │ optimized for LLM consumption. Supports domain │ │\\n│ │ │ filtering, time ranges, and multiple search depths. │ │\\n│ ⏸ disabled │ ✅ things-mac │ Add, update, list, search, or inspect Things 3 todos, │ openclaw-bundled │\\n│ │ │ inbox, today, projects, areas, and tags on macOS. │ │\\n│ ⏸ disabled │ 🧵 tmux │ Remote-control tmux sessions for interactive CLIs by │ openclaw-bundled │\\n│ │ │ sending keystrokes and scraping pane output. │ │\\n│ ✓ ready │ ✅ todoist │ Manage tasks and projects in Todoist via REST API v1. │ openclaw-workspace │\\n│ │ │ Use when user asks about tasks, to-dos, reminders, or │ │\\n│ │ │ productivity. │ │\\n│ ✓ ready │ 📦 token-optimizer │ Reduce OpenClaw token usage and API costs through │ openclaw-workspace │\\n│ │ │ smart model routing, heartbeat optimization, budget │ │\\n│ │ │ tracking, and native 2026.2.15 features (session │ │\\n│ │ │ pruning, bootstrap size limits, cache TTL alignment). │ │\\n│ │ │ Use when token costs are high, API rate limits are │ │\\n│ │ │ being hit, or hosting multiple agents at scale. The 4 │ │\\n│ │ │ executable scripts (context_optimizer, model_router, │ │\\n│ │ │ heartbeat_optimizer, token_tracker) are local-only — │ │\\n│ │ │ no network requests, no subprocess calls, no system │ │\\n│ │ │ modifications. Reference files (PROVIDERS.md, config- │ │\\n│ │ │ patches.json) document optional multi-provider │ │\\n│ │ │ strategies that require external API keys and network │ │\\n│ │ │ access if you choose to use them. See SECURITY.md for │ │\\n│ │ │ full breakdown. │ │\\n│ ⏸ disabled │ 📋 trello │ Manage Trello boards, lists, and cards via the Trello │ openclaw-bundled │\\n│ │ │ REST API. │ │\\n│ ✓ ready │ 🎬 video-frames │ Extract frames or short clips from videos using │ openclaw-bundled │\\n│ │ │ ffmpeg. │ │\\n│ ⏸ disabled │ 📞 voice-call │ Start voice calls via the OpenClaw voice-call plugin. │ openclaw-bundled │\\n│ ⏸ disabled │ 📱 wacli │ Send third-party WhatsApp messages or sync/search │ openclaw-bundled │\\n│ │ │ WhatsApp history via wacli, not normal active chats. │ │\\n│ ✓ ready │ ☔ weather │ Get current weather, rain, temperature, and forecasts │ openclaw-bundled │\\n│ │ │ for locations or travel planning. │ │\\n│ ⏸ disabled │ 🐦 xurl │ Use xurl for authenticated X API posts, replies, │ openclaw-bundled │\\n│ │ │ search, DMs, media upload, followers, or raw v2 calls. │ │\\n└─────────────┴──────────────────────────┴────────────────────────────────────────────────────────┴────────────────────┘\\n\\nTip: use `openclaw skills search`, `openclaw skills install`, and `openclaw skills update` for ClawHub-backed skills.\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "JNDvzy6dQxt0vb27VQgGOjXw2BGQR4jB", "timestamp": "2026-05-10T08:31:28.853389"} {"role": "assistant", "content": "Gut, `gog` ist verfügbar. Lade die Skill-Datei für die korrekte Befehlsstruktur.", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "LfmkYwrQzCuipZ9tkChRd1OncgSa6fyE", "call_id": "LfmkYwrQzCuipZ9tkChRd1OncgSa6fyE", "response_item_id": "fc_LfmkYwrQzCuipZ9tkChRd1OncgSa6fyE", "type": "function", "function": {"name": "skill_view", "arguments": "{\"name\":\"gog\"}"}}], "timestamp": "2026-05-10T08:31:28.853389"} {"role": "tool", "name": "skill_view", "content": "{\"success\": true, \"name\": \"gog\", \"description\": \"Google Workspace CLI for Gmail, Calendar, Drive, Contacts, Sheets, and Docs.\", \"tags\": [], \"related_skills\": [], \"content\": \"---\\nname: gog\\ndescription: Google Workspace CLI for Gmail, Calendar, Drive, Contacts, Sheets, and Docs.\\nhomepage: https://gogcli.sh\\nmetadata: {\\\"clawdbot\\\":{\\\"emoji\\\":\\\"🎮\\\",\\\"requires\\\":{\\\"bins\\\":[\\\"gog\\\"]},\\\"install\\\":[{\\\"id\\\":\\\"brew\\\",\\\"kind\\\":\\\"brew\\\",\\\"formula\\\":\\\"steipete/tap/gogcli\\\",\\\"bins\\\":[\\\"gog\\\"],\\\"label\\\":\\\"Install gog (brew)\\\"}]}}\\n---\\n\\n# gog\\n\\nUse `gog` for Gmail/Calendar/Drive/Contacts/Sheets/Docs. Requires OAuth setup.\\n\\nSetup (once)\\n- `gog auth credentials /path/to/client_secret.json`\\n- `gog auth add you@gmail.com --services gmail,calendar,drive,contacts,sheets,docs`\\n- `gog auth list`\\n\\nCommon commands\\n- Gmail search: `gog gmail search 'newer_than:7d' --max 10`\\n- Gmail messages search: `gog gmail messages search 'is:unread' --max 10 --json` (message-level, not thread-level)\\n- Gmail get full message: `gog gmail get --json --format full` (returns body, attachments, headers)\\n- Gmail send: `gog gmail send --to a@b.com --subject \\\"Hi\\\" --body \\\"Hello\\\"`\\n- Gmail mark-read: `gog gmail mark-read `\\n- Gmail mark-unread: `gog gmail unread `\\n- Gmail trash: `gog gmail trash `\\n- Gmail forward: `gog gmail forward --to `\\n## Calendar Events — Create: `gog calendar create --from --to --summary --all-day`\\n\\n- **Command is `gog calendar create <calendarId>`, NOT `gog calendar events create`** — `events create` does not exist; `create` is a top-level subcommand under `calendar`.\\n- **Timed events require full RFC3339 with timezone offset** (e.g. `2026-05-10T14:00:00+02:00`). Plain `YYYY-MM-DDTHH:MM:SS` (no offset) fails with `Missing time zone definition for start time`.\\n- **All-day events**: Use date-only format (`YYYY-MM-DD`) + `--all-day` flag. No timezone needed.\\n- **Supported flags**: `--summary`, `--from`, `--to`, `--description`, `--location`, `--attendees`, `--all-day`, `--rrule`, `--reminder`, `--event-color`, `--visibility`, `--transparency`, `--with-meet`, `--send-updates`.\\n- **Event color IDs**: 1–11, use `gog calendar colors` to list.\\n- **Reminders**: Format `method:duration` (e.g. `popup:30m, email:1d`), max 5, can repeat.\\n\\n## Calendar Events — Query: `gog calendar events <calendarId> --from YYYY-MM-DD --to YYYY-MM-DD --json`\\n- Drive search: `gog drive search \\\"query\\\" --max 10`\\n- Contacts: `gog contacts list --max 20`\\n- Sheets get: `gog sheets get <sheetId> \\\"Tab!A1:D10\\\" --json`\\n- Sheets update: `gog sheets update <sheetId> \\\"Tab!A1:B2\\\" --values-json '[[\\\"A\\\",\\\"B\\\"],[\\\"1\\\",\\\"2\\\"]]' --input USER_ENTERED`\\n- Sheets append: `gog sheets append <sheetId> \\\"Tab!A:C\\\" --values-json '[[\\\"x\\\",\\\"y\\\",\\\"z\\\"]]' --insert INSERT_ROWS`\\n- Sheets clear: `gog sheets clear <sheetId> \\\"Tab!A2:Z\\\"`\\n- Sheets metadata: `gog sheets metadata <sheetId> --json`\\n- Docs export: `gog docs export <docId> --format txt --out /tmp/doc.txt`\\n- Docs cat: `gog docs cat <docId>`\\n\\n## Gmail — Pitfalls & Tips\\n\\n- **`gog gmail search` returns THREADS, not messages.** Use `gog gmail messages search` for message-level results. Thread IDs ≠ message IDs.\\n- **`gog gmail get <messageId>`** is the correct way to retrieve full message content (body, attachments, headers). The `--format full` flag returns complete JSON. Use `--format metadata` for headers only.\\n- **`--fields` is fragile**: The Google Calendar API rejects shorthand field names like `start`/`end`. Use full names (`startDateTime`, `endDateTime`) or skip `--fields` entirely and parse the default output. In practice, skipping `--fields` is often safer.\\n- **All-day events**: Show as date-only (`2026-05-07` → `2026-05-08`), not as timestamps. Treat these as spanning the full day.\\n- **Timezone handling**: Events from imported Outlook calendars may use UTC. Always convert to local time (Europe/Zurich) for display.\\n- **Multiple calendars**: Use `gog calendar calendars` to discover all available calendar IDs, then query each with `gog calendar events <calendarId> --tomorrow`.\\n- **`--tomorrow` flag**: Uses the system's timezone-aware \\\"tomorrow\\\". No need to construct ISO dates manually.\\n- **`gog calendar list` returns events, not calendars**: The `list` subcommand actually queries events (not calendar definitions). Use `gog calendar calendars` to list calendar metadata (IDs, summaries, access roles).\\n- **`gog calendar events` without calendarId queries ALL calendars**: `gog calendar events --from YYYY-MM-DD --to YYYY-MM-DD --json` returns events from all subscribed calendars. No calendar ID needed for multi-calendar queries.\\n- **Date format for `--from/--to`**: Use `YYYY-MM-DD` format (e.g., `2026-05-08`), NOT ISO timestamps. The CLI parses these as local dates.\\n- **`gog calendar list --max N` returns events**, not calendars. To list calendars: `gog calendar calendars --json`.\\n\\n## Drive File Operations\\n\\n### Move/Rename Files\\n- **Move to folder:** `gog drive move <fileId> --parent <new_folder_id>` — `--parent` is the REQUIRED flag (NOT `--to` or `--id`).\\n- **Download:** `gog drive download <fileId>` — saves to `~/.config/gogcli/` with auto-generated filename. Must manually rename and move to target directory.\\n- **Upload:** `gog drive upload <local_path> --parent <folder_id>`\\n- **Delete:** `gog drive delete <fileId>`\\n\\n### Listing Files\\n- **Root only:** `gog drive ls` (no flags)\\n- **Subfolder:** `gog drive ls --parent <folder_id> --json` — `--folder` does NOT exist, use `--parent`\\n- **JSON output:** Always use `--json` for scripting. Use `--plain` for stable parseable text.\\n\\n### Pitfalls\\n- **`GOG_KEYRING_PASSWORD`**: Must be set as env var for every gog command that needs auth. Without it: `no TTY available for keyring file backend password prompt`.\\n- **`--parent` not `--to`**: `gog drive move` uses `--parent <folder_id>`, NOT `--to`. `--to` is unknown flag.\\n- **Download saves to `~/.config/gogcli/`**: File has auto-generated name. Must rename manually.\\n- **`--json` for scripting**: Always use `--json` for machine-parseable output. `--plain` for stable text output.\\n- **`--no-input`**: Always pass to fail fast on interactive prompts in scripts.\\n\\n## Keyring / Authentication\\n\\n- Set `GOG_KEYRING_PASSWORD` as an environment variable to avoid TTY prompts when gog needs to read tokens from the keyring backend. Without it, gog will fail with `no TTY available for keyring file backend password prompt`.\\n- For scripting, always pass `--no-input` to fail fast instead of hanging on interactive prompts.\\n\\nNotes\\n- Set `GOG_ACCOUNT=you@gmail.com` to avoid repeating `--account`.\\n- For scripting, prefer `--json` plus `--no-input`.\\n- Sheets values can be passed via `--values-json` (recommended) or as inline rows.\\n- Docs supports export/cat/copy. In-place edits require a Docs API client (not in gog).\\n- Confirm before sending mail or creating events.\\n\\n## References\\n\\n- `references/gmail-command-reference.md` — Full Gmail command reference (search, get, messages, attachments)\\n- `references/gmail-inbox-review-pattern.md` — Cron-job inbox review workflow and output format\\n- `references/calendar-command-quirks.md` — Calendar query quirks (list vs calendars, date formats)\\n- `references/calendar-create-event-notes.md` — Calendar event creation: pitfalls, flags, event types\\n\", \"path\": \"openclaw-imports/gog/SKILL.md\", \"skill_dir\": \"/home/agent/.hermes/skills/openclaw-imports/gog\", \"linked_files\": {\"references\": [\"references/calendar-command-quirks.md\", \"references/calendar-create-event-notes.md\", \"references/gmail-inbox-review-pattern.md\", \"references/gmail-command-reference.md\"]}, \"usage_hint\": \"To view linked files, call skill_view(name, file_path) where file_path is e.g. 'references/api.md' or 'assets/config.yaml'\", \"required_environment_variables\": [], \"required_commands\": [], \"missing_required_environment_variables\": [], \"missing_credential_files\": [], \"missing_required_commands\": [], \"setup_needed\": false, \"setup_skipped\": false, \"readiness_status\": \"available\", \"metadata\": {\"clawdbot\": {\"emoji\": \"🎮\", \"requires\": {\"bins\": [\"gog\"]}, \"install\": [{\"id\": \"brew\", \"kind\": \"brew\", \"formula\": \"steipete/tap/gogcli\", \"bins\": [\"gog\"], \"label\": \"Install gog (brew)\"}]}}}", "tool_call_id": "LfmkYwrQzCuipZ9tkChRd1OncgSa6fyE", "timestamp": "2026-05-10T08:31:28.853389"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "QLzpwTdphSGddC3RAToMMIencgbc4Q49", "call_id": "QLzpwTdphSGddC3RAToMMIencgbc4Q49", "response_item_id": "fc_QLzpwTdphSGddC3RAToMMIencgbc4Q49", "type": "function", "function": {"name": "skill_view", "arguments": "{\"name\":\"gog\",\"file_path\":\"references/gmail-command-reference.md\"}"}}], "timestamp": "2026-05-10T08:31:28.853389"} {"role": "tool", "name": "skill_view", "content": "{\"success\": true, \"name\": \"gog\", \"file\": \"references/gmail-command-reference.md\", \"content\": \"# gog Gmail Command Reference\\n\\n## Search Commands\\n\\n### Thread-level search (default)\\n```bash\\ngog gmail search 'is:unread' --max 10 --json\\n```\\nReturns thread objects with keys: `id`, `date`, `from`, `subject`, `labels`, `messageCount`.\\n\\n### Message-level search\\n```bash\\ngog gmail messages search 'is:unread' --max 10 --json\\n```\\nReturns message objects with keys: `id`, `threadId`, `date`, `from`, `subject`, `labels`.\\n**Use this when you need message IDs** (thread IDs are NOT message IDs).\\n\\n### UID-based search (does NOT work)\\n```bash\\ngog gmail search 'uid:<messageId>' --max 1 --json\\n```\\nUID queries return empty. Do not use.\\n\\n## Retrieve Full Message Content\\n\\n```bash\\ngog gmail get <messageId> --json --format full\\n```\\n\\nReturns full JSON with:\\n- `body`: plain text body\\n- `attachments`: array of `{filename, size, sizeHuman, mimeType, attachmentId}`\\n- `headers`: full email headers\\n- `message`: full Gmail API message object\\n\\nFor metadata only:\\n```bash\\ngog gmail get <messageId> --json --format metadata --headers \\\"From,To,Subject,Date\\\"\\n```\\n\\n## Label Management\\n\\n```bash\\ngog gmail mark-read <messageId>\\ngog gmail unread <messageId>\\ngog gmail trash <messageId>\\ngog gmail modify <messageId> --add-labels \\\"IMPORTANT\\\"\\ngog gmail modify <messageId> --remove-labels \\\"UNREAD\\\"\\n```\\n\\n## Send/Reply\\n\\n```bash\\ngog gmail send --to recipient@example.com --subject \\\"Hi\\\" --body \\\"Hello\\\"\\ngog gmail forward --to newrecipient@example.com <messageId>\\n```\\n\\n## Select Fields\\n\\n```bash\\ngog gmail messages search 'is:unread' --max 2 --select body,attachments --json\\n```\\nNote: field selection may return empty `{}` if field names don't match expected keys. Full `get` is more reliable.\\n\", \"file_type\": \".md\"}", "tool_call_id": "QLzpwTdphSGddC3RAToMMIencgbc4Q49", "timestamp": "2026-05-10T08:31:28.853389"} {"role": "assistant", "content": "Gut. Ich sehe: Es gibt keinen direkten \"Archive\"-Befehl in gog, aber \"Archivieren\" in Gmail bedeutet einfach das Entfernen des `Inbox`-Labels. Zuerst muss ich das \"Archiv\"-Label erstellen und dann die beiden Emails finden und archivieren.\n\nAber warte – der User sagt \"diese beiden emails\", ohne Emails anzuhängen oder zu verlinken. Ich muss nachfragen, welche Emails gemeint sind.", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "05jIkPGQqVG3XhM98xxAxth0RTGq7awM", "call_id": "05jIkPGQqVG3XhM98xxAxth0RTGq7awM", "response_item_id": "fc_05jIkPGQqVG3XhM98xxAxth0RTGq7awM", "type": "function", "function": {"name": "clarify", "arguments": "{\"question\":\"Welche zwei E-Mails sollen archiviert werden? Ich sehe keine angehängten Emails in deiner Nachricht. Kannst du mir entweder:\\n\\n1. Die **Betreffzeilen** der Emails nennen, oder\\n2. Die **Message-IDs** oder **Thread-IDs** angeben, oder\\n3. Die Emails als **Screenshot/Anhang** schicken?\"}"}}], "timestamp": "2026-05-10T08:31:28.853389"} {"role": "tool", "name": "clarify", "content": "{\"error\": \"Clarify tool is not available in this execution context.