sjlL@dZddlZddlmZddlmZmZmZddlm Z edGdd Z d-d e d e d e ddfdZ d e d e de fdZdedeeeffdZehdZdede fdZddddedzdededefdZdddde d e dedzd e de de defdZd eeeefdeeeeffd!Zd"eeeefdzdeeeeffd#Zd$eeeefd"eeeefdefd%Zd$eeeefd e deefd&Zd'd dd d(d eeeefd"eeeefdzd)ed e dedzd*e de fd+Z d.d eeeefd)ed e dedzdeeeeff d,ZdS)/aAnthropic prompt caching strategy. The default layout uses 4 cache_control breakpoints: the static system prefix, the end of the system prompt, and the last 2 non-system messages. When a static system prefix is unavailable, it falls back to one system breakpoint plus the last 3 messages. All markers use the same TTL (5m or 1h). This preserves intra-session caching while allowing new sessions to reuse the stable system-prompt prefix. Pure functions -- no class state, no AIAgent dependency. N) dataclass)AnyDictList)find_stable_prefixT)frozenc|eZdZUdZeeeefed<eeeefed<e de fdZ dS)PromptCachePlanzARequest-local message and tool sections with their cache markers.messagestoolsreturnc6t|j|jS)zWire-visible cache markers in this plan (computed on demand). Only tests consume this; keeping it lazy avoids walking every message part and tool schema on the per-request hot path. )_count_cache_markersr r )selfs 8/home/agent/.hermes/hermes-agent/agent/prompt_caching.py marker_countzPromptCachePlan.marker_counts$DM4:>>>N) __name__ __module__ __qualname____doc__rrstrr__annotations__propertyintrrrr r ssKK4S>"""" S#X  ?c???X???rr Fmsg cache_markernative_anthropicr c|dd}|d}|dkr |r||d<dS||dkr|dkr|sdS|dkr|sdS||d<dSt|trH|dkr6t|}|%d ||d d |t |dd g|d<dSd ||d g|d<dSt|t r&|r&|d }t|t r ||d<dSdSdSdS) zFAdd cache_control to a single message, handling all format variations.rolecontenttool cache_controlN assistantusertexttyper(r%r*r()get isinstancerrlenlistdict)rrrr!r# stable_prefixlasts r_apply_cache_markerr4%s 7762  Dggi  G v~~*~ ,O'R-- 6>>"2> F ;  '7  F+O'3 6>>.w77M(!' -)5 $WS5G5G5H5H-IJJ "IW| L L I '4  1W1r{ dD ! ! 1$0D ! ! !1111 1 1rc|rdS|d}||dkrdSt|tr*t|ot|dtSt|t S)uTrue if a marker on this message is actually honored by the provider. On the native Anthropic layout every message works (top-level markers are relocated by the adapter). On the envelope layout (OpenRouter et al.) only markers inside content parts are honored: empty-content messages (e.g. assistant turns that are pure tool_calls) and empty tool messages would receive a top-level marker the provider ignores — wasting one of the four breakpoints. Skip those so the breakpoints land on messages that count. Tr#Nr"Fr,)r-r.r0boolr1r)rrr#s r_can_carry_markerr7[swtggi  G'R--u'4  ? G}}>GBK!>!>> gs # ##rttlc$ddi}|dkrd|d<|S)zCBuild a cache_control marker dict for the given TTL ('5m' or '1h').r* ephemeral1hr8r)r8markers r _build_markerr=ss#$k2F d{{u Mr> opencode-go opencode-zenalibabaopencodemodelc2d|pdvS)zTrue when ``model`` names a Qwen-family model (case-insensitive). Shared by the TTL clamp below and ``agent_runtime_helpers.anthropic_prompt_cache_policy`` so the cache-policy opt-in and the clamp can never desync (#84733). qwenr")lower)rBs r is_qwen_modelrFs ekr((** **rr")rBproviderrGcz|dkr|pdSt|rdS|pdtvrdSdS)aClamp a requested cache TTL to what the destination route