Skip to content

Latest commit

 

History

History
70 lines (54 loc) · 5.22 KB

README.md

File metadata and controls

70 lines (54 loc) · 5.22 KB

API Index

Each EverAlgo distribution has its own README with a quick-start, public API surface, prompt customisation guide, and testing patterns.


Per-distribution READMEs

Distribution README What it provides
everalgo-core packages/everalgo-core/README.md ChatMessage, MemCell, Episode, RankInput/RankOutput, LLMClient, OpenAICompatClient, FakeLLMClient
everalgo-boundary packages/everalgo-boundary/README.md BoundaryDetector (adetect_batch / detect_batch, adetect_step / detect_step), Cuts, to_memcells
everalgo-clustering packages/everalgo-clustering/README.md Cluster, cluster_by_geometry, acluster_by_llm, cluster_by_llm
everalgo-rank packages/everalgo-rank/README.md hybrid, agentic, cluster, maxsim, category, business rankers, fusion, weight, rerank
everalgo-parser packages/everalgo-parser/README.md aparse, parse, ParsedContent, Modality, RawFile; image / audio / document / URL parsers (video deferred)
everalgo-user-memory packages/everalgo-user-memory/README.md BoundaryDetector, Cuts, to_memcells (re-exported from everalgo-boundary), EpisodeExtractor, EpisodeReflector, EpisodeSummarizer, ForesightExtractor, AtomicFactExtractor, ProfileExtractor, OutputLanguage
everalgo-agent-memory packages/everalgo-agent-memory/README.md BoundaryDetector, Cuts, to_memcells (re-exported from everalgo-boundary), three extractors, extraction result / skip reason types, OpOutcome
everalgo-knowledge packages/everalgo-knowledge/README.md KnowledgeExtractor, aclassify_category, classify_category; KnowledgeMemory / CategorySpec live in everalgo.types

Integration contracts


Key types at a glance

All shared data contracts are in everalgo.types:

from everalgo.types import (
    ChatMessage,        # single conversation turn — kind="text"
    ToolCallRequest,    # assistant-emitted tool invocation — kind="tool_call"
    ToolCallResult,     # tool execution result — kind="tool_result"
    ConversationItem,   # ChatMessage | ToolCallRequest | ToolCallResult (discriminated union)
    MemCell,            # boundary-segmented conversation slice; items: list[ConversationItem]
    Episode,            # narrative memory — owner_id, episode, subject, summary, timestamp
    Foresight,          # anticipated future event — owner_id, foresight, evidence, start_time, end_time, duration_days
    AtomicFact,         # single verifiable assertion — owner_id (str | None), content, timestamp
    Profile,            # structured user profile — owner_id, summary, timestamp, explicit_info, implicit_traits
    ExplicitInfoItem,   # one Profile.explicit_info item — description, category, evidence, observed_at; extra keys kept
    ImplicitTraitItem,  # one Profile.implicit_traits item — description, trait, basis, observed_at; extra keys kept
    AgentCase,          # distilled agent trajectory — id, timestamp, task_intent, approach, quality_score, key_insight
    AgentSkill,         # aggregated skill — id, cluster_id (caller-stamped), name, description, content, confidence
    AgentProfilePatch,  # one section-level SOUL.md / AGENTS.md edit — file, action, section, old_text, new_text
    AgentProfileUpdate, # AgentProfileExtractor result — patches, soul_diff / agents_diff, new_*_md, signals
    AgentProfileSignal, # below-gate implicit signal — caller persists and feeds back as pending_signals
    KnowledgeMemory,    # extracted knowledge from a file
    RankInput,          # multi-route recall candidates + cross-memory linkage
    RankOutput,         # ranked memory list
    ParsedContent,      # multimodal file parsed into structured content
)

The clustering value object lives in everalgo.clustering:

from everalgo.clustering import Cluster
# Cluster: id (caller-supplied), centroid, count, last_ts, preview, members (caller entity ids)

The LLM contract lives in everalgo.llm:

from everalgo.llm import LLMClient, LLMMessage, LLMOutputError, StrictOutputModel, parse_reply

LLMClient.achat(messages, *, response_format, temperature=None, max_tokens=None, **extra) takes LLMMessages and a StrictOutputModel subclass and returns the reply text; an operator validates it with parse_reply and raises LLMOutputError on a bad reply. everalgo.llm.LLMMessage is the request message sent to the model; everalgo.types.ChatMessage is the domain type (conversation message). They are distinct classes. The full contract is concepts/llm-contract.md.