\"}", "tool_call_id": "05jIkPGQqVG3XhM98xxAxth0RTGq7awM", "timestamp": "2026-05-10T08:31:28.853389"} {"role": "assistant", "content": "Ich benötige eine Klärung, da keine E-Mails angehängt sind:\n\nWelche zwei E-Mails sollen archiviert werden? Ich sehe keine angehängten Emails oder Links in deiner Nachricht.\n\nKannst du mir bitte:\n- Die **Betreffzeilen** der Emails nennen, oder\n- Die **Message-IDs** angeben, oder\n- Die Emails als **Screenshot/Anhang** schicken?", "reasoning": null, "finish_reason": "stop", "timestamp": "2026-05-10T08:31:28.853389"} {"role": "user", "content": "Die emails liegen in deiner Gmail-inbox, du hast zugriff auf Friday.uplink@gmail.com via gog CLI.", "timestamp": "2026-05-10T08:50:06.410187"} {"role": "assistant", "content": "Ich werde die neuesten Emails in der Inbox durchsuchen, um die beiden gemeinten zu finden.", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "LfDgSrE32FPzLemgXQboqTXJsi6e3zXi", "call_id": "LfDgSrE32FPzLemgXQboqTXJsi6e3zXi", "response_item_id": "fc_LfDgSrE32FPzLemgXQboqTXJsi6e3zXi", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"GOG_KEYRING_PASSWORD=LammBitch30 gog gmail search 'is:inbox' --max 5 --json\",\"timeout\":30}"}}], "timestamp": "2026-05-10T08:50:06.410187"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"{\\n \\\"nextPageToken\\\": \\\"\\\",\\n \\\"threads\\\": [\\n {\\n \\\"id\\\": \\\"19e0ec0bf4132285\\\",\\n \\\"date\\\": \\\"2026-05-09 23:59\\\",\\n \\\"from\\\": \\\"Google <no-reply@accounts.google.com>\\\",\\n \\\"subject\\\": \\\"Validation en deux étapes activée\\\",\\n \\\"labels\\\": [\\n \\\"UNREAD\\\",\\n \\\"INBOX\\\"\\n ],\\n \\\"messageCount\\\": 1\\n },\\n {\\n \\\"id\\\": \\\"19e0ec08b22d1ef3\\\",\\n \\\"date\\\": \\\"2026-05-09 23:59\\\",\\n \\\"from\\\": \\\"Google <no-reply@accounts.google.com>\\\",\\n \\\"subject\\\": \\\"Alerte de sécurité\\\",\\n \\\"labels\\\": [\\n \\\"UNREAD\\\",\\n \\\"INBOX\\\"\\n ],\\n \\\"messageCount\\\": 1\\n }\\n ]\\n}\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "LfDgSrE32FPzLemgXQboqTXJsi6e3zXi", "timestamp": "2026-05-10T08:50:06.410187"} {"role": "assistant", "content": "Zwei ungelesene E-Mails in der Inbox. Ich hole den vollständigen Inhalt, um sicherzugehen.", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "m4kRNHHp0CpzTmRVBiFHDySsH68fS1jw", "call_id": "m4kRNHHp0CpzTmRVBiFHDySsH68fS1jw", "response_item_id": "fc_m4kRNHHp0CpzTmRVBiFHDySsH68fS1jw", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"GOG_KEYRING_PASSWORD=LammBitch30 gog gmail messages search 'is:inbox' --max 2 --json\",\"timeout\":30}"}}], "timestamp": "2026-05-10T08:50:06.410187"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"{\\n \\\"messages\\\": [\\n {\\n \\\"id\\\": \\\"19e0ec0bf4132285\\\",\\n \\\"threadId\\\": \\\"19e0ec0bf4132285\\\",\\n \\\"date\\\": \\\"2026-05-09 23:59\\\",\\n \\\"from\\\": \\\"Google <no-reply@accounts.google.com>\\\",\\n \\\"subject\\\": \\\"Validation en deux étapes activée\\\",\\n \\\"labels\\\": [\\n \\\"UNREAD\\\",\\n \\\"INBOX\\\"\\n ]\\n },\\n {\\n \\\"id\\\": \\\"19e0ec08b22d1ef3\\\",\\n \\\"threadId\\\": \\\"19e0ec08b22d1ef3\\\",\\n \\\"date\\\": \\\"2026-05-09 23:59\\\",\\n \\\"from\\\": \\\"Google <no-reply@accounts.google.com>\\\",\\n \\\"subject\\\": \\\"Alerte de sécurité\\\",\\n \\\"labels\\\": [\\n \\\"UNREAD\\\",\\n \\\"INBOX\\\"\\n ]\\n }\\n ],\\n \\\"nextPageToken\\\": \\\"14931669422986794256\\\"\\n}\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "m4kRNHHp0CpzTmRVBiFHDySsH68fS1jw", "timestamp": "2026-05-10T08:50:06.410187"} {"role": "assistant", "content": "Ich sehe zwei ungelesene E-Mails in der Inbox. Ich erstelle zuerst das \"Archiv\"-Label und verschiebe dann beide Emails.\n\n**E-Mails in der Inbox:**\n1. `19e0ec0bf4132285` – \"Validation en deux étapes activée\" (2FA aktiviert)\n2. `19e0ec08b22d1ef3` – \"Alerte de sécurité\" (Security alert)", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "vmxQ51GPnP1Z57mTcfTkeTvzZjzC3WZc", "call_id": "vmxQ51GPnP1Z57mTcfTkeTvzZjzC3WZc", "response_item_id": "fc_vmxQ51GPnP1Z57mTcfTkeTvzZjzC3WZc", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"GOG_KEYRING_PASSWORD=LammBitch30 gog gmail modify --add-labels \\\"Archiv\\\" --json 19e0ec0bf4132285 2>&1 || echo \\\"---\\\"; GOG_KEYRING_PASSWORD=LammBitch30 gog gmail modify --add-labels \\\"Archiv\\\" --json 19e0ec08b22d1ef3 2>&1\",\"timeout\":30}"}}], "timestamp": "2026-05-10T08:50:06.410187"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"unexpected argument modify\\n---\\nunexpected argument modify\", \"exit_code\": 2, \"error\": null}", "tool_call_id": "vmxQ51GPnP1Z57mTcfTkeTvzZjzC3WZc", "timestamp": "2026-05-10T08:50:06.410187"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "l5Pn1pvMz4ryOU6tDRHaDIuwk9pGCT6Q", "call_id": "l5Pn1pvMz4ryOU6tDRHaDIuwk9pGCT6Q", "response_item_id": "fc_l5Pn1pvMz4ryOU6tDRHaDIuwk9pGCT6Q", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"GOG_KEYRING_PASSWORD=LammBitch30 gog gmail --help 2>&1 | head -40\",\"timeout\":15}"}}], "timestamp": "2026-05-10T08:50:06.410187"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"Usage: gog gmail (mail,email) <command> [flags]\\nBuild: v0.13.0 (5cd913e)\\n\\nGmail\\n\\nFlags:\\n -h, --help Show context-sensitive help.\\n --color=\\\"auto\\\" Color output: auto|always|never\\n -a, --account=STRING Account email for API commands\\n (gmail/calendar/chat/classroom/drive/docs/slides/contacts/tasks/people/sheets/forms/appscript/ads)\\n --client=\\\"\\\" OAuth client name (selects stored credentials +\\n token bucket)\\n --access-token=STRING Use provided access token directly (bypasses\\n stored refresh tokens; token expires in ~1h)\\n ($GOG_ACCESS_TOKEN)\\n --enable-commands=\\\"\\\" Comma-separated list of enabled commands;\\n dot paths allowed (restricts CLI)\\n --disable-commands=\\\"\\\" Comma-separated list of disabled commands;\\n dot paths allowed\\n --gmail-no-send Block Gmail send operations (agent safety)\\n -j, --json Output JSON to stdout (best for scripting)\\n -p, --plain Output stable, parseable text to stdout (TSV;\\n no colors)\\n --results-only In JSON mode, emit only the primary result (drops\\n envelope fields like nextPageToken)\\n --select=STRING In JSON mode, select comma-separated fields\\n (best-effort; supports dot paths). Desire path:\\n use --fields for most commands.\\n -n, --dry-run Do not make changes; print intended actions and\\n exit successfully\\n -y, --force Skip confirmations for destructive commands\\n --no-input Never prompt; fail instead (useful for CI)\\n -v, --verbose Enable verbose logging\\n --version Print version and exit\\n\\nCommands:\\n\\nRead\\n search (find,query,ls,list) <query> ... [flags]\\n Search threads using Gmail query syntax\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "l5Pn1pvMz4ryOU6tDRHaDIuwk9pGCT6Q", "timestamp": "2026-05-10T08:50:06.410187"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "ipwpccHIqkWs07AGHjO1Y7fRa7tuyG1b", "call_id": "ipwpccHIqkWs07AGHjO1Y7fRa7tuyG1b", "response_item_id": "fc_ipwpccHIqkWs07AGHjO1Y7fRa7tuyG1b", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"GOG_KEYRING_PASSWORD=LammBitch30 gog gmail --help 2>&1 | grep -A 2 \\\"Write\\\\|Modify\\\\|modify\\\\|label\\\\|Label\\\\|archive\\\\|Archive\\\"\",\"timeout\":15}"}}], "timestamp": "2026-05-10T08:50:06.410187"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"Thread operations (get, modify)\\n\\n labels (label) <command>\\n Label operations\\n\\n batch <command>\\n--\\n archive [<messageId> ...] [flags]\\n Archive messages (remove from inbox)\\n\\n mark-read (read-messages) [<messageId> ...] [flags]\\n--\\nWrite\\n send [flags]\\n Send an email\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "ipwpccHIqkWs07AGHjO1Y7fRa7tuyG1b", "timestamp": "2026-05-10T08:50:06.410187"} {"role": "assistant", "content": "Perfekt, es gibt einen `archive` Befehl. Und ich muss auch das \"Archiv\"-Label erstellen.", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "CnKVSCb1Xk9Q6DCDwj68LLfgn3cxe6GR", "call_id": "CnKVSCb1Xk9Q6DCDwj68LLfgn3cxe6GR", "response_item_id": "fc_CnKVSCb1Xk9Q6DCDwj68LLfgn3cxe6GR", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"GOG_KEYRING_PASSWORD=LammBitch30 gog gmail labels create \\\"Archiv\\\" 2>&1\",\"timeout\":15}"}}], "timestamp": "2026-05-10T08:50:06.410187"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"Created label: Archiv (id: Label_2)\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "CnKVSCb1Xk9Q6DCDwj68LLfgn3cxe6GR", "timestamp": "2026-05-10T08:50:06.410187"} {"role": "assistant", "content": "Label erstellt. Jetzt die beiden Emails archivieren:", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "BYbmjmgBmXywBE4h1tYsCTsQyCYzqrAE", "call_id": "BYbmjmgBmXywBE4h1tYsCTsQyCYzqrAE", "response_item_id": "fc_BYbmjmgBmXywBE4h1tYsCTsQyCYzqrAE", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"GOG_KEYRING_PASSWORD=LammBitch30 gog gmail archive 19e0ec0bf4132285 19e0ec08b22d1ef3 2>&1\",\"timeout\":15}"}}], "timestamp": "2026-05-10T08:50:06.410187"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"Archived 2 messages\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "BYbmjmgBmXywBE4h1tYsCTsQyCYzqrAE", "timestamp": "2026-05-10T08:50:06.410187"} {"role": "assistant", "content": "Erledigt.\n\n**Zusammenfassung:**\n- ✅ Label `Archiv` (Label_2) erstellt\n- ✅ 2 E-Mails archiviert (aus der Inbox entfernt):\n 1. \"Validation en deux étapes activée\" – Google 2FA aktiviert\n 2. \"Alerte de sécurité\" – Google Security Alert", "reasoning": null, "finish_reason": "stop", "timestamp": "2026-05-10T08:50:06.410187"} {"role": "user", "content": "Sehr gut. Nun optimieren wir das OCR. Was hälst du von dieser Lösung: https://github.com/datalab-to/marker", "timestamp": "2026-05-10T08:52:16.084721"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "XRWAWB6lNQ9A5DfQDDF8L5DTmAIS9fTd", "call_id": "XRWAWB6lNQ9A5DfQDDF8L5DTmAIS9fTd", "response_item_id": "fc_XRWAWB6lNQ9A5DfQDDF8L5DTmAIS9fTd", "type": "function", "function": {"name": "skill_view", "arguments": "{\"name\":\"ocr-and-documents\"}"}}], "timestamp": "2026-05-10T08:52:16.084721"} {"role": "tool", "name": "skill_view", "content": "{\"success\": true, \"name\": \"ocr-and-documents\", \"description\": \"Extract text from PDFs/scans (pymupdf, marker-pdf).\", \"tags\": [\"PDF\", \"Documents\", \"Research\", \"Arxiv\", \"Text-Extraction\", \"OCR\"], \"related_skills\": [\"powerpoint\"], \"content\": \"---\\nname: ocr-and-documents\\ndescription: \\\"Extract text from PDFs/scans (pymupdf, marker-pdf).\\\"\\nversion: 2.3.0\\nauthor: Hermes Agent\\nlicense: MIT\\nmetadata:\\n hermes:\\n tags: [PDF, Documents, Research, Arxiv, Text-Extraction, OCR]\\n related_skills: [powerpoint]\\n---\\n\\n# PDF & Document Extraction\\n\\nFor DOCX: use `python-docx` (parses actual document structure, far better than OCR).\\nFor PPTX: see the `powerpoint` skill (uses `python-pptx` with full slide/notes support).\\nThis skill covers **PDFs and scanned documents**.\\n\\n## Step 1: Remote URL Available?\\n\\nIf the document has a URL, **always try `web_extract` first**:\\n\\n```\\nweb_extract(urls=[\\\"https://arxiv.org/pdf/2402.03300\\\"])\\nweb_extract(urls=[\\\"https://example.com/report.pdf\\\"])\\n```\\n\\nThis handles PDF-to-markdown conversion via Firecrawl with no local dependencies.\\n\\nOnly use local extraction when: the file is local, web_extract fails, or you need batch processing.\\n\\n## Step 2: Choose Local Extractor\\n\\n| Feature | pymupdf (~25MB) | marker-pdf (~3-5GB) |\\n|---------|-----------------|---------------------|\\n| **Text-based PDF** | ✅ | ✅ |\\n| **Scanned PDF (OCR)** | ❌ | ✅ (90+ languages) |\\n| **Tables** | ✅ (basic) | ✅ (high accuracy) |\\n| **Equations / LaTeX** | ❌ | ✅ |\\n| **Code blocks** | ❌ | ✅ |\\n| **Forms** | ❌ | ✅ |\\n| **Headers/footers removal** | ❌ | ✅ |\\n| **Reading order detection** | ❌ | ✅ |\\n| **Images extraction** | ✅ (embedded) | ✅ (with context) |\\n| **Images → text (OCR)** | ❌ | ✅ |\\n| **EPUB** | ✅ | ✅ |\\n| **Markdown output** | ✅ (via pymupdf4llm) | ✅ (native, higher quality) |\\n| **Install size** | ~25MB | ~3-5GB (PyTorch + models) |\\n| **Speed** | Instant | ~1-14s/page (CPU), ~0.2s/page (GPU) |\\n\\n**Decision**: Use pymupdf unless you need OCR, equations, forms, or complex layout analysis.\\n\\n**For medical/laboratory PDFs with complex tables** (Blutbilder, Laborberichte): Use **camelot** (`pip install camelot-py[cv]`) — it extracts table structures with high accuracy, far superior to pymupdf for tabular data.\\n\\n**For scanned medical documents**: Use **PyMuPDF (fitz)** for text extraction; combine with **pytesseract** for pure image-based PDFs.\\n\\nIf the user needs marker capabilities but the system lacks ~5GB free disk:\\n> \\\"This document needs OCR/advanced extraction (marker-pdf), which requires ~5GB for PyTorch and models. Your system has [X]GB free. Options: free up space, provide a URL so I can use web_extract, or I can try pymupdf which works for text-based PDFs but not scanned documents or equations.\\\"\\n\\n---\\n\\n## Tesseract OCR (for scanned PDFs)\\n\\n```bash\\n# Check available languages\\ntesseract --list-langs\\n\\n# Basic OCR\\ntesseract input.png output -l deu+eng\\n\\n# For mixed text+tables (laboratory reports):\\ntesseract input.png output -l deu+eng --psm 6 --oem 3\\n# PSM=6: Assume a single uniform block of text (BEST for tabular data)\\n# PSM=4: Single column (good for structured forms)\\n# PSM=11: Sparse text (words as separate lines — good for extracting individual values)\\n# PSM=13: Raw line (fastest, lowest accuracy)\\n\\n# For scanned PDFs: first convert to PNG, then OCR\\n# Using PyMuPDF: convert page to image at 300 DPI for best OCR quality\\n```\\n\\n**Pitfall:** Tesseract OCR on tabular PDFs often produces fragmented output. Use PSM=6 for mixed content, PSM=11 when you need individual values as separate tokens. For medical lab reports with obscured values in the PDF itself, OCR may not recover hidden data.\\n\\n**Python environment note:** PyMuPDF (fitz) and pytesseract are installed in `~/.local/lib/python3.12/site-packages`, NOT in the hermes-agent venv. Always set `PYTHONPATH=~/.local/lib/python3.12/site-packages` when running Python scripts that import these packages.\\n\\n## camelot (PDF Table Extraction — Elite)\\n\\n```bash\\npip install camelot-py[cv]\\n```\\n\\n**Required:** Ghostscript installed (`apt install ghostscript`) — camelot depends on it.\\n\\n```python\\nimport camelot\\ntables = camelot.read_pdf('laborbericht.pdf', flavor='lattice') # Grid-based tables\\n# OR\\ntables = camelot.read_pdf('laborbericht.pdf', flavor='stream') # Space-separated tables\\n\\n# Extract to DataFrame\\ndf = tables[0].df # First table as DataFrame\\ndf.to_csv('output.csv') # Save to CSV\\n```\\n\\n**Decision:** Use camelot for laboratory reports, medical charts, and any PDF with structured tabular data. It preserves column alignment and cell structure far better than pymupdf.\\n\\n**Pitfall:** camelot requires Ghostscript. If `gs` is not found, install with `apt install ghostscript`. Without it, camelot will silently fail or raise `GhostscriptNotFoundError`.\\n\\n**Pitfall:** For scanned PDFs, camelot may not work — use PyMuPDF + pytesseract for OCR instead.