supports. Qwen/Alibaba context caching documents an explicit five-minute window (renewed on hit); the Anthropic ``1h`` tier is ignored/rejected there, so a configured ``1h`` regresses to ``5m`` instead of shipping a marker the provider drops and creating a false 1h-cache expectation (#84733). All other caching routes keep the requested TTL. ``None`` (caching active with no explicit tier) resolves to ``5m``. r;5mr")rFrEALIBABA_FAMILY_PROVIDERS)r8rBrGs reffective_cache_ttlrKsR  d{{{dUtB#;;;t 4r) mark_suffixfallback_to_wholemessagestatic_system_prefixrLrMcr|d}t|trv|rtt|tr_||rJ|t |d}|rd|d}|r||d<d||d|g|d<|rdndSt ||| dS|sd St ||| dS) uMark the static system prefix (and optionally the full prompt). The system prompt remains one stored string. Splitting it only in the outgoing request keeps session persistence and non-Anthropic transports unchanged while making the stable prefix independently cacheable. ``mark_suffix=False`` is the tool-cache-plan layout: only the static prefix carries a marker, the volatile suffix rides unmarked (its breakpoint budget is spent on the tools array instead). ``fallback_to_whole=False`` skips marking entirely when the prefix split is not possible (no prefix, mismatched prefix, non-string content) instead of marking the whole message. When the prompt IS exactly the static prefix (empty suffix), the whole message is marked as a single block — never a two-part split with an empty text block, which Anthropic rejects. Returns the number of markers applied (0, 1, or 2). r#Nr(r+r%r)rr)r-r.r startswithr/r4) rNrrOrrLrMr#suffix suffix_parts r_apply_system_cache_markersrWs:kk)$$G'--  w $ $   3 4 4  122334  +)/ @ @K z0strip_anthropic_cache_control..s6VVdz$%%A/T*AVVVVVVrcg|];}t|tr"d|vrd|Dn|z..s(GGG$!Q!2F2FA2F2F2Fr)r.r1itemsr]s r z1strip_anthropic_cache_control..sfdD)).=.E.EHG$**,,GGGGrc3K|]}t|toh|dddkoNt|dto&t |ddhkVdS)r*r(N)r.r1r-rsetkeysr]s rr`z0strip_anthropic_cache_control..!s+ +  tT " " 5((F2 5488F++S11 5DIIKK  VV$44 + + + + + + rsystemr"c3&K|] }|dV dS)r(Nrr]s rr`z0strip_anthropic_cache_control..-s&$F$FdT&\$F$F$F$F$F$Fr) r.r1popr-r0r/anyalljoin)rXrr#skill_split_shapedecoration_shapes rstrip_anthropic_cache_controlrss6*G*G#t$$   &&&'')$$'4((   GGFOOv % 2G ! 271:t,, 271:t,, 2 71:-  2  wqz1  VVgVVV V V %$ G %C N"  s+ +  + + + ( (   LLA  !8+AG 0A !    GWW$F$Fg$F$F$FFFC N rr ctj|pg}|D]-}t|tr|dd.|S)zBReturn copied tools without request-local Anthropic cache markers.r%N)copydeepcopyr.r1rm)r cleanedr$s r"strip_anthropic_tool_cache_controlrx1sPmEKR((G,, dD ! ! , HH_d + + + Nrr ctd|D}|td|Dz }|td|DzS)z=Count the wire-visible cache markers in a request-local plan.c3LK|]}t|trd|vdV dSr%rRNr\)r^rNs rr`z'_count_cache_markers..<sM  gt $ $*9G)C)C )C)C)C)Crc3K|]g}t|trPt|dtr(|dD]}t|trd|vdV hdS)r#r%rRN)r.r1r-r0)r^rNr_s rr`z'_count_cache_markers..As  gt $ $*4GKK 4J4JD)Q)QI&   dD ! !  '6&=&= '>&=&=&=&= rc3LK|]}t|trd|vdV dSr{r\)r^r$s rr`z'_count_cache_markers..HsGz$55:IT:Q:Q:Q:Q:Q:Qr)sum)r r counts rrr:s   E  SE 3 rcg}d}|t|kr5||}t|tr|ddkr|dz }P|ddkr|dr|dz}|}|t|krO||}t|tr|ddkrn|dz }|t|kO||kr0|dz }t |||r|||}!|ddkrd|t|krO||}t|tr|ddkrn|dz }|t|kO|ddkr|dzt|kr|dz }|ddkr|d d vr|dz } t ||r|||dz }|t|k5|S) z