List view
Goal: Position contextweaver in the AI agent ecosystem with framework-specific integration guides, interop documentation, and adapter recipes. Scope: Framework integration guides (LlamaIndex, OpenAI ADK, Google ADK, LangChain, LangGraph, Pipecat), interop positioning doc + runtime matrix, "zero deps" messaging reframe, FastMCP CodeMode adapter, interop/policy-engine epic. Dependency notes: Depends on stable core (v0.3+) and observability (v0.5) so guides can reference real metrics. #86 (interop epic) and #123 (integration guides epic) are tracking parents. All child issues (#130, #132, #134, #136) from previous splits land here with their epics. Done when: ≥5 framework integration guides published, runtime compatibility matrix documented, interop positioning page live, FastMCP CodeMode adapter working.
No due date•0/13 issues closedFollow-up to v0.2.0 (Core Hardening & DX). Broadens project maturity across three pillars: ### Documentation depth (7 issues) - #21 — MCP annotation security limitations - #24 — Reference agent loop example (all 4 phases) - #81 — "Why Trust contextweaver?" section - #82 — Troubleshooting guide + FAQ - #103 — llms.txt for AI-consumable docs - #105 — Integration cookbook (FastMCP, A2A recipes) - #110 — Auto-generated API reference (mkdocstrings) ### Infrastructure hardening (5 issues) - #49 — Optional-dependencies extras in pyproject.toml - #50 — `contextweaver[tokens]` extra (depends on #49) - #90 — Test coverage reporting (non-gating) - #92 — SECURITY.md with disclosure policy - #94 — CI dependency caching ### Code quality & internal cleanup (6 issues) - #64 — Replace bare ValueError with custom exceptions - #65 — Enforce or remove dead ttl_behavior field - #66 — StoreBundle.from_dict() for symmetric serialization - #96 — Reusable .github/prompts/ for agent tasks - #101 — Decompose ContextManager (≤300-line guideline) - #106 — BuildStats diagnostic report + CLI **Entry criteria:** v0.2.0 shipped **Exit criteria:** All 18 issues closed, `make ci` green
No due date•0/17 issues closedSame as before. MCP proxy/gateway surface, shared ProxyRuntime, gateway specs, tool_view meta-tool, routing pipeline decomposition, embedding-based retrieval. The "production-ready" release.
No due date•0/9 issues closedGoal: Add production observability primitives, quantitative evaluation harnesses, and explore internal extensibility (CLI, YAML, pydantic, networkx). Scope: MetricsCollector/MetricsHook, OpenTelemetry integration, routing/context benchmark harness, CLI extra (typer+rich), YAML config extra, pydantic evaluation spike, networkx evaluation spike, plugin scaffold template. Dependency notes: #57 (OTel) depends on #10 (MetricsCollector) — both in this milestone, correct ordering. #122 (plugin scaffold) depends on #47 (v0.4). #119 (benchmark) depends on stable routing from v0.4. Spikes (#58, #60) are exploratory and inform v1.0 decisions. Done when: MetricsHook emitting production metrics, OTel spans available via extra, benchmark harness measuring routing recall@k, CLI colorized, at least 2 spike reports filed.
No due date•3/12 issues closedLevel up routing: EngineRegistry with pluggable Retriever/Reranker/Clustering protocols, ProfileConfig presets, CatalogNormalizer, structured RouteTrace, GraphManifest, toolset gating, BM25/fuzzy retrieval backends, uncertainty/clarifying questions API, dynamic per-step retrieval. The "smart routing" release.
No due date•1/13 issues closedComplete the store layer and remaining protocol wiring: SQLite-backed stores, JSON file store, EpisodicStore/FactStore protocols, TTL/eviction, timestamp query, CSV drilldown, RuleEngine dispatcher, Summarizer/Extractor LLM plugin. The "feature-complete core" release.
No due date•1/18 issues closedStabilize core: wire Summarizer/Extractor, clean dead config, extras infra, reference examples, README rewrite, CI/infra hardening. The "contributor-ready" release.
No due date•10/12 issues closed