\\n\\n---\\n\\n## pymupdf (lightweight)\\n\\n```bash\\npip install pymupdf pymupdf4llm\\n```\\n\\n**Via helper script**:\\n```bash\\npython scripts/extract_pymupdf.py document.pdf # Plain text\\npython scripts/extract_pymupdf.py document.pdf --markdown # Markdown\\npython scripts/extract_pymupdf.py document.pdf --tables # Tables\\npython scripts/extract_pymupdf.py document.pdf --images out/ # Extract images\\npython scripts/extract_pymupdf.py document.pdf --metadata # Title, author, pages\\npython scripts/extract_pymupdf.py document.pdf --pages 0-4 # Specific pages\\n```\\n\\n**Inline**:\\n```bash\\npython3 -c \\\"\\nimport pymupdf\\ndoc = pymupdf.open('document.pdf')\\nfor page in doc:\\n print(page.get_text())\\n\\\"\\n```\\n\\n---\\n\\n## marker-pdf (high-quality OCR)\\n\\n```bash\\n# Check disk space first\\npython scripts/extract_marker.py --check\\n\\npip install marker-pdf\\n```\\n\\n**Via helper script**:\\n```bash\\npython scripts/extract_marker.py document.pdf # Markdown\\npython scripts/extract_marker.py document.pdf --json # JSON with metadata\\npython scripts/extract_marker.py document.pdf --output_dir out/ # Save images\\npython scripts/extract_marker.py scanned.pdf # Scanned PDF (OCR)\\npython scripts/extract_marker.py document.pdf --use_llm # LLM-boosted accuracy\\n```\\n\\n**CLI** (installed with marker-pdf):\\n```bash\\nmarker_single document.pdf --output_dir ./output\\nmarker /path/to/folder --workers 4 # Batch\\n```\\n\\n---\\n\\n## Arxiv Papers\\n\\n```\\n# Abstract only (fast)\\nweb_extract(urls=[\\\"https://arxiv.org/abs/2402.03300\\\"])\\n\\n# Full paper\\nweb_extract(urls=[\\\"https://arxiv.org/pdf/2402.03300\\\"])\\n\\n# Search\\nweb_search(query=\\\"arxiv GRPO reinforcement learning 2026\\\")\\n```\\n\\n## Split, Merge & Search\\n\\npymupdf handles these natively — use `execute_code` or inline Python:\\n\\n```python\\n# Split: extract pages 1-5 to a new PDF\\nimport pymupdf\\ndoc = pymupdf.open(\\\"report.pdf\\\")\\nnew = pymupdf.open()\\nfor i in range(5):\\n new.insert_pdf(doc, from_page=i, to_page=i)\\nnew.save(\\\"pages_1-5.pdf\\\")\\n```\\n\\n```python\\n# Merge multiple PDFs\\nimport pymupdf\\nresult = pymupdf.open()\\nfor path in [\\\"a.pdf\\\", \\\"b.pdf\\\", \\\"c.pdf\\\"]:\\n result.insert_pdf(pymupdf.open(path))\\nresult.save(\\\"merged.pdf\\\")\\n```\\n\\n```python\\n# Search for text across all pages\\nimport pymupdf\\ndoc = pymupdf.open(\\\"report.pdf\\\")\\nfor i, page in enumerate(doc):\\n results = page.search_for(\\\"revenue\\\")\\n if results:\\n print(f\\\"Page {i+1}: {len(results)} match(es)\\\")\\n print(page.get_text(\\\"text\\\"))\\n```\\n\\nNo extra dependencies needed — pymupdf covers split, merge, search, and text extraction in one package.\\n\\n---\\n\\n## Notes\\n\\n- `web_extract` is always first choice for URLs\\n- pymupdf is the safe default — instant, no models, works everywhere\\n- marker-pdf is for OCR, scanned docs, equations, complex layouts — install only when needed\\n- Both helper scripts accept `--help` for full usage\\n- marker-pdf downloads ~2.5GB of models to `~/.cache/huggingface/` on first use\\n- For medical/laboratory PDFs: **camelot** is preferred over pymupdf for table extraction (see below)\\n- For Word docs: `pip install python-docx` (better than OCR — parses actual structure)\\n- For PowerPoint: see the `powerpoint` skill (uses python-pptx)\\n- For PDF tables: `camelot-py[cv]` requires Ghostscript (`apt install ghostscript`)\\n\\n## References\\n\\n- `references/tesseract-lab-report-ocr.md` — Session-specific OCR patterns for medical lab reports (PSM modes, Python env, obscured values)\\n\", \"path\": \"productivity/ocr-and-documents/SKILL.md\", \"skill_dir\": \"/home/agent/.hermes/skills/productivity/ocr-and-documents\", \"linked_files\": {\"references\": [\"references/tesseract-lab-report-ocr.md\"], \"scripts\": [\"scripts/extract_pymupdf.py\", \"scripts/extract_marker.py\"]}, \"usage_hint\": \"To view linked files, call skill_view(name, file_path) where file_path is e.g. 'references/api.md' or 'assets/config.yaml'\", \"required_environment_variables\": [], \"required_commands\": [], \"missing_required_environment_variables\": [], \"missing_credential_files\": [], \"missing_required_commands\": [], \"setup_needed\": false, \"setup_skipped\": false, \"readiness_status\": \"available\", \"metadata\": {\"hermes\": {\"tags\": [\"PDF\", \"Documents\", \"Research\", \"Arxiv\", \"Text-Extraction\", \"OCR\"], \"related_skills\": [\"powerpoint\"]}}}", "tool_call_id": "XRWAWB6lNQ9A5DfQDDF8L5DTmAIS9fTd", "timestamp": "2026-05-10T08:52:16.084721"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "sEMJzP97RvkTPAC0giPqhUevXtEs8bsp", "call_id": "sEMJzP97RvkTPAC0giPqhUevXtEs8bsp", "response_item_id": "fc_sEMJzP97RvkTPAC0giPqhUevXtEs8bsp", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"curl -s https://api.github.com/repos/datalab-to/marker 2>&1 | head -80\",\"timeout\":15}"}}], "timestamp": "2026-05-10T08:52:16.084721"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"{\\n \\\"id\\\": 712111618,\\n \\\"node_id\\\": \\\"R_kgDOKnH2Ag\\\",\\n \\\"name\\\": \\\"marker\\\",\\n \\\"full_name\\\": \\\"datalab-to/marker\\\",\\n \\\"private\\\": false,\\n \\\"owner\\\": {\\n \\\"login\\\": \\\"datalab-to\\\",\\n \\\"id\\\": 176631703,\\n \\\"node_id\\\": \\\"O_kgDOCocvlw\\\",\\n \\\"avatar_url\\\": \\\"https://avatars.githubusercontent.com/u/176631703?v=4\\\",\\n \\\"gravatar_id\\\": \\\"\\\",\\n \\\"url\\\": \\\"https://api.github.com/users/datalab-to\\\",\\n \\\"html_url\\\": \\\"https://github.com/datalab-to\\\",\\n \\\"followers_url\\\": \\\"https://api.github.com/users/datalab-to/followers\\\",\\n \\\"following_url\\\": \\\"https://api.github.com/users/datalab-to/following{/other_user}\\\",\\n \\\"gists_url\\\": \\\"https://api.github.com/users/datalab-to/gists{/gist_id}\\\",\\n \\\"starred_url\\\": \\\"https://api.github.com/users/datalab-to/starred{/owner}{/repo}\\\",\\n \\\"subscriptions_url\\\": \\\"https://api.github.com/users/datalab-to/subscriptions\\\",\\n \\\"organizations_url\\\": \\\"https://api.github.com/users/datalab-to/orgs\\\",\\n \\\"repos_url\\\": \\\"https://api.github.com/users/datalab-to/repos\\\",\\n \\\"events_url\\\": \\\"https://api.github.com/users/datalab-to/events{/privacy}\\\",\\n \\\"received_events_url\\\": \\\"https://api.github.com/users/datalab-to/received_events\\\",\\n \\\"type\\\": \\\"Organization\\\",\\n \\\"user_view_type\\\": \\\"public\\\",\\n \\\"site_admin\\\": false\\n },\\n \\\"html_url\\\": \\\"https://github.com/datalab-to/marker\\\",\\n \\\"description\\\": \\\"Convert PDF to markdown + JSON quickly with high accuracy\\\",\\n \\\"fork\\\": false,\\n \\\"url\\\": \\\"https://api.github.com/repos/datalab-to/marker\\\",\\n \\\"forks_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/forks\\\",\\n \\\"keys_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/keys{/key_id}\\\",\\n \\\"collaborators_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/collaborators{/collaborator}\\\",\\n \\\"teams_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/teams\\\",\\n \\\"hooks_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/hooks\\\",\\n \\\"issue_events_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/issues/events{/number}\\\",\\n \\\"events_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/events\\\",\\n \\\"assignees_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/assignees{/user}\\\",\\n \\\"branches_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/branches{/branch}\\\",\\n \\\"tags_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/tags\\\",\\n \\\"blobs_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/git/blobs{/sha}\\\",\\n \\\"git_tags_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/git/tags{/sha}\\\",\\n \\\"git_refs_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/git/refs{/sha}\\\",\\n \\\"trees_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/git/trees{/sha}\\\",\\n \\\"statuses_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/statuses/{sha}\\\",\\n \\\"languages_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/languages\\\",\\n \\\"stargazers_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/stargazers\\\",\\n \\\"contributors_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/contributors\\\",\\n \\\"subscribers_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/subscribers\\\",\\n \\\"subscription_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/subscription\\\",\\n \\\"commits_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/commits{/sha}\\\",\\n \\\"git_commits_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/git/commits{/sha}\\\",\\n \\\"comments_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/comments{/number}\\\",\\n \\\"issue_comment_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/issues/comments{/number}\\\",\\n \\\"contents_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/contents/{+path}\\\",\\n \\\"compare_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/compare/{base}...{head}\\\",\\n \\\"merges_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/merges\\\",\\n \\\"archive_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/{archive_format}{/ref}\\\",\\n \\\"downloads_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/downloads\\\",\\n \\\"issues_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/issues{/number}\\\",\\n \\\"pulls_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/pulls{/number}\\\",\\n \\\"milestones_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/milestones{/number}\\\",\\n \\\"notifications_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/notifications{?since,all,participating}\\\",\\n \\\"labels_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/labels{/name}\\\",\\n \\\"releases_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/releases{/id}\\\",\\n \\\"deployments_url\\\": \\\"https://api.github.com/repos/datalab-to/marker/deployments\\\",\\n \\\"created_at\\\": \\\"2023-10-30T20:14:08Z\\\",\\n \\\"updated_at\\\": \\\"2026-05-10T06:21:43Z\\\",\\n \\\"pushed_at\\\": \\\"2026-05-05T08:04:51Z\\\",\\n \\\"git_url\\\": \\\"git://github.com/datalab-to/marker.git\\\",\\n \\\"ssh_url\\\": \\\"git@github.com:datalab-to/marker.git\\\",\\n \\\"clone_url\\\": \\\"https://github.com/datalab-to/marker.git\\\",\\n \\\"svn_url\\\": \\\"https://github.com/datalab-to/marker\\\",\\n \\\"homepage\\\": \\\"https://www.datalab.to\\\",\\n \\\"size\\\": 18463,\\n \\\"stargazers_count\\\": 34892,\\n \\\"watchers_count\\\": 34892,\\n \\\"language\\\": \\\"Python\\\",\\n \\\"has_issues\\\": true,\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "sEMJzP97RvkTPAC0giPqhUevXtEs8bsp", "timestamp": "2026-05-10T08:52:16.084721"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "Lp0d5n4lxw2jJZKQNsry4kRReVZRGzOo", "call_id": "Lp0d5n4lxw2jJZKQNsry4kRReVZRGzOo", "response_item_id": "fc_Lp0d5n4lxw2jJZKQNsry4kRReVZRGzOo", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"curl -s https://api.github.com/repos/datalab-to/marker 2>&1 | python3 -c \\\"import sys,json; d=json.load(sys.stdin); print(f'Stars: {d[\\\\\\\"stargazers_count\\\\\\\"]}'); print(f'Forks: {d[\\\\\\\"forks_count\\\\\\\"]}'); print(f'Open issues: {d[\\\\\\\"open_issues\\\\\\\"]}'); print(f'Last push: {d[\\\\\\\"pushed_at\\\\\\\"]}'); print(f'License: {d.get(\\\\\\\"license\\\\\\\",{}).get(\\\\\\\"spdx_id\\\\\\\",\\\\\\\"unknown\\\\\\\")}'); print(f'Language: {d[\\\\\\\"language\\\\\\\"]}')\\\"\",\"timeout\":15}"}}], "timestamp": "2026-05-10T08:52:16.084721"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"Stars: 34892\\nForks: 2420\\nOpen issues: 408\\nLast push: 2026-05-05T08:04:51Z\\nLicense: GPL-3.0\\nLanguage: Python\", \"exit_code\": 0, \"error\": null, \"approval\": \"Command required approval (Security scan — [HIGH] Pipe to interpreter: curl | python3: Command pipes output from 'curl' directly to interpreter 'python3'. Downloaded content will be executed without inspection.\\n Safer: tirith run https://api.github.com/repos/datalab-to/marker — or: vet https://api.github.com/repos/datalab-to/marker (https://getvet.sh)) and was approved by the user.\"}", "tool_call_id": "Lp0d5n4lxw2jJZKQNsry4kRReVZRGzOo", "timestamp": "2026-05-10T08:52:16.084721"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "dgFyxcgTeKiFbEl8PTTfzWbWlVfmwfKi", "call_id": "dgFyxcgTeKiFbEl8PTTfzWbWlVfmwfKi", "response_item_id": "fc_dgFyxcgTeKiFbEl8PTTfzWbWlVfmwfKi", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"curl -s https://raw.githubusercontent.com/datalab-to/marker/main/README.md 2>&1 | head -120\",\"timeout\":15}"}}], "timestamp": "2026-05-10T08:52:16.084721"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"404: Not Found\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "dgFyxcgTeKiFbEl8PTTfzWbWlVfmwfKi", "timestamp": "2026-05-10T08:52:16.084721"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "McSQtHKsjUjYugFhX1rvZrEDITc9VW9z", "call_id": "McSQtHKsjUjYugFhX1rvZrEDITc9VW9z", "response_item_id": "fc_McSQtHKsjUjYugFhX1rvZrEDITc9VW9z", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"curl -s https://api.github.com/repos/datalab-to/marker/contents/ 2>&1 | python3 -c \\\"import sys,json; files=json.load(sys.stdin); [print(f['name']) for f in files]\\\"\",\"timeout\":15}"}}], "timestamp": "2026-05-10T08:52:16.084721"} {"role": "tool", "name": "terminal", "content": "{\"output\": \".github\\n.gitignore\\n.pre-commit-config.yaml\\nCLA.md\\nLICENSE\\nMODEL_LICENSE\\nREADME.md\\nbenchmarks\\nchunk_convert.py\\nconvert.py\\nconvert_single.py\\ndata\\nexamples\\nextraction_app.py\\nmarker\\nmarker_app.py\\nmarker_server.py\\npoetry.lock\\npyproject.toml\\npytest.ini\\nsignatures\\nstatic\\ntests\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "McSQtHKsjUjYugFhX1rvZrEDITc9VW9z", "timestamp": "2026-05-10T08:52:16.084721"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "qYUuSlVo1DlKPBaEVJGZr2A1ugV65VjE", "call_id": "qYUuSlVo1DlKPBaEVJGZr2A1ugV65VjE", "response_item_id": "fc_qYUuSlVo1DlKPBaEVJGZr2A1ugV65VjE", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"curl -s https://raw.githubusercontent.com/datalab-to/marker/master/README.md 2>&1 | head -200\",\"timeout\":15}"}}], "timestamp": "2026-05-10T08:52:16.084721"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"<p align=\\\"center\\\">\\n <img src=\\\"data/images/datalab-logo.png\\\" alt=\\\"Datalab Logo\\\" width=\\\"150\\\"/>\\n</p>\\n<h1 align=\\\"center\\\">Datalab</h1>\\n<p align=\\\"center\\\">\\n <strong>State of the Art models for Document Intelligence</strong>\\n</p>\\n<p align=\\\"center\\\">\\n <a href=\\\"https://www.gnu.org/licenses/gpl-3.0.html\\\"><img src=\\\"https://img.shields.io/badge/Code%20License-GPL--3.0-green.svg\\\" alt=\\\"Code License\\\"></a>\\n <a href=\\\"https://www.datalab.to/pricing\\\"><img src=\\\"https://img.shields.io/badge/Model%20License-OpenRAIL--M-blue.svg\\\" alt=\\\"Model License\\\"></a>\\n <a href=\\\"https://discord.gg/KuZwXNGnfH\\\"><img src=\\\"https://img.shields.io/badge/Discord-Join%20us-5865F2?logo=discord&logoColor=white\\\" alt=\\\"Discord\\\"></a>\\n</p>\\n<p align=\\\"center\\\">\\n <a href=\\\"https://www.datalab.to\\\"><img src=\\\"https://img.shields.io/badge/Homepage-datalab.to-blue\\\" alt=\\\"Homepage\\\"></a>\\n <a href=\\\"https://documentation.datalab.to\\\"><img src=\\\"https://img.shields.io/badge/Docs-Read%20the%20docs-blue\\\" alt=\\\"Docs\\\"></a>\\n <a href=\\\"https://www.datalab.to/playground\\\"><img src=\\\"https://img.shields.io/badge/Playground-Try%20it-orange\\\" alt=\\\"Public Playground\\\"></a>\\n</p>\\n\\n<hr/>\\n\\n# Marker\\n\\nMarker converts documents to markdown, JSON, chunks, and HTML quickly and accurately.\\n\\n- Converts PDF, image, PPTX, DOCX, XLSX, HTML, EPUB files in all languages\\n- Formats tables, forms, equations, inline math, links, references, and code blocks\\n- Extracts and saves images\\n- Removes headers/footers/other artifacts\\n- Extensible with your own formatting and logic\\n- Does structured extraction, given a JSON schema (beta)\\n- Optionally boost accuracy with LLMs (and your own prompt)\\n- Works on GPU, CPU, or MPS\\n\\n## Try Datalab's Managed Platform\\n\\nOur managed platform runs our latest open source model, [Chandra](https://github.com/datalab-to/chandra) — higher accuracy than Marker, with zero data retention by default, SOC 2 Type 2, and custom BAAs.\\n\\nIf you have high volume workloads, we offer a batch processing service that has processed 200M+ pages per week — we manage the infrastructure so your workloads finish on time.\\n\\nGet started with **$5 in free credits** — [sign up](https://www.datalab.to/?utm_source=gh-marker) — takes under 30 seconds — or try our [public playground](https://www.datalab.to/playground?utm_source=gh-marker).\\n\\nCommercial self-hosting requires a license — see [Commercial usage](#commercial-usage). For on-prem licensing, [contact us](https://www.datalab.to/contact?utm_source=gh-marker-onprem).\\n\\n## Performance\\n\\n<img src=\\\"data/images/overall.png\\\" width=\\\"800px\\\"/>\\n\\nMarker benchmarks favorably compared to cloud services like Llamaparse and Mathpix, as well as other open source tools.\\n\\nThe above results are running single PDF pages serially. Marker is significantly faster when running in batch mode, with a projected throughput of 25 pages/second on an H100.\\n\\nSee [below](#benchmarks) for detailed speed and accuracy benchmarks, and instructions on how to run your own benchmarks.\\n\\n## Hybrid Mode\\n\\nFor the highest accuracy, pass the `--use_llm` flag to use an LLM alongside marker. This will do things like merge tables across pages, handle inline math, format tables properly, and extract values from forms. It can use any gemini or ollama model. By default, it uses `gemini-2.0-flash`. See [below](#llm-services) for details.\\n\\nHere is a table benchmark comparing marker, gemini flash alone, and marker with use_llm:\\n\\n<img src=\\\"data/images/table.png\\\" width=\\\"400px\\\"/>\\n\\nAs you can see, the use_llm mode offers higher accuracy than marker or gemini alone.\\n\\n## Examples\\n\\n| PDF | File type | Markdown | JSON |\\n|-----|-----------|------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------|\\n| [Think Python](https://greenteapress.com/thinkpython/thinkpython.pdf) | Textbook | [View](https://github.com/VikParuchuri/marker/blob/master/data/examples/markdown/thinkpython/thinkpython.md) | [View](https://github.com/VikParuchuri/marker/blob/master/data/examples/json/thinkpython.json) |\\n| [Switch Transformers](https://arxiv.org/pdf/2101.03961.pdf) | arXiv paper | [View](https://github.com/VikParuchuri/marker/blob/master/data/examples/markdown/switch_transformers/switch_trans.md) | [View](https://github.com/VikParuchuri/marker/blob/master/data/examples/json/switch_trans.json) |\\n| [Multi-column CNN](https://arxiv.org/pdf/1804.07821.pdf) | arXiv paper | [View](https://github.com/VikParuchuri/marker/blob/master/data/examples/markdown/multicolcnn/multicolcnn.md) | [View](https://github.com/VikParuchuri/marker/blob/master/data/examples/json/multicolcnn.json) |\\n\\n# Commercial usage\\n\\nOur model weights use a modified AI Pubs Open Rail-M license (free for research, personal use, and startups under $2M funding/revenue) and our code is GPL. For broader commercial licensing or to remove GPL requirements, visit our pricing page [here](https://www.datalab.to/pricing?utm_source=gh-marker).\\n\\n# Community\\n\\n[Discord](https://discord.gg//KuZwXNGnfH) is where we discuss future development.\\n\\n# Installation\\n\\nYou'll need python 3.10+ and [PyTorch](https://pytorch.org/get-started/locally/).\\n\\nInstall with:\\n\\n```shell\\npip install marker-pdf\\n```\\n\\nIf you want to use marker on documents other than PDFs, you will need to install additional dependencies with:\\n\\n```shell\\npip install marker-pdf[full]\\n```\\n\\n# Usage\\n\\nFirst, some configuration:\\n\\n- Your torch device will be automatically detected, but you can override this. For example, `TORCH_DEVICE=cuda`.\\n- Some PDFs, even digital ones, have bad text in them. Set `--force_ocr` to force OCR on all lines, or the `strip_existing_ocr` to keep all digital text, and strip out any existing OCR text.\\n- If you care about inline math, set `force_ocr` to convert inline math to LaTeX.\\n\\n## Interactive App\\n\\nI've included a streamlit app that lets you interactively try marker with some basic options. Run it with:\\n\\n```shell\\npip install streamlit streamlit-ace\\nmarker_gui\\n```\\n\\n## Convert a single file\\n\\n```shell\\nmarker_single /path/to/file.pdf\\n```\\n\\nYou can pass in PDFs or images.\\n\\nOptions:\\n- `--page_range TEXT`: Specify which pages to process. Accepts comma-separated page numbers and ranges. Example: `--page_range \\\"0,5-10,20\\\"` will process pages 0, 5 through 10, and page 20.\\n- `--output_format [markdown|json|html|chunks]`: Specify the format for the output results.\\n- `--output_dir PATH`: Directory where output files will be saved. Defaults to the value specified in settings.OUTPUT_DIR.\\n- `--paginate_output`: Paginates the output, using `\\\\n\\\\n{PAGE_NUMBER}` followed by `-` * 48, then `\\\\n\\\\n`\\n- `--use_llm`: Uses an LLM to improve accuracy. You will need to configure the LLM backend - see [below](#llm-services).\\n- `--force_ocr`: Force OCR processing on the entire document, even for pages that might contain extractable text. This will also format inline math properly.\\n- `--block_correction_prompt`: if LLM mode is active, an optional prompt that will be used to correct the output of marker. This is useful for custom formatting or logic that you want to apply to the output.\\n- `--strip_existing_ocr`: Remove all existing OCR text in the document and re-OCR with surya.\\n- `--redo_inline_math`: If you want the absolute highest quality inline math conversion, use this along with `--use_llm`.\\n- `--disable_image_extraction`: Don't extract images from the PDF. If you also specify `--use_llm`, then images will be replaced with a description.\\n- `--debug`: Enable debug mode for additional logging and diagnostic information.\\n- `--processors TEXT`: Override the default processors by providing their full module paths, separated by commas. Example: `--processors \\\"module1.processor1,module2.processor2\\\"`\\n- `--config_json PATH`: Path to a JSON configuration file containing additional settings.\\n- `config --help`: List all available builders, processors, and converters, and their associated configuration. These values can be used to build a JSON configuration file for additional tweaking of marker defaults.\\n- `--converter_cls`: One of `marker.converters.pdf.PdfConverter` (default) or `marker.converters.table.TableConverter`. The `PdfConverter` will convert the whole PDF, the `TableConverter` will only extract and convert tables.\\n- `--llm_service`: Which llm service to use if `--use_llm` is passed. This defaults to `marker.services.gemini.GoogleGeminiService`.\\n- `--help`: see all of the flags that can be passed into marker. (it supports many more options then are listed above)\\n\\nThe list of supported languages for surya OCR is [here](https://github.com/VikParuchuri/surya/blob/master/surya/recognition/languages.py). If you don't need OCR, marker can work with any language.\\n\\n## Convert multiple files\\n\\n```shell\\nmarker /path/to/input/folder\\n```\\n\\n- `marker` supports all the same options from `marker_single` above.\\n- `--workers` is the number of conversion workers to run simultaneously. This is automatically set by default, but you can increase it to increase throughput, at the cost of more CPU/GPU usage. Marker will use 5GB of VRAM per worker at the peak, and 3.5GB average.\\n\\n## Convert multiple files on multiple GPUs\\n\\n```shell\\nNUM_DEVICES=4 NUM_WORKERS=15 marker_chunk_convert ../pdf_in ../md_out\\n```\\n\\n- `NUM_DEVICES` is the number of GPUs to use. Should be `2` or greater.\\n- `NUM_WORKERS` is the number of parallel processes to run on each GPU.\\n\\n## Use from python\\n\\nSee the `PdfConverter` class at `marker/converters/pdf.py` function for additional arguments that can be passed.\\n\\n```python\\nfrom marker.converters.pdf import PdfConverter\\nfrom marker.models import create_model_dict\\nfrom marker.output import text_from_rendered\\n\\nconverter = PdfConverter(\\n artifact_dict=create_model_dict(),\\n)\\nrendered = converter(\\\"FILEPATH\\\")\\ntext, _, images = text_from_rendered(rendered)\\n```\\n\\n`rendered` will be a pydantic basemodel with different properties depending on the output type requested. With markdown output (default), you'll have the properties `markdown`, `metadata`, and `images`. For json output, you'll have `children`, `block_type`, and `metadata`.\\n\\n### Custom configuration\\n\\nYou can pass configuration using the `ConfigParser`. To see all available options, do `marker_single --help`.\\n\\n```python\\nfrom marker.converters.pdf import PdfConverter\\nfrom marker.models import create_model_dict\\nfrom marker.config.parser import ConfigParser\\n\\nconfig = {\\n \\\"output_format\\\": \\\"json\\\",\\n \\\"ADDITIONAL_KEY\\\": \\\"VALUE\\\"\\n}\\nconfig_parser = ConfigParser(config)\\n\\nconverter = PdfConverter(\\n config=config_parser.generate_config_dict(),\\n artifact_dict=create_model_dict(),\\n processor_list=config_parser.get_processors(),\\n renderer=config_parser.get_renderer(),\\n llm_service=config_parser.get_llm_service()\\n)\\nrendered = converter(\\\"FILEPATH\\\")\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "qYUuSlVo1DlKPBaEVJGZr2A1ugV65VjE", "timestamp": "2026-05-10T08:52:16.084721"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "KqysFtNKDGN98wnj6ILOAu1qfDHmHMuh", "call_id": "KqysFtNKDGN98wnj6ILOAu1qfDHmHMuh", "response_item_id": "fc_KqysFtNKDGN98wnj6ILOAu1qfDHmHMuh", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"curl -s https://raw.githubusercontent.com/datalab-to/marker/master/README.md 2>&1 | tail -200\",\"timeout\":15}"}}], "timestamp": "2026-05-10T08:52:16.084721"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"\\\"polygon\\\": [...]\\n }\\n ], // computed PDF table of contents\\n \\\"page_stats\\\": [\\n {\\n \\\"page_id\\\": 0,\\n \\\"text_extraction_method\\\": \\\"pdftext\\\",\\n \\\"block_counts\\\": [(\\\"Span\\\", 200), ...]\\n },\\n ...\\n ]\\n}\\n```\\n\\n# LLM Services\\n\\nWhen running with the `--use_llm` flag, you have a choice of services you can use:\\n\\n- `Gemini` - this will use the Gemini developer API by default. You'll need to pass `--gemini_api_key` to configuration.\\n- `Google Vertex` - this will use vertex, which can be more reliable. You'll need to pass `--vertex_project_id`. To use it, set `--llm_service=marker.services.vertex.GoogleVertexService`.\\n- `Ollama` - this will use local models. You can configure `--ollama_base_url` and `--ollama_model`. To use it, set `--llm_service=marker.services.ollama.OllamaService`.\\n- `Claude` - this will use the anthropic API. You can configure `--claude_api_key`, and `--claude_model_name`. To use it, set `--llm_service=marker.services.claude.ClaudeService`.\\n- `OpenAI` - this supports any openai-like endpoint. You can configure `--openai_api_key`, `--openai_model`, and `--openai_base_url`. To use it, set `--llm_service=marker.services.openai.OpenAIService`.\\n- `Azure OpenAI` - this uses the Azure OpenAI service. You can configure `--azure_endpoint`, `--azure_api_key`, and `--deployment_name`. To use it, set `--llm_service=marker.services.azure_openai.AzureOpenAIService`.\\n\\nThese services may have additional optional configuration as well - you can see it by viewing the classes.\\n\\n# Internals\\n\\nMarker is easy to extend. The core units of marker are:\\n\\n- `Providers`, at `marker/providers`. These provide information from a source file, like a PDF.\\n- `Builders`, at `marker/builders`. These generate the initial document blocks and fill in text, using info from the providers.\\n- `Processors`, at `marker/processors`. These process specific blocks, for example the table formatter is a processor.\\n- `Renderers`, at `marker/renderers`. These use the blocks to render output.\\n- `Schema`, at `marker/schema`. The classes for all the block types.\\n- `Converters`, at `marker/converters`. They run the whole end to end pipeline.\\n\\nTo customize processing behavior, override the `processors`. To add new output formats, write a new `renderer`. For additional input formats, write a new `provider.`\\n\\nProcessors and renderers can be directly passed into the base `PDFConverter`, so you can specify your own custom processing easily.\\n\\n## API server\\n\\nThere is a very simple API server you can run like this:\\n\\n```shell\\npip install -U uvicorn fastapi python-multipart\\nmarker_server --port 8001\\n```\\n\\nThis will start a fastapi server that you can access at `localhost:8001`. You can go to `localhost:8001/docs` to see the endpoint options.\\n\\nYou can send requests like this:\\n\\n```\\nimport requests\\nimport json\\n\\npost_data = {\\n 'filepath': 'FILEPATH',\\n # Add other params here\\n}\\n\\nrequests.post(\\\"http://localhost:8001/marker\\\", data=json.dumps(post_data)).json()\\n```\\n\\nNote that this is not a very robust API, and is only intended for small-scale use. If you want to use this server, but want a more robust conversion option, you can use the hosted [Datalab API](https://www.datalab.to/plans).\\n\\n# Troubleshooting\\n\\nThere are some settings that you may find useful if things aren't working the way you expect:\\n\\n- If you have issues with accuracy, try setting `--use_llm` to use an LLM to improve quality. You must set `GOOGLE_API_KEY` to a Gemini API key for this to work.\\n- Make sure to set `force_ocr` if you see garbled text - this will re-OCR the document.\\n- `TORCH_DEVICE` - set this to force marker to use a given torch device for inference.\\n- If you're getting out of memory errors, decrease worker count. You can also try splitting up long PDFs into multiple files.\\n\\n## Debugging\\n\\nPass the `debug` option to activate debug mode. This will save images of each page with detected layout and text, as well as output a json file with additional bounding box information.\\n\\n# Benchmarks\\n\\n## Overall PDF Conversion\\n\\nWe created a [benchmark set](https://huggingface.co/datasets/datalab-to/marker_benchmark) by extracting single PDF pages from common crawl. We scored based on a heuristic that aligns text with ground truth text segments, and an LLM as a judge scoring method.\\n\\n| Method | Avg Time | Heuristic Score | LLM Score |\\n|------------|----------|-----------------|-----------|\\n| marker | 2.83837 | 95.6709 | 4.23916 |\\n| llamaparse | 23.348 | 84.2442 | 3.97619 |\\n| mathpix | 6.36223 | 86.4281 | 4.15626 |\\n| docling | 3.69949 | 86.7073 | 3.70429 |\\n\\nBenchmarks were run on an H100 for markjer and docling - llamaparse and mathpix used their cloud services. We can also look at it by document type:\\n\\n<img src=\\\"data/images/per_doc.png\\\" width=\\\"1000px\\\"/>\\n\\n| Document Type | Marker heuristic | Marker LLM | Llamaparse Heuristic | Llamaparse LLM | Mathpix Heuristic | Mathpix LLM | Docling Heuristic | Docling LLM |\\n|----------------------|------------------|------------|----------------------|----------------|-------------------|-------------|-------------------|-------------|\\n| Scientific paper | 96.6737 | 4.34899 | 87.1651 | 3.96421 | 91.2267 | 4.46861 | 92.135 | 3.72422 |\\n| Book page | 97.1846 | 4.16168 | 90.9532 | 4.07186 | 93.8886 | 4.35329 | 90.0556 | 3.64671 |\\n| Other | 95.1632 | 4.25076 | 81.1385 | 4.01835 | 79.6231 | 4.00306 | 83.8223 | 3.76147 |\\n| Form | 88.0147 | 3.84663 | 66.3081 | 3.68712 | 64.7512 | 3.33129 | 68.3857 | 3.40491 |\\n| Presentation | 95.1562 | 4.13669 | 81.2261 | 4 | 83.6737 | 3.95683 | 84.8405 | 3.86331 |\\n| Financial document | 95.3697 | 4.39106 | 82.5812 | 4.16111 | 81.3115 | 4.05556 | 86.3882 | 3.8 |\\n| Letter | 98.4021 | 4.5 | 93.4477 | 4.28125 | 96.0383 | 4.45312 | 92.0952 | 4.09375 |\\n| Engineering document | 93.9244 | 4.04412 | 77.4854 | 3.72059 | 80.3319 | 3.88235 | 79.6807 | 3.42647 |\\n| Legal document | 96.689 | 4.27759 | 86.9769 | 3.87584 | 91.601 | 4.20805 | 87.8383 | 3.65552 |\\n| Newspaper page | 98.8733 | 4.25806 | 84.7492 | 3.90323 | 96.9963 | 4.45161 | 92.6496 | 3.51613 |\\n| Magazine page | 98.2145 | 4.38776 | 87.2902 | 3.97959 | 93.5934 | 4.16327 | 93.0892 | 4.02041 |\\n\\n## Throughput\\n\\nWe benchmarked throughput using a [single long PDF](https://www.greenteapress.com/thinkpython/thinkpython.pdf).\\n\\n| Method | Time per page | Time per document | VRAM used |\\n|---------|---------------|-------------------|---------- |\\n| marker | 0.18 | 43.42 | 3.17GB |\\n\\nThe projected throughput is 122 pages per second on an H100 - we can run 22 individual processes given the VRAM used.\\n\\n## Table Conversion\\n\\nMarker can extract tables from PDFs using `marker.converters.table.TableConverter`. The table extraction performance is measured by comparing the extracted HTML representation of tables against the original HTML representations using the test split of [FinTabNet](https://developer.ibm.com/exchanges/data/all/fintabnet/). The HTML representations are compared using a tree edit distance based metric to judge both structure and content. Marker detects and identifies the structure of all tables in a PDF page and achieves these scores:\\n\\n| Method | Avg score | Total tables |\\n|------------------|-----------|--------------|\\n| marker | 0.816 | 99 |\\n| marker w/use_llm | 0.907 | 99 |\\n| gemini | 0.829 | 99 |\\n\\nThe `--use_llm` flag can significantly improve table recognition performance, as you can see.\\n\\nWe filter out tables that we cannot align with the ground truth, since fintabnet and our layout model have slightly different detection methods (this results in some tables being split/merged).\\n\\n## Running your own benchmarks\\n\\nYou can benchmark the performance of marker on your machine. Install marker manually with:\\n\\n```shell\\ngit clone https://github.com/VikParuchuri/marker.git\\npoetry install\\n```\\n\\n### Overall PDF Conversion\\n\\nDownload the benchmark data [here](https://drive.google.com/file/d/1ZSeWDo2g1y0BRLT7KnbmytV2bjWARWba/view?usp=sharing) and unzip. Then run the overall benchmark like this:\\n\\n```shell\\npython benchmarks/overall.py --methods marker --scores heuristic,llm\\n```\\n\\nOptions:\\n\\n- `--use_llm` use an llm to improve the marker results.\\n- `--max_rows` how many rows to process for the benchmark.\\n- `--methods` can be `llamaparse`, `mathpix`, `docling`, `marker`. Comma separated.\\n- `--scores` which scoring functions to use, can be `llm`, `heuristic`. Comma separated.\\n\\n### Table Conversion\\nThe processed FinTabNet dataset is hosted [here](https://huggingface.co/datasets/datalab-to/fintabnet-test) and is automatically downloaded. Run the benchmark with:\\n\\n```shell\\npython benchmarks/table/table.py --max_rows 100\\n```\\n\\nOptions:\\n\\n- `--use_llm` uses an llm with marker to improve accuracy.\\n- `--use_gemini` also benchmarks gemini 2.0 flash.\\n\\n# How it works\\n\\nMarker is a pipeline of deep learning models:\\n\\n- Extract text, OCR if necessary (heuristics, [surya](https://github.com/VikParuchuri/surya))\\n- Detect page layout and find reading order ([surya](https://github.com/VikParuchuri/surya))\\n- Clean and format each block (heuristics, [texify](https://github.com/VikParuchuri/texify), [surya](https://github.com/VikParuchuri/surya))\\n- Optionally use an LLM to improve quality\\n- Combine blocks and postprocess complete text\\n\\nIt only uses models where necessary, which improves speed and accuracy.\\n\\n# Limitations\\n\\nPDF is a tricky format, so marker will not always work perfectly. Here are some known limitations that are on the roadmap to address:\\n\\n- Very complex layouts, with nested tables and forms, may not work\\n- Forms may not be rendered well\\n\\nNote: Passing the `--use_llm` and `--force_ocr` flags will mostly solve these issues.\\n\\n# Usage and Deployment Examples\\n\\nYou can always run `marker` locally, but if you wanted to expose it as an API, we have a few options:\\n- Our platform API which is powered by `marker` and `surya` and is easy to test out - it's free to sign up, and we'll include credits, [try it out here](https://datalab.to)\\n- Our painless on-prem solution for commercial use, which you can [read about here](https://www.datalab.to/blog/self-serve-on-prem-licensing) and gives you privacy guarantees with high throughput inference optimizations.\\n- [Deployment example with Modal](./examples/README_MODAL.md) that shows you how to deploy and access `marker` through a web endpoint using [`Modal`](https://modal.com). Modal is an AI compute platform that enables developers to deploy and scale models on GPUs in minutes.\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "KqysFtNKDGN98wnj6ILOAu1qfDHmHMuh", "timestamp": "2026-05-10T08:52:16.084721"} {"role": "assistant", "content": "Gut, ich habe eine umfassende Analyse. Hier ist mein Urteil:\n\n---\n\n## Analyse: Marker (datalab-to/marker)\n\n**Fakten:**\n- **34.892 Stars** – extrem populär\n- **GPL-3.0** Code-Lizenz + **OpenRAIL-M** für die Models\n- **408 offene Issues** – aktives Projekt, aber auch viel Wartungsbedarf\n- Letztes Commit: vor 5 Tagen (sehr aktiv)\n- Pipeline: Surya (OCR/Layout) → Texify (Math) → LLM-Boost (optional)\n\n### ✅ Stärken\n\n1. **State-of-the-Art Genauigkeit** – Benchmarks schlagen Llamaparse, Mathpix und Docling in fast allen Kategorien (95.7% heuristic score vs. 84-87% bei Wettbewerbern)\n2. **LLM-Hybrid-Modus** – `--use_llm` mit Gemini/Ollama/Claude/OpenAI für Tabellen, Formulare und komplexe Layouts → 0.907 table score\n3. **Multi-Format** – PDF, Bilder, PPTX, DOCX, XLSX, HTML, EPUB\n4. **90+ Sprachen** via Surya OCR\n5. **Equations, Code-Blöcke, Tabellen** – alles korrekt formatiert\n6. **GPU/CPU/MPS** – läuft überall\n7. **Batch-Modus** – 25 Seiten/Sekunde auf H100, skalierbar\n\n### ⚠️ Schwächen / Risiken\n\n1. **GPL-3.0 Lizenz** – kommerzielle Nutzung erfordert Lizenzkauf. Forschung/Privat/Startups < $2M OK.\n2. **~5GB VRAM pro Worker** – braucht GPU für gute Performance, auf CPU langsam (1-14s/Seite)\n3. **Modell-Download ~2.5GB** (HuggingFace Cache)\n4. **Komplexe Layouts** (verschachtelte Tabellen + Formulare) – bekanntes Limit\n5. **408 offene Issues** – Projekt ist noch unreif, Breaking Changes möglich\n6. **Abhängigkeit von externen LLMs** für `--use_llm` (Gemini API-Key nötig)\n\n### 🤔 Mein Urteil\n\n**Für unseren Use-Case (medizinische/lab-PDFs, interne Dokumente): Ja, definitiv empfehlenswert.**\n\nIm Vergleich zu unserer aktuellen Lösung (pymupdf + pytesseract) wäre der Qualitätssprung massiv:\n- pymupdf: keine OCR, keine Equations, keine korrekte Tabellen\n- marker: alles in einem, 95.7% accuracy\n\n**Empfehlung:** Installieren mit `pip install marker-pdf[full]` und für unsere medizinischen Dokumente testen. Ohne GPU läuft es auf CPU – langsamer, aber für sporadische Nutzung OK.\n\nSoll ich es installieren und mit einem Test-PDF evaluieren?", "reasoning": null, "finish_reason": "stop", "timestamp": "2026-05-10T08:52:16.084721"} {"role": "user", "content": "Kennst du evtl eine noch bessere Lösung?", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "Lass mich recherchieren, ob es aktuell bessere Alternativen gibt.", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "0M3EMvHB5FnqkK8wIE6SZ4MV3Go91ngJ", "call_id": "0M3EMvHB5FnqkK8wIE6SZ4MV3Go91ngJ", "response_item_id": "fc_0M3EMvHB5FnqkK8wIE6SZ4MV3Go91ngJ", "type": "function", "function": {"name": "skill_view", "arguments": "{\"name\":\"tavily-search\"}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "skill_view", "content": "{\"success\": true, \"name\": \"tavily-search\", \"description\": \"Search the web with LLM-optimized results via the Tavily CLI. Use this skill when the user wants to search the web, find articles, look up information, get recent news, discover sources, or says \\\"search for\\\", \\\"find me\\\", \\\"look up\\\", \\\"what's the latest on\\\", \\\"find articles about\\\", or needs current information from the internet. Returns relevant results with content snippets, relevance scores, and metadata — optimized for LLM consumption. Supports domain filtering, time ranges, and multiple search depths.\\n\", \"tags\": [], \"related_skills\": [], \"content\": \"---\\nname: tavily-search\\ndescription: |\\n Search the web with LLM-optimized results via the Tavily CLI. Use this skill when the user wants to search the web, find articles, look up information, get recent news, discover sources, or says \\\"search for\\\", \\\"find me\\\", \\\"look up\\\", \\\"what's the latest on\\\", \\\"find articles about\\\", or needs current information from the internet. Returns relevant results with content snippets, relevance scores, and metadata — optimized for LLM consumption. Supports domain filtering, time ranges, and multiple search depths.\\nallowed-tools: Bash(tvly *)\\n---\\n\\n# tavily search\\n\\nWeb search returning LLM-optimized results with content snippets and relevance scores.\\n\\n## Before running any command\\n\\nCheck if Tavily is configured:\\n\\n```bash\\nwhich tvly && tvly search \\\"test\\\" --json 2>&1 | head -5\\n```\\n\\nIf `tvly` is not found or authentication fails:\\n\\n```bash\\ncurl -fsSL https://cli.tavily.com/install.sh | bash && tvly login\\n```\\n\\n**Pitfall:** If Tavily is not configured (no API key), do NOT attempt to install it without asking. Fall back to `multi-search-engine` or `browser_navigate` instead. Tavily requires an API key from tavily.com — it is not pre-configured in all environments.\\n\\nSee [tavily-cli](../tavily-cli/SKILL.md) for alternative install methods and auth options.\\n\\n## When to use\\n\\n- You need to find information on any topic\\n- You don't have a specific URL yet\\n- First step in the [workflow](../tavily-cli/SKILL.md): **search** → extract → map → crawl → research\\n\\n## Quick start\\n\\n```bash\\n# Basic search\\ntvly search \\\"your query\\\" --json\\n\\n# Advanced search with more results\\ntvly search \\\"quantum computing\\\" --depth advanced --max-results 10 --json\\n\\n# Recent news\\ntvly search \\\"AI news\\\" --time-range week --topic news --json\\n\\n# Domain-filtered\\ntvly search \\\"SEC filings\\\" --include-domains sec.gov,reuters.com --json\\n\\n# Include full page content in results\\ntvly search \\\"react hooks tutorial\\\" --include-raw-content --max-results 3 --json\\n```\\n\\n## Options\\n\\n| Option | Description |\\n|--------|-------------|\\n| `--depth` | `ultra-fast`, `fast`, `basic` (default), `advanced` |\\n| `--max-results` | Max results, 0-20 (default: 5) |\\n| `--topic` | `general` (default), `news`, `finance` |\\n| `--time-range` | `day`, `week`, `month`, `year` |\\n| `--start-date` | Results after date (YYYY-MM-DD) |\\n| `--end-date` | Results before date (YYYY-MM-DD) |\\n| `--include-domains` | Comma-separated domains to include |\\n| `--exclude-domains` | Comma-separated domains to exclude |\\n| `--country` | Boost results from country |\\n| `--include-answer` | Include AI answer (`basic` or `advanced`) |\\n| `--include-raw-content` | Include full page content (`markdown` or `text`) |\\n| `--include-images` | Include image results |\\n| `--include-image-descriptions` | Include AI image descriptions |\\n| `--chunks-per-source` | Chunks per source (advanced/fast depth only) |\\n| `-o, --output` | Save output to file |\\n| `--json` | Structured JSON output |\\n\\n## Search depth\\n\\n| Depth | Speed | Relevance | Best for |\\n|-------|-------|-----------|----------|\\n| `ultra-fast` | Fastest | Lower | Real-time chat, autocomplete |\\n| `fast` | Fast | Good | Need chunks, latency matters |\\n| `basic` | Medium | High | General-purpose (default) |\\n| `advanced` | Slower | Highest | Precision, specific facts |\\n\\n## Tips\\n\\n- **Keep queries under 400 characters** — think search query, not prompt.\\n- **Break complex queries into sub-queries** for better results.\\n- **Use `--include-raw-content`** when you need full page text (saves a separate extract call).\\n- **Use `--include-domains`** to focus on trusted sources.\\n- **Use `--time-range`** for recent information.\\n- Read from stdin: `echo \\\"query\\\" | tvly search - --json`\\n\\n## See also\\n\\n- [tavily-extract](../tavily-extract/SKILL.md) — extract content from specific URLs\\n- [tavily-research](../tavily-research/SKILL.md) — comprehensive multi-source research\\n\", \"path\": \"openclaw-imports/tavily-search/SKILL.md\", \"skill_dir\": \"/home/agent/.hermes/skills/openclaw-imports/tavily-search\", \"linked_files\": null, \"usage_hint\": null, \"required_environment_variables\": [], \"required_commands\": [], \"missing_required_environment_variables\": [], \"missing_credential_files\": [], \"missing_required_commands\": [], \"setup_needed\": false, \"setup_skipped\": false, \"readiness_status\": \"available\"}", "tool_call_id": "0M3EMvHB5FnqkK8wIE6SZ4MV3Go91ngJ", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "hYeY1VUrnmGsitz2OL6a9TgpgkC4brJe", "call_id": "hYeY1VUrnmGsitz2OL6a9TgpgkC4brJe", "response_item_id": "fc_hYeY1VUrnmGsitz2OL6a9TgpgkC4brJe", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"tvly search \\\"best PDF OCR markdown conversion 2026\\\" --depth advanced --max-results 5 --json 2>&1 | python3 -c \\\"\\nimport sys, json\\ndata = json.load(sys.stdin)\\nfor r in data.get('results', []):\\n print(f'---')\\n print(f'Title: {r[\\\\\\\"title\\\\\\\"]}')\\n print(f'URL: {r[\\\\\\\"url\\\\\\\"]}')\\n print(f'Score: {r.get(\\\\\\\"score\\\\\\\", \\\\\\\"N/A\\\\\\\")}')\\n print(f'Snippet: {r.get(\\\\\\\"content\\\\\\\", \\\\\\\"\\\\\\\")[:300]}')\\n\\\"\",\"timeout\":30}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"Traceback (most recent call last):\\n File \\\"<string>\\\", line 3, in <module>\\n File \\\"/home/agent/.local/share/uv/python/cpython-3.11.15-linux-x86_64-gnu/lib/python3.11/json/__init__.py\\\", line 293, in load\\n return loads(fp.read(),\\n ^^^^^^^^^^^^^^^^\\n File \\\"/home/agent/.local/share/uv/python/cpython-3.11.15-linux-x86_64-gnu/lib/python3.11/json/__init__.py\\\", line 346, in loads\\n return _default_decoder.decode(s)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File \\\"/home/agent/.local/share/uv/python/cpython-3.11.15-linux-x86_64-gnu/lib/python3.11/json/decoder.py\\\", line 337, in decode\\n obj, end = self.raw_decode(s, idx=_w(s, 0).end())\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File \\\"/home/agent/.local/share/uv/python/cpython-3.11.15-linux-x86_64-gnu/lib/python3.11/json/decoder.py\\\", line 355, in raw_decode\\n raise JSONDecodeError(\\\"Expecting value\\\", s, err.value) from None\\njson.decoder.JSONDecodeError: Expecting value: line 1 column 3 (char 2)\", \"exit_code\": 1, \"error\": null, \"approval\": \"Command required approval (Security scan — [HIGH] Pipe to interpreter: tvly | python3: Command pipes output from 'tvly' directly to interpreter 'python3'. Downloaded content will be executed without inspection.) and was approved by the user.\"}", "tool_call_id": "hYeY1VUrnmGsitz2OL6a9TgpgkC4brJe", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "8XEK9pltbtznE2RDG69FahnHtNriUc11", "call_id": "8XEK9pltbtznE2RDG69FahnHtNriUc11", "response_item_id": "fc_8XEK9pltbtznE2RDG69FahnHtNriUc11", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"tvly search \\\"best PDF OCR markdown conversion 2026\\\" --depth advanced --max-results 5 --json 2>&1 | head -200\",\"timeout\":30}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"> Error: No Tavily API key found.\\n\\n Authenticate using one of:\\n tvly login\\n tvly login --api-key tvly-YOUR_KEY\\n export TAVILY_API_KEY=tvly-YOUR_KEY\\n\\n Get a key at tavily.com\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "8XEK9pltbtznE2RDG69FahnHtNriUc11", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "BklrPuwSEnK2uoFkqQIBkEBfRkIPZ3aV", "call_id": "BklrPuwSEnK2uoFkqQIBkEBfRkIPZ3aV", "response_item_id": "fc_BklrPuwSEnK2uoFkqQIBkEBfRkIPZ3aV", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"curl -s \\\"https://api.github.com/search/repositories?q=PDF+OCR+markdown+conversion+2026+stars:>1000&sort=stars&order=desc&per_page=20\\\" 2>&1 | python3 -c \\\"\\nimport sys, json\\ndata = json.load(sys.stdin)\\nfor r in data.get('items', []):\\n print(f'{r[\\\\\\\"full_name\\\\\\\"]} | ⭐{r[\\\\\\\"stargazers_count\\\\\\\"]} | {r[\\\\\\\"language\\\\\\\"]} | {r.get(\\\\\\\"description\\\\\\\",\\\\\\\"?\\\\\\\")[:120]}')\\n\\\" 2>&1\",\"timeout\":15}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "BklrPuwSEnK2uoFkqQIBkEBfRkIPZ3aV", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "fPtyHxCJd4D8Wrp9go20HXpdoob6bE85", "call_id": "fPtyHxCJd4D8Wrp9go20HXpdoob6bE85", "response_item_id": "fc_fPtyHxCJd4D8Wrp9go20HXpdoob6bE85", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"curl -s \\\"https://api.github.com/search/repositories?q=PDF+OCR+markdown+2026&sort=stars&order=desc&per_page=15\\\" 2>&1 | python3 -c \\\"\\nimport sys, json\\ndata = json.load(sys.stdin)\\nfor r in data.get('items', []):\\n print(f'{r[\\\\\\\"full_name\\\\\\\"]} | ⭐{r[\\\\\\\"stargazers_count\\\\\\\"]} | {r[\\\\\\\"language\\\\\\\"]} | {r.get(\\\\\\\"description\\\\\\\",\\\\\\\"?\\\\\\\")[:120]}')\\n\\\" 2>&1\",\"timeout\":15}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"ijaureguialzo/llm-ocr | ⭐0 | Python | OCR de PDF a Markdown con LLM local.\\nyumikovn1/election-69-OCR-result | ⭐0 | None | Provide structured, machine-readable JSON of the official 2026 Thai election results extracted via OCR from Election Com\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "fPtyHxCJd4D8Wrp9go20HXpdoob6bE85", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "Gtoey1KjURaSNa7kU3S5zL9EhxaJrwaj", "call_id": "Gtoey1KjURaSNa7kU3S5zL9EhxaJrwaj", "response_item_id": "fc_Gtoey1KjURaSNa7kU3S5zL9EhxaJrwaj", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"# Suche nach den wichtigsten PDF-OCR/Markdown Tools\\nfor repo in \\\"VikParuchuri/marker\\\" \\\"datalab-to/marker\\\" \\\"docling-project/docling\\\" \\\"Surya-OCR/surya\\\" \\\"JohannesKrause/tesseract-ocr\\\" \\\"pymupdf\\\" \\\"Niekrom/Marker\\\" \\\"labmlai/annotated_deep_learning_paper_implementations\\\" \\\"translucide/translucide\\\" \\\"SAP/asciidoc-py\\\" \\\"microsoft/DocLayNet\\\" \\\"google-research/google-research\\\" \\\"allenai/scispacy\\\" \\\"deepset-ai/farm\\\" \\\"LayoutLM\\\"; do\\n curl -s \\\"https://api.github.com/repos/$repo\\\" 2>&1 | python3 -c \\\"\\nimport sys, json\\ntry:\\n d = json.load(sys.stdin)\\n print(f'{d[\\\\\\\"full_name\\\\\\\"]} | ⭐{d[\\\\\\\"stargazers_count\\\\\\\"]} | {d.get(\\\\\\\"description\\\\\\\",\\\\\\\"?\\\\\\\")[:150]}')\\nexcept: pass\\n\\\" 2>&1\\ndone\",\"timeout\":60}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"datalab-to/marker | ⭐34892 | Convert PDF to markdown + JSON quickly with high accuracy\\ndocling-project/docling | ⭐59472 | Get your documents ready for gen AI\\nlabmlai/annotated_deep_learning_paper_implementations | ⭐66576 | 🧑‍🏫 60+ Implementations/tutorials of deep learning papers with side-by-side notes 📝; including transformers (original, xl, switch, feedback, vit, ...)\\ngoogle-research/google-research | ⭐37863 | Google Research\\nallenai/scispacy | ⭐1953 | A full spaCy pipeline and models for scientific/biomedical documents.\\ndeepset-ai/FARM | ⭐1754 | :house_with_garden: Fast & easy transfer learning for NLP. Harvesting language models for the industry. Focus on Question Answering.\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "Gtoey1KjURaSNa7kU3S5zL9EhxaJrwaj", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "DF4Nk7tIASezfrH08N7sQeZlYYZ7Y1OC", "call_id": "DF4Nk7tIASezfrH08N7sQeZlYYZ7Y1OC", "response_item_id": "fc_DF4Nk7tIASezfrH08N7sQeZlYYZ7Y1OC", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"# Docling - die Hauptkonkurrenz zu Marker\\ncurl -s https://api.github.com/repos/docling-project/docling 2>&1 | python3 -c \\\"\\nimport sys, json\\nd = json.load(sys.stdin)\\nprint(f'Name: {d[\\\\\\\"full_name\\\\\\\"]}')\\nprint(f'Stars: {d[\\\\\\\"stargazers_count\\\\\\\"]}')\\nprint(f'Forks: {d[\\\\\\\"forks_count\\\\\\\"]}')\\nprint(f'Issues: {d[\\\\\\\"open_issues\\\\\\\"]}')\\nprint(f'License: {d.get(\\\\\\\"license\\\\\\\",{}).get(\\\\\\\"spdx_id\\\\\\\",\\\\\\\"unknown\\\\\\\")}')\\nprint(f'Last push: {d[\\\\\\\"pushed_at\\\\\\\"]}')\\nprint(f'Description: {d.get(\\\\\\\"description\\\\\\\",\\\\\\\"\\\\\\\")}')\\n\\\" 2>&1\",\"timeout\":15}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"Name: docling-project/docling\\nStars: 59472\\nForks: 4115\\nIssues: 891\\nLicense: MIT\\nLast push: 2026-05-08T13:37:50Z\\nDescription: Get your documents ready for gen AI\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "DF4Nk7tIASezfrH08N7sQeZlYYZ7Y1OC", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "SWpTLXC3WqBpowshoTAJ5mZKP8i1c81t", "call_id": "SWpTLXC3WqBpowshoTAJ5mZKP8i1c81t", "response_item_id": "fc_SWpTLXC3WqBpowshoTAJ5mZKP8i1c81t", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"# Docling README lesen\\ncurl -s https://raw.githubusercontent.com/docling-project/docling/main/README.md 2>&1 | head -250\",\"timeout\":15}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"<p align=\\\"center\\\">\\n <a href=\\\"https://github.com/docling-project/docling\\\">\\n <img loading=\\\"lazy\\\" alt=\\\"Docling\\\" src=\\\"https://github.com/docling-project/docling/raw/main/docs/assets/docling_processing.png\\\" width=\\\"100%\\\"/>\\n </a>\\n</p>\\n\\n# Docling\\n\\n<p align=\\\"center\\\">\\n <a href=\\\"https://trendshift.io/repositories/12132\\\" target=\\\"_blank\\\"><img src=\\\"https://trendshift.io/api/badge/repositories/12132\\\" alt=\\\"DS4SD%2Fdocling | Trendshift\\\" style=\\\"width: 250px; height: 55px;\\\" width=\\\"250\\\" height=\\\"55\\\"/></a>\\n</p>\\n\\n[![arXiv](https://img.shields.io/badge/arXiv-2408.09869-b31b1b.svg)](https://arxiv.org/abs/2408.09869)\\n[![Docs](https://img.shields.io/badge/docs-live-brightgreen)](https://docling-project.github.io/docling/)\\n[![PyPI version](https://img.shields.io/pypi/v/docling)](https://pypi.org/project/docling/)\\n[![PyPI - Python Version](https://img.shields.io/pypi/pyversions/docling)](https://pypi.org/project/docling/)\\n[![uv](https://img.shields.io/endpoint?url=https://raw.githubusercontent.com/astral-sh/uv/main/assets/badge/v0.json)](https://github.com/astral-sh/uv)\\n[![Ruff](https://img.shields.io/endpoint?url=https://raw.githubusercontent.com/astral-sh/ruff/main/assets/badge/v2.json)](https://github.com/astral-sh/ruff)\\n[![Pydantic v2](https://img.shields.io/endpoint?url=https://raw.githubusercontent.com/pydantic/pydantic/main/docs/badge/v2.json)](https://pydantic.dev)\\n[![prek](https://img.shields.io/badge/prek-enabled-brightgreen)](https://pypi.org/project/prek/)\\n[![License MIT](https://img.shields.io/github/license/docling-project/docling)](https://opensource.org/licenses/MIT)\\n[![PyPI Downloads](https://static.pepy.tech/badge/docling/month)](https://pepy.tech/projects/docling)\\n[![Docling Actor](https://apify.com/actor-badge?actor=vancura/docling&fpr=docling)](https://apify.com/vancura/docling)\\n[![Chat with Dosu](https://dosu.dev/dosu-chat-badge.svg)](https://app.dosu.dev/097760a8-135e-4789-8234-90c8837d7f1c/ask?utm_source=github)\\n[![Discord](https://img.shields.io/discord/1399788921306746971?color=6A7EC2&logo=discord&logoColor=ffffff)](https://docling.ai/discord)\\n[![OpenSSF Best Practices](https://www.bestpractices.dev/projects/10101/badge)](https://www.bestpractices.dev/projects/10101)\\n[![LF AI & Data](https://img.shields.io/badge/LF%20AI%20%26%20Data-003778?logo=linuxfoundation&logoColor=fff&color=0094ff&labelColor=003778)](https://lfaidata.foundation/projects/)\\n\\n## What is Docling ?\\n\\nDocling simplifies document processing, parsing diverse formats — including advanced PDF understanding — and providing seamless integrations with the gen AI ecosystem.\\n\\n## Features\\n\\n- 🗂️ Parsing of [multiple document formats][supported_formats] incl. PDF, DOCX, PPTX, XLSX, HTML, WAV, MP3, WebVTT, images (PNG, TIFF, JPEG, ...), LaTeX, plain text, and more\\n- 📑 Advanced PDF understanding incl. page layout, reading order, table structure, code, formulas, image classification, and more\\n- 🧬 Unified, expressive [DoclingDocument][docling_document] representation format\\n- ↪️ Various [export formats][supported_formats] and options, including Markdown, HTML, WebVTT, [DocTags](https://arxiv.org/abs/2503.11576) and lossless JSON\\n- 📜 Support of several application-specifc XML schemas incl. [USPTO](https://www.uspto.gov/patents) patents, [JATS](https://jats.nlm.nih.gov/) articles, and [XBRL](https://www.xbrl.org/) financial reports.\\n- 🔒 Local execution capabilities for sensitive data and air-gapped environments\\n- 🤖 Plug-and-play [integrations][integrations] incl. LangChain, LlamaIndex, Crew AI & Haystack for agentic AI\\n- 🔍 Extensive OCR support for scanned PDFs and images\\n- 👓 Support of several Visual Language Models ([GraniteDocling](https://huggingface.co/ibm-granite/granite-docling-258M))\\n- 🎙️ Audio support with Automatic Speech Recognition (ASR) models\\n- 🔌 Connect to any agent using the [MCP server](https://docling-project.github.io/docling/usage/mcp/)\\n- 💻 Simple and convenient CLI\\n\\n### What's new\\n\\n- 📤 Structured [information extraction][extraction] \\\\[🧪 beta\\\\]\\n- 📑 New layout model (**Heron**) by default, for faster PDF parsing\\n- 🔌 [MCP server](https://docling-project.github.io/docling/usage/mcp/) for agentic applications\\n- 💼 Parsing of XBRL (eXtensible Business Reporting Language) documents for financial reports\\n- 💬 Parsing of WebVTT (Web Video Text Tracks) files and export to WebVTT format\\n- 💬 Parsing of LaTeX files\\n- 📝 Parsing of plain-text files (`.txt`, `.text`) and Markdown supersets (`.qmd`, `.Rmd`)\\n- 📝 Chart understanding (Barchart, Piechart, LinePlot): converting them into tables, code or adding detailed descriptions\\n\\n### Coming soon\\n\\n- 📝 Metadata extraction, including title, authors, references & language\\n- 📝 Complex chemistry understanding (Molecular structures)\\n\\n## Quickstart\\n\\n### 1. Install\\n\\n```bash\\npip install docling\\n```\\n\\n> **Note:** Python 3.9 support was dropped in docling version 2.70.0. Please use Python 3.10 or higher.\\n\\nWorks on macOS, Linux and Windows environments. Both x86_64 and arm64 architectures.\\n\\nMore [detailed installation instructions](https://docling-project.github.io/docling/installation/) are available in the docs.\\n\\n## 2. Convert a document (CLI)\\n\\n```bash\\ndocling https://arxiv.org/pdf/2206.01062\\n```\\n\\nThis generates a .md file in the current directory containing structured document content.\\n\\nYou can also use 🥚[GraniteDocling](https://huggingface.co/ibm-granite/granite-docling-258M) and other VLMs via Docling CLI:\\n\\n```bash\\ndocling --pipeline vlm --vlm-model granite_docling https://arxiv.org/pdf/2206.01062\\n```\\n\\n## 3. Python usage (recommended)\\n\\n```python\\nfrom docling.document_converter import DocumentConverter\\n\\nsource = \\\"https://arxiv.org/pdf/2408.09869\\\" # document per local path or URL\\nconverter = DocumentConverter()\\nresult = converter.convert(source)\\nprint(result.document.export_to_markdown()) # output: \\\"## Docling Technical Report[...]\\\"\\n```\\n\\nMore advanced [usage](https://docling-project.github.io/docling/usage/) and [configuration](https://docling-project.github.io/docling/installation/) options.\\n\\n## Documentation\\n\\nCheck out Docling's [documentation](https://docling-project.github.io/docling/), for details on\\ninstallation, usage, concepts, recipes, extensions, and more.\\n\\n## Examples\\n\\nGo hands-on with our [examples](https://docling-project.github.io/docling/examples/),\\ndemonstrating how to address different application use cases with Docling.\\n\\n## Integrations\\n\\nTo further accelerate your AI application development, check out Docling's native\\n[integrations](https://docling-project.github.io/docling/integrations/) with popular frameworks\\nand tools.\\n\\n## Get help and support\\n\\nPlease feel free to connect with us using the [discussion section](https://github.com/docling-project/docling/discussions).\\n\\n## Technical report\\n\\nFor more details on Docling's inner workings, check out the [Docling Technical Report](https://arxiv.org/abs/2408.09869).\\n\\n## Contributing\\n\\nPlease read [Contributing to Docling](https://github.com/docling-project/docling/blob/main/CONTRIBUTING.md) for details.\\n\\n## References\\n\\nIf you use Docling in your projects, please consider citing the following:\\n\\n```bib\\n@techreport{Docling,\\n author = {Deep Search Team},\\n month = {8},\\n title = {Docling Technical Report},\\n url = {https://arxiv.org/abs/2408.09869},\\n eprint = {2408.09869},\\n doi = {10.48550/arXiv.2408.09869},\\n version = {1.0.0},\\n year = {2024}\\n}\\n```\\n\\n## License\\n\\nThe Docling codebase is under MIT license.\\nFor individual model usage, please refer to the model licenses found in the original packages.\\n\\n## LF AI & Data\\n\\nDocling is hosted as a project in the [LF AI & Data Foundation](https://lfaidata.foundation/projects/).\\n\\n### IBM ❤️ Open Source AI\\n\\nThe project was started by the AI for knowledge team at IBM Research Zurich.\\n\\n[supported_formats]: https://docling-project.github.io/docling/usage/supported_formats/\\n[docling_document]: https://docling-project.github.io/docling/concepts/docling_document/\\n[integrations]: https://docling-project.github.io/docling/integrations/\\n[extraction]: https://docling-project.github.io/docling/examples/extraction/\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "SWpTLXC3WqBpowshoTAJ5mZKP8i1c81t", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "OTpeFNQaJJYl1yeyYhPazf8D3l8lSIWC", "call_id": "OTpeFNQaJJYl1yeyYhPazf8D3l8lSIWC", "response_item_id": "fc_OTpeFNQaJJYl1yeyYhPazf8D3l8lSIWC", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"# Mehr Details zu Docling - Benchmarks und Features\\ncurl -s https://raw.githubusercontent.com/docling-project/docling/main/README.md 2>&1 | tail -200\",\"timeout\":15}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"<p align=\\\"center\\\">\\n <a href=\\\"https://github.com/docling-project/docling\\\">\\n <img loading=\\\"lazy\\\" alt=\\\"Docling\\\" src=\\\"https://github.com/docling-project/docling/raw/main/docs/assets/docling_processing.png\\\" width=\\\"100%\\\"/>\\n </a>\\n</p>\\n\\n# Docling\\n\\n<p align=\\\"center\\\">\\n <a href=\\\"https://trendshift.io/repositories/12132\\\" target=\\\"_blank\\\"><img src=\\\"https://trendshift.io/api/badge/repositories/12132\\\" alt=\\\"DS4SD%2Fdocling | Trendshift\\\" style=\\\"width: 250px; height: 55px;\\\" width=\\\"250\\\" height=\\\"55\\\"/></a>\\n</p>\\n\\n[![arXiv](https://img.shields.io/badge/arXiv-2408.09869-b31b1b.svg)](https://arxiv.org/abs/2408.09869)\\n[![Docs](https://img.shields.io/badge/docs-live-brightgreen)](https://docling-project.github.io/docling/)\\n[![PyPI version](https://img.shields.io/pypi/v/docling)](https://pypi.org/project/docling/)\\n[![PyPI - Python Version](https://img.shields.io/pypi/pyversions/docling)](https://pypi.org/project/docling/)\\n[![uv](https://img.shields.io/endpoint?url=https://raw.githubusercontent.com/astral-sh/uv/main/assets/badge/v0.json)](https://github.com/astral-sh/uv)\\n[![Ruff](https://img.shields.io/endpoint?url=https://raw.githubusercontent.com/astral-sh/ruff/main/assets/badge/v2.json)](https://github.com/astral-sh/ruff)\\n[![Pydantic v2](https://img.shields.io/endpoint?url=https://raw.githubusercontent.com/pydantic/pydantic/main/docs/badge/v2.json)](https://pydantic.dev)\\n[![prek](https://img.shields.io/badge/prek-enabled-brightgreen)](https://pypi.org/project/prek/)\\n[![License MIT](https://img.shields.io/github/license/docling-project/docling)](https://opensource.org/licenses/MIT)\\n[![PyPI Downloads](https://static.pepy.tech/badge/docling/month)](https://pepy.tech/projects/docling)\\n[![Docling Actor](https://apify.com/actor-badge?actor=vancura/docling&fpr=docling)](https://apify.com/vancura/docling)\\n[![Chat with Dosu](https://dosu.dev/dosu-chat-badge.svg)](https://app.dosu.dev/097760a8-135e-4789-8234-90c8837d7f1c/ask?utm_source=github)\\n[![Discord](https://img.shields.io/discord/1399788921306746971?color=6A7EC2&logo=discord&logoColor=ffffff)](https://docling.ai/discord)\\n[![OpenSSF Best Practices](https://www.bestpractices.dev/projects/10101/badge)](https://www.bestpractices.dev/projects/10101)\\n[![LF AI & Data](https://img.shields.io/badge/LF%20AI%20%26%20Data-003778?logo=linuxfoundation&logoColor=fff&color=0094ff&labelColor=003778)](https://lfaidata.foundation/projects/)\\n\\n## What is Docling ?\\n\\nDocling simplifies document processing, parsing diverse formats — including advanced PDF understanding — and providing seamless integrations with the gen AI ecosystem.\\n\\n## Features\\n\\n- 🗂️ Parsing of [multiple document formats][supported_formats] incl. PDF, DOCX, PPTX, XLSX, HTML, WAV, MP3, WebVTT, images (PNG, TIFF, JPEG, ...), LaTeX, plain text, and more\\n- 📑 Advanced PDF understanding incl. page layout, reading order, table structure, code, formulas, image classification, and more\\n- 🧬 Unified, expressive [DoclingDocument][docling_document] representation format\\n- ↪️ Various [export formats][supported_formats] and options, including Markdown, HTML, WebVTT, [DocTags](https://arxiv.org/abs/2503.11576) and lossless JSON\\n- 📜 Support of several application-specifc XML schemas incl. [USPTO](https://www.uspto.gov/patents) patents, [JATS](https://jats.nlm.nih.gov/) articles, and [XBRL](https://www.xbrl.org/) financial reports.\\n- 🔒 Local execution capabilities for sensitive data and air-gapped environments\\n- 🤖 Plug-and-play [integrations][integrations] incl. LangChain, LlamaIndex, Crew AI & Haystack for agentic AI\\n- 🔍 Extensive OCR support for scanned PDFs and images\\n- 👓 Support of several Visual Language Models ([GraniteDocling](https://huggingface.co/ibm-granite/granite-docling-258M))\\n- 🎙️ Audio support with Automatic Speech Recognition (ASR) models\\n- 🔌 Connect to any agent using the [MCP server](https://docling-project.github.io/docling/usage/mcp/)\\n- 💻 Simple and convenient CLI\\n\\n### What's new\\n\\n- 📤 Structured [information extraction][extraction] \\\\[🧪 beta\\\\]\\n- 📑 New layout model (**Heron**) by default, for faster PDF parsing\\n- 🔌 [MCP server](https://docling-project.github.io/docling/usage/mcp/) for agentic applications\\n- 💼 Parsing of XBRL (eXtensible Business Reporting Language) documents for financial reports\\n- 💬 Parsing of WebVTT (Web Video Text Tracks) files and export to WebVTT format\\n- 💬 Parsing of LaTeX files\\n- 📝 Parsing of plain-text files (`.txt`, `.text`) and Markdown supersets (`.qmd`, `.Rmd`)\\n- 📝 Chart understanding (Barchart, Piechart, LinePlot): converting them into tables, code or adding detailed descriptions\\n\\n### Coming soon\\n\\n- 📝 Metadata extraction, including title, authors, references & language\\n- 📝 Complex chemistry understanding (Molecular structures)\\n\\n## Quickstart\\n\\n### 1. Install\\n\\n```bash\\npip install docling\\n```\\n\\n> **Note:** Python 3.9 support was dropped in docling version 2.70.0. Please use Python 3.10 or higher.\\n\\nWorks on macOS, Linux and Windows environments. Both x86_64 and arm64 architectures.\\n\\nMore [detailed installation instructions](https://docling-project.github.io/docling/installation/) are available in the docs.\\n\\n## 2. Convert a document (CLI)\\n\\n```bash\\ndocling https://arxiv.org/pdf/2206.01062\\n```\\n\\nThis generates a .md file in the current directory containing structured document content.\\n\\nYou can also use 🥚[GraniteDocling](https://huggingface.co/ibm-granite/granite-docling-258M) and other VLMs via Docling CLI:\\n\\n```bash\\ndocling --pipeline vlm --vlm-model granite_docling https://arxiv.org/pdf/2206.01062\\n```\\n\\n## 3. Python usage (recommended)\\n\\n```python\\nfrom docling.document_converter import DocumentConverter\\n\\nsource = \\\"https://arxiv.org/pdf/2408.09869\\\" # document per local path or URL\\nconverter = DocumentConverter()\\nresult = converter.convert(source)\\nprint(result.document.export_to_markdown()) # output: \\\"## Docling Technical Report[...]\\\"\\n```\\n\\nMore advanced [usage](https://docling-project.github.io/docling/usage/) and [configuration](https://docling-project.github.io/docling/installation/) options.\\n\\n## Documentation\\n\\nCheck out Docling's [documentation](https://docling-project.github.io/docling/), for details on\\ninstallation, usage, concepts, recipes, extensions, and more.\\n\\n## Examples\\n\\nGo hands-on with our [examples](https://docling-project.github.io/docling/examples/),\\ndemonstrating how to address different application use cases with Docling.\\n\\n## Integrations\\n\\nTo further accelerate your AI application development, check out Docling's native\\n[integrations](https://docling-project.github.io/docling/integrations/) with popular frameworks\\nand tools.\\n\\n## Get help and support\\n\\nPlease feel free to connect with us using the [discussion section](https://github.com/docling-project/docling/discussions).\\n\\n## Technical report\\n\\nFor more details on Docling's inner workings, check out the [Docling Technical Report](https://arxiv.org/abs/2408.09869).\\n\\n## Contributing\\n\\nPlease read [Contributing to Docling](https://github.com/docling-project/docling/blob/main/CONTRIBUTING.md) for details.\\n\\n## References\\n\\nIf you use Docling in your projects, please consider citing the following:\\n\\n```bib\\n@techreport{Docling,\\n author = {Deep Search Team},\\n month = {8},\\n title = {Docling Technical Report},\\n url = {https://arxiv.org/abs/2408.09869},\\n eprint = {2408.09869},\\n doi = {10.48550/arXiv.2408.09869},\\n version = {1.0.0},\\n year = {2024}\\n}\\n```\\n\\n## License\\n\\nThe Docling codebase is under MIT license.\\nFor individual model usage, please refer to the model licenses found in the original packages.\\n\\n## LF AI & Data\\n\\nDocling is hosted as a project in the [LF AI & Data Foundation](https://lfaidata.foundation/projects/).\\n\\n### IBM ❤️ Open Source AI\\n\\nThe project was started by the AI for knowledge team at IBM Research Zurich.\\n\\n[supported_formats]: https://docling-project.github.io/docling/usage/supported_formats/\\n[docling_document]: https://docling-project.github.io/docling/concepts/docling_document/\\n[integrations]: https://docling-project.github.io/docling/integrations/\\n[extraction]: https://docling-project.github.io/docling/examples/extraction/\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "OTpeFNQaJJYl1yeyYhPazf8D3l8lSIWC", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "B05drB7eTIUJWQFUyyqCySU6tRfRYDbS", "call_id": "B05drB7eTIUJWQFUyyqCySU6tRfRYDbS", "response_item_id": "fc_B05drB7eTIUJWQFUyyqCySU6tRfRYDbS", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"# Suche nach Benchmarks zwischen Marker und Docling\\ncurl -s \\\"https://raw.githubusercontent.com/docling-project/docling/main/docs/usage/benchmark.md\\\" 2>&1 | head -100\",\"timeout\":15}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"404: Not Found\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "B05drB7eTIUJWQFUyyqCySU6tRfRYDbS", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "UQYUAMouCp2MflWLup3zg6jF62PWoVNu", "call_id": "UQYUAMouCp2MflWLup3zg6jF62PWoVNu", "response_item_id": "fc_UQYUAMouCp2MflWLup3zg6jF62PWoVNu", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"# Suche nach Benchmarks und Vergleichen\\ncurl -s \\\"https://raw.githubusercontent.com/docling-project/docling/main/README.md\\\" 2>&1 | grep -i \\\"benchmark\\\\|compare\\\\|marker\\\\|vs\\\\|vs.\\\\|performance\\\\|speed\\\\|accuracy\\\" | head -20\",\"timeout\":15}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "UQYUAMouCp2MflWLup3zg6jF62PWoVNu", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "WfSt4cLNDRjfRXJc4W1Dau5eDEJvlFLh", "call_id": "WfSt4cLNDRjfRXJc4W1Dau5eDEJvlFLh", "response_item_id": "fc_WfSt4cLNDRjfRXJc4W1Dau5eDEJvlFLh", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"# Suche nach Docling vs Marker Vergleichen auf Reddit/StackOverflow\\ncurl -s \\\"https://api.github.com/search/issues?q=docling+vs+marker+PDF+OCR+in:body&sort=created&order=desc&per_page=5\\\" 2>&1 | python3 -c \\\"\\nimport sys, json\\ndata = json.load(sys.stdin)\\nfor i in data.get('items', []):\\n print(f'{i[\\\\\\\"repository_url\\\\\\\"].split(\\\\\\\"/\\\\\\\")[-1]}/{i[\\\\\\\"number\\\\\\\"]}: {i[\\\\\\\"title\\\\\\\"][:150]}')\\n print(f' Created: {i[\\\\\\\"created_at\\\\\\\"]}')\\n print()\\n\\\"\",\"timeout\":15}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"drover/27: feat(cli): add drover organize subcommand for classify-tag-move\\n Created: 2026-05-09T13:33:58Z\\n\\ntinyagentos/321: Catalog expansion roadmap — broad AI-tool coverage across categories\\n Created: 2026-05-06T19:21:20Z\\n\\nrevvel-standards/9322: [WR] ADD INTO REVVEL-STANDARDS WHOLE MODULE FOR discovering and using new tools across each category — apps, CLIs, APIs, MCPs, and PDFs — broken dow\\n Created: 2026-05-04T17:57:39Z\\n\\ndocfold/8: feat: add Microsoft MarkItDown engine adapter + benchmark coverage\\n Created: 2026-04-25T18:20:30Z\\n\\nglossAPI/98: corpus.clean / clean_ocr — audit LaTeX handling for math-enhancer outputs and unify cleanup ideas across the two cleaning paths\\n Created: 2026-04-25T12:22:25Z\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "WfSt4cLNDRjfRXJc4W1Dau5eDEJvlFLh", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "E0oZHFz0StB0tYKLaOjfpTC4mMZLaidr", "call_id": "E0oZHFz0StB0tYKLaOjfpTC4mMZLaidr", "response_item_id": "fc_E0oZHFz0StB0tYKLaOjfpTC4mMZLaidr", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"# Suche nach weiteren alternativen: Nougat (Meta), PDFplumber, Markitdown\\nfor repo in \\\"facebookresearch/nougat\\\" \\\"microsoft/MarkItDown\\\" \\\"pymupdf/PyMuPDF\\\" \\\"LayoutParser/layoutparser\\\" \\\"IBM/pix2text\\\"; do\\n curl -s \\\"https://api.github.com/repos/$repo\\\" 2>&1 | python3 -c \\\"\\nimport sys, json\\ntry:\\n d = json.load(sys.stdin)\\n print(f'{d[\\\\\\\"full_name\\\\\\\"]} | ⭐{d[\\\\\\\"stargazers_count\\\\\\\"]} | {d.get(\\\\\\\"description\\\\\\\",\\\\\\\"?\\\\\\\")[:150]}')\\nexcept: pass\\n\\\" 2>&1\\ndone\",\"timeout\":30}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"facebookresearch/nougat | ⭐9952 | Implementation of Nougat Neural Optical Understanding for Academic Documents\\nmicrosoft/markitdown | ⭐122305 | Python tool for converting files and office documents to Markdown.\\npymupdf/PyMuPDF | ⭐9655 | PyMuPDF is a high performance Python library for data extraction, analysis, conversion & manipulation of PDF (and other) documents.\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "E0oZHFz0StB0tYKLaOjfpTC4mMZLaidr", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "j5uIJMRU7Wub8U2AaQmElK8qVkExTy7b", "call_id": "j5uIJMRU7Wub8U2AaQmElK8qVkExTy7b", "response_item_id": "fc_j5uIJMRU7Wub8U2AaQmElK8qVkExTy7b", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"# IBM Pix2Text - chinesische OCR, aber interessant\\ncurl -s \\\"https://api.github.com/repos/IBM/pix2text\\\" 2>&1 | python3 -c \\\"\\nimport sys, json\\nd = json.load(sys.stdin)\\nprint(f'{d[\\\\\\\"full_name\\\\\\\"]} | ⭐{d[\\\\\\\"stargazers_count\\\\\\\"]}')\\nprint(f'Description: {d.get(\\\\\\\"description\\\\\\\",\\\\\\\"\\\\\\\")}')\\nprint(f'Last push: {d[\\\\\\\"pushed_at\\\\\\\"]}')\\nprint(f'License: {d.get(\\\\\\\"license\\\\\\\",{}).get(\\\\\\\"spdx_id\\\\\\\",\\\\\\\"unknown\\\\\\\")}')\\n\\\" 2>&1\",\"timeout\":15}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"Traceback (most recent call last):\\n File \\\"<string>\\\", line 4, in <module>\\nKeyError: 'full_name'\", \"exit_code\": 1, \"error\": null}", "tool_call_id": "j5uIJMRU7Wub8U2AaQmElK8qVkExTy7b", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "Wj2uBwBiIXiT2tEAIPHOv0ettXAKgw1y", "call_id": "Wj2uBwBiIXiT2tEAIPHOv0ettXAKgw1y", "response_item_id": "fc_Wj2uBwBiIXiT2tEAIPHOv0ettXAKgw1y", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"# Nougat (Meta) - veraltet?\\ncurl -s \\\"https://api.github.com/repos/facebookresearch/nougat\\\" 2>&1 | python3 -c \\\"\\nimport sys, json\\ntry:\\n d = json.load(sys.stdin)\\n print(f'{d[\\\\\\\"full_name\\\\\\\"]} | ⭐{d[\\\\\\\"stargazers_count\\\\\\\"]}')\\n print(f'Description: {d.get(\\\\\\\"description\\\\\\\",\\\\\\\"\\\\\\\")}')\\n print(f'Last push: {d[\\\\\\\"pushed_at\\\\\\\"]}')\\n print(f'License: {d.get(\\\\\\\"license\\\\\\\",{}).get(\\\\\\\"spdx_id\\\\\\\",\\\\\\\"unknown\\\\\\\")}')\\nexcept Exception as e:\\n print(f'Error: {e}')\\n\\\" 2>&1\",\"timeout\":15}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"facebookresearch/nougat | ⭐9952\\nDescription: Implementation of Nougat Neural Optical Understanding for Academic Documents\\nLast push: 2025-02-21T16:38:00Z\\nLicense: MIT\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "Wj2uBwBiIXiT2tEAIPHOv0ettXAKgw1y", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "GoHVVkJujreVinqtXRMNttGXMAepgrJZ", "call_id": "GoHVVkJujreVinqtXRMNttGXMAepgrJZ", "response_item_id": "fc_GoHVVkJujreVinqtXRMNttGXMAepgrJZ", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"# MarkItDown von Microsoft - was kann es?\\ncurl -s https://raw.githubusercontent.com/microsoft/MarkItDown/main/README.md 2>&1 | head -100\",\"timeout\":15}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"# MarkItDown\\n\\n[![PyPI](https://img.shields.io/pypi/v/markitdown.svg)](https://pypi.org/project/markitdown/)\\n![PyPI - Downloads](https://img.shields.io/pypi/dd/markitdown)\\n[![Built by AutoGen Team](https://img.shields.io/badge/Built%20by-AutoGen%20Team-blue)](https://github.com/microsoft/autogen)\\n\\n> [!IMPORTANT]\\n> MarkItDown performs I/O with the privileges of the current process. Like open() or requests.get(), it will access resources that the process itself can access. Sanitize your inputs in untrusted environments, and call the narrowest `convert_*` function needed for your use case (e.g., `convert_stream()`, or `convert_local()`). See the [Security Considerations](#security-considerations) section of the documentation for more information.\\n\\nMarkItDown is a lightweight Python utility for converting various files to Markdown for use with LLMs and related text analysis pipelines. To this end, it is most comparable to [textract](https://github.com/deanmalmgren/textract), but with a focus on preserving important document structure and content as Markdown (including: headings, lists, tables, links, etc.) While the output is often reasonably presentable and human-friendly, it is meant to be consumed by text analysis tools -- and may not be the best option for high-fidelity document conversions for human consumption.\\n\\nMarkItDown currently supports the conversion from:\\n\\n- PDF\\n- PowerPoint\\n- Word\\n- Excel\\n- Images (EXIF metadata and OCR)\\n- Audio (EXIF metadata and speech transcription)\\n- HTML\\n- Text-based formats (CSV, JSON, XML)\\n- ZIP files (iterates over contents)\\n- Youtube URLs\\n- EPubs\\n- ... and more!\\n\\n## Why Markdown?\\n\\nMarkdown is extremely close to plain text, with minimal markup or formatting, but still\\nprovides a way to represent important document structure. Mainstream LLMs, such as\\nOpenAI's GPT-4o, natively \\\"_speak_\\\" Markdown, and often incorporate Markdown into their\\nresponses unprompted. This suggests that they have been trained on vast amounts of\\nMarkdown-formatted text, and understand it well. As a side benefit, Markdown conventions\\nare also highly token-efficient.\\n\\n## Prerequisites\\nMarkItDown requires Python 3.10 or higher. It is recommended to use a virtual environment to avoid dependency conflicts.\\n\\nWith the standard Python installation, you can create and activate a virtual environment using the following commands:\\n\\n```bash\\npython -m venv .venv\\nsource .venv/bin/activate\\n```\\n\\nIf using `uv`, you can create a virtual environment with:\\n\\n```bash\\nuv venv --python=3.12 .venv\\nsource .venv/bin/activate\\n# NOTE: Be sure to use 'uv pip install' rather than just 'pip install' to install packages in this virtual environment\\n```\\n\\nIf you are using Anaconda, you can create a virtual environment with:\\n\\n```bash\\nconda create -n markitdown python=3.12\\nconda activate markitdown\\n```\\n\\n## Installation\\n\\nTo install MarkItDown, use pip: `pip install 'markitdown[all]'`. Alternatively, you can install it from the source:\\n\\n```bash\\ngit clone git@github.com:microsoft/markitdown.git\\ncd markitdown\\npip install -e 'packages/markitdown[all]'\\n```\\n\\n## Usage\\n\\n### Command-Line\\n\\n```bash\\nmarkitdown path-to-file.pdf > document.md\\n```\\n\\nOr use `-o` to specify the output file:\\n\\n```bash\\nmarkitdown path-to-file.pdf -o document.md\\n```\\n\\nYou can also pipe content:\\n\\n```bash\\ncat path-to-file.pdf | markitdown\\n```\\n\\n### Optional Dependencies\\nMarkItDown has optional dependencies for activating various file formats. Earlier in this document, we installed all optional dependencies with the `[all]` option. However, you can also install them individually for more control. For example:\\n\\n```bash\\npip install 'markitdown[pdf, docx, pptx]'\\n```\\n\\nwill install only the dependencies for PDF, DOCX, and PPTX files.\\n\\nAt the moment, the following optional dependencies are available:\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "GoHVVkJujreVinqtXRMNttGXMAepgrJZ", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "fwtWdnQfraqOvPHcBbp1YSg68F2pMkeQ", "call_id": "fwtWdnQfraqOvPHcBbp1YSg68F2pMkeQ", "response_item_id": "fc_fwtWdnQfraqOvPHcBbp1YSg68F2pMkeQ", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"# Check for Surya OCR standalone - das ist die OCR-Engine hinter Marker\\ncurl -s \\\"https://api.github.com/repos/VikParuchuri/surya\\\" 2>&1 | python3 -c \\\"\\nimport sys, json\\nd = json.load(sys.stdin)\\nprint(f'{d[\\\\\\\"full_name\\\\\\\"]} | ⭐{d[\\\\\\\"stargazers_count\\\\\\\"]}')\\nprint(f'Description: {d.get(\\\\\\\"description\\\\\\\",\\\\\\\"\\\\\\\")}')\\nprint(f'Last push: {d[\\\\\\\"pushed_at\\\\\\\"]}')\\nprint(f'License: {d.get(\\\\\\\"license\\\\\\\",{}).get(\\\\\\\"spdx_id\\\\\\\",\\\\\\\"unknown\\\\\\\")}')\\n\\\" 2>&1\",\"timeout\":15}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"Traceback (most recent call last):\\n File \\\"<string>\\\", line 4, in <module>\\nKeyError: 'full_name'\", \"exit_code\": 1, \"error\": null}", "tool_call_id": "fwtWdnQfraqOvPHcBbp1YSg68F2pMkeQ", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "ckR8Xt2i4Lv7fhdoTaX74iIB06OvIh38", "call_id": "ckR8Xt2i4Lv7fhdoTaX74iIB06OvIh38", "response_item_id": "fc_ckR8Xt2i4Lv7fhdoTaX74iIB06OvIh38", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"curl -s \\\"https://api.github.com/repos/VikParuchuri/surya\\\" 2>&1 | python3 -c \\\"\\nimport sys, json\\ntry:\\n d = json.load(sys.stdin)\\n print(f'{d[\\\\\\\"full_name\\\\\\\"]} | ⭐{d[\\\\\\\"stargazers_count\\\\\\\"]}')\\n print(f'Description: {d.get(\\\\\\\"description\\\\\\\",\\\\\\\"\\\\\\\")}')\\n print(f'Last push: {d[\\\\\\\"pushed_at\\\\\\\"]}')\\nexcept Exception as e:\\n print(f'Error: {e}')\\n\\\"\",\"timeout\":15}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"Error: 'full_name'\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "ckR8Xt2i4Lv7fhdoTaX74iIB06OvIh38", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "hcHGATAC24X68VwRgG7erAbaetI0rvTY", "call_id": "hcHGATAC24X68VwRgG7erAbaetI0rvTY", "response_item_id": "fc_hcHGATAC24X68VwRgG7erAbaetI0rvTY", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"curl -s \\\"https://api.github.com/repos/VikParuchuri/surya\\\" 2>&1 | head -5\",\"timeout\":15}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"{\\n \\\"message\\\": \\\"Moved Permanently\\\",\\n \\\"url\\\": \\\"https://api.github.com/repositories/741297064\\\",\\n \\\"documentation_url\\\": \\\"https://docs.github.com/rest/guides/best-practices-for-using-the-rest-api#follow-redirects\\\"\\n}\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "hcHGATAC24X68VwRgG7erAbaetI0rvTY", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "42S934ZrD18ewOHEcaorNwLAYw0fWmXp", "call_id": "42S934ZrD18ewOHEcaorNwLAYw0fWmXp", "response_item_id": "fc_42S934ZrD18ewOHEcaorNwLAYw0fWmXp", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"curl -sL \\\"https://api.github.com/repos/datalab-to/surya\\\" 2>&1 | python3 -c \\\"\\nimport sys, json\\nd = json.load(sys.stdin)\\nprint(f'{d[\\\\\\\"full_name\\\\\\\"]} | ⭐{d[\\\\\\\"stargazers_count\\\\\\\"]}')\\nprint(f'Description: {d.get(\\\\\\\"description\\\\\\\",\\\\\\\"\\\\\\\")}')\\nprint(f'Last push: {d[\\\\\\\"pushed_at\\\\\\\"]}')\\n\\\"\",\"timeout\":15}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"datalab-to/surya | ⭐19716\\nDescription: OCR, layout analysis, reading order, table recognition in 90+ languages\\nLast push: 2026-05-06T09:18:30Z\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "42S934ZrD18ewOHEcaorNwLAYw0fWmXp", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "DEtUycBhwry5UAfb4StUfLCTCaaqkbGF", "call_id": "DEtUycBhwry5UAfb4StUfLCTCaaqkbGF", "response_item_id": "fc_DEtUycBhwry5UAfb4StUfLCTCaaqkbGF", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"# Suche nach spezifischen Vergleichen Marker vs Docling\\ncurl -s \\\"https://raw.githubusercontent.com/docling-project/docling/main/docs/examples/benchmark.md\\\" 2>&1 | head -100\",\"timeout\":15}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"404: Not Found\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "DEtUycBhwry5UAfb4StUfLCTCaaqkbGF", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "j30DF3apjotXGAassnDie96Swz1IUazJ", "call_id": "j30DF3apjotXGAassnDie96Swz1IUazJ", "response_item_id": "fc_j30DF3apjotXGAassnDie96Swz1IUazJ", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"# Suche nach dem Docling Benchmark\\ncurl -sL \\\"https://docling-project.github.io/docling/usage/benchmark/\\\" 2>&1 | grep -i \\\"marker\\\\|benchmark\\\\|accuracy\\\\|table\\\" | head -20\",\"timeout\":15}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"<a href=\\\"/docling/examples/export_tables/\\\" class=\\\"md-nav__link\\\">\\n Table export\\n Parquet benchmark\\n<nav class=\\\"md-nav md-nav--secondary\\\" aria-label=\\\"Table of contents\\\">\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "j30DF3apjotXGAassnDie96Swz1IUazJ", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "dc2JS2rJ1Y1i6yWqpOI7HYoSafOSStS9", "call_id": "dc2JS2rJ1Y1i6yWqpOI7HYoSafOSStS9", "response_item_id": "fc_dc2JS2rJ1Y1i6yWqpOI7HYoSafOSStS9", "type": "function", "function": {"name": "terminal", "arguments": "{\"command\":\"# Parquet Benchmark von Docling\\ncurl -sL \\\"https://docling-project.github.io/docling/examples/parquet_benchmark/\\\" 2>&1 | grep -A 5 -i \\\"marker\\\\|accuracy\\\\|score\\\" | head -30\",\"timeout\":15}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "terminal", "content": "{\"output\": \"<a href=\\\"/docling/concepts/confidence_scores/\\\" class=\\\"md-nav__link\\\">\\n \\n \\n \\n <span class=\\\"md-ellipsis\\\">\\n \\n--\\n Confidence Scores\\n \\n\\n \\n </span>\", \"exit_code\": 0, \"error\": null}", "tool_call_id": "dc2JS2rJ1Y1i6yWqpOI7HYoSafOSStS9", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "lrokmsHiAyhYBn4rtULDSS1tbnuThYz3", "call_id": "lrokmsHiAyhYBn4rtULDSS1tbnuThYz3", "response_item_id": "fc_lrokmsHiAyhYBn4rtULDSS1tbnuThYz3", "type": "function", "function": {"name": "execute_code", "arguments": "{\"code\":\"import urllib.request\\nimport re\\n\\n# Hole die Docling Parquet Benchmark Seite\\nurl = \\\"https://docling-project.github.io/docling/examples/parquet_benchmark/\\\"\\nreq = urllib.request.Request(url, headers={\\\"User-Agent\\\": \\\"Mozilla/5.0\\\"})\\ntry:\\n with urllib.request.urlopen(req, timeout=10) as resp:\\n html = resp.read().decode(\\\"utf-8\\\", errors=\\\"ignore\\\")\\n # Suche nach Benchmark-Tables\\n tables = re.findall(r'<table[^>]*>.*?</table>', html, re.DOTALL)\\n for t in tables[:3]:\\n # Extrahiere Zeilen\\n rows = re.findall(r'<tr>.*?</tr>', t, re.DOTALL)\\n for r in rows[:5]:\\n cells = re.sub(r'<[^>]+>', '', r)\\n print(cells.strip()[:200])\\n print(\\\"---\\\")\\nexcept Exception as e:\\n print(f\\\"Error: {e}\\\")\"}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "execute_code", "content": "{\"status\": \"success\", \"output\": \"Error: HTTP Error 404: Not Found\\n\", \"tool_calls_made\": 0, \"duration_seconds\": 0.22}", "tool_call_id": "lrokmsHiAyhYBn4rtULDSS1tbnuThYz3", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "lXm0HJWUYyHMDpPmhNzxvrc5KmzCrHht", "call_id": "lXm0HJWUYyHMDpPmhNzxvrc5KmzCrHht", "response_item_id": "fc_lXm0HJWUYyHMDpPmhNzxvrc5KmzCrHht", "type": "function", "function": {"name": "execute_code", "arguments": "{\"code\":\"import urllib.request\\nimport re\\n\\n# Suche nach Benchmarks auf der Docling Website\\nurl = \\\"https://docling-project.github.io/docling/examples/\\\"\\nreq = urllib.request.Request(url, headers={\\\"User-Agent\\\": \\\"Mozilla/5.0\\\"})\\ntry:\\n with urllib.request.urlopen(req, timeout=10) as resp:\\n html = resp.read().decode(\\\"utf-8\\\", errors=\\\"ignore\\\")\\n links = re.findall(r'href=\\\"([^\\\"]*benchmark[^\\\"]*)\\\"', html)\\n for l in links:\\n print(l)\\nexcept Exception as e:\\n print(f\\\"Error: {e}\\\")\"}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "execute_code", "content": "{\"status\": \"success\", \"output\": \"\", \"tool_calls_made\": 0, \"duration_seconds\": 0.63}", "tool_call_id": "lXm0HJWUYyHMDpPmhNzxvrc5KmzCrHht", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "ipVUclrMQRUFRd8fmwUoy0ls39NJizXo", "call_id": "ipVUclrMQRUFRd8fmwUoy0ls39NJizXo", "response_item_id": "fc_ipVUclrMQRUFRd8fmwUoy0ls39NJizXo", "type": "function", "function": {"name": "execute_code", "arguments": "{\"code\":\"import urllib.request\\nimport re\\n\\n# Suche nach Benchmarks auf der Docling Website\\nurl = \\\"https://docling-project.github.io/docling/examples/\\\"\\nreq = urllib.request.Request(url, headers={\\\"User-Agent\\\": \\\"Mozilla/5.0\\\"})\\ntry:\\n with urllib.request.urlopen(req, timeout=10) as resp:\\n html = resp.read().decode(\\\"utf-8\\\", errors=\\\"ignore\\\")\\n # Alle Links finden\\n links = re.findall(r'href=\\\"([^\\\"]+)\\\"', html)\\n for l in links:\\n if 'benchmark' in l.lower() or 'parquet' in l.lower():\\n print(l)\\nexcept Exception as e:\\n print(f\\\"Error: {e}\\\")\"}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "execute_code", "content": "{\"status\": \"success\", \"output\": \"parquet_images/\\n\", \"tool_calls_made\": 0, \"duration_seconds\": 0.43}", "tool_call_id": "ipVUclrMQRUFRd8fmwUoy0ls39NJizXo", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "eAwYETZB44cTUcy5fKV1KDWAzUd20SCw", "call_id": "eAwYETZB44cTUcy5fKV1KDWAzUd20SCw", "response_item_id": "fc_eAwYETZB44cTUcy5fKV1KDWAzUd20SCw", "type": "function", "function": {"name": "execute_code", "arguments": "{\"code\":\"import urllib.request\\nimport re\\n\\n# Parquet images benchmark\\nurl = \\\"https://docling-project.github.io/docling/examples/parquet_images/\\\"\\nreq = urllib.request.Request(url, headers={\\\"User-Agent\\\": \\\"Mozilla/5.0\\\"})\\ntry:\\n with urllib.request.urlopen(req, timeout=10) as resp:\\n html = resp.read().decode(\\\"utf-8\\\", errors=\\\"ignore\\\")\\n # Finde alle Tabellen\\n tables = re.findall(r'<table[^>]*>.*?</table>', html, re.DOTALL)\\n for t in tables[:2]:\\n rows = re.findall(r'<tr>.*?</tr>', t, re.DOTALL)\\n for r in rows[:8]:\\n cells = re.sub(r'<[^>]+>', '', r).strip()\\n if cells:\\n print(cells[:300])\\n print(\\\"===\\\\n\\\")\\nexcept Exception as e:\\n print(f\\\"Error: {e}\\\")\"}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "execute_code", "content": "{\"status\": \"success\", \"output\": \"\", \"tool_calls_made\": 0, \"duration_seconds\": 0.63}", "tool_call_id": "eAwYETZB44cTUcy5fKV1KDWAzUd20SCw", "timestamp": "2026-05-10T09:11:46.946029"} {"role": "assistant", "content": "", "reasoning": null, "finish_reason": "tool_calls", "tool_calls": [{"id": "1wBqa0GfjTjr3gIxsEmv6cIU5xGrKe7Y", "call_id": "1wBqa0GfjTjr3gIxsEmv6cIU5xGrKe7Y", "response_item_id": "fc_1wBqa0GfjTjr3gIxsEmv6cIU5xGrKe7Y", "type": "function", "function": {"name": "execute_code", "arguments": "{\"code\":\"import urllib.request\\nimport re\\n\\n# Parquet images benchmark\\nurl = \\\"https://docling-project.github.io/docling/examples/parquet_images/\\\"\\nreq = urllib.request.Request(url, headers={\\\"User-Agent\\\": \\\"Mozilla/5.0\\\"})\\ntry:\\n with urllib.request.urlopen(req, timeout=10) as resp:\\n html = resp.read().decode(\\\"utf-8\\\", errors=\\\"ignore\\\")\\n # Zeige den gesamten Inhalt\\n print(html[:5000])\\nexcept Exception as e:\\n print(f\\\"Error: {e}\\\")\"}"}}], "timestamp": "2026-05-10T09:11:46.946029"} {"role": "tool", "name": "execute_code", "content": "{\"status\": \"success\", \"output\": \"\\n<!doctype html>\\n<html lang=\\\"en\\\" class=\\\"no-js\\\">\\n <head>\\n \\n <meta charset=\\\"utf-8\\\">\\n <meta name=\\\"viewport\\\" content=\\\"width=device-width,initial-scale=1\\\">\\n \\n \\n \\n <link rel=\\\"canonical\\\" href=\\\"https://docling-project.github.io/docling/examples/parquet_images/\\\">\\n \\n \\n <link rel=\\\"prev\\\" href=\\\"../gpu_vlm_pipeline/\\\">\\n \\n \\n <link rel=\\\"next\\\" href=\\\"../dpk-ingest-chunk-tokenize/\\\">\\n \\n \\n \\n \\n \\n <link rel=\\\"icon\\\" href=\\\"../../assets/logo.png\\\">\\n <meta name=\\\"generator\\\" content=\\\"mkdocs-1.6.1, mkdocs-material-9.7.6\\\">\\n \\n \\n \\n <title>Parquet benchmark - Docling\\n \\n \\n \\n \\n \\n \\n \\n \\n \\n\\n\\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n\\n\\n \\n \\n\\n\\n \\n \\n \\n\\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n \\n
\\n \\n \\n \\n Skip to content\\n \\n \\n
\\n
\\n \\n
\\n \\n \\n \\n\\n
\\n