edwin-hao-ai/Awareness-SDK
Official SDKs & plugins for Awareness Memory Cloud — persistent, cross-session memory for AI agents. Python, TypeScript, MCP, Claude Code, OpenClaw, Cursor, Windsurf. Drop-in memory layer for LangChain, CrewAI, AutoGen & more.
What's novel
Official SDKs & plugins for Awareness Memory Cloud — persistent, cross-session memory for AI agents. Python, TypeScript, MCP, Claude Code, OpenClaw, Cursor, Windsurf. Drop-in memory layer for LangChain, CrewAI, AutoGen & more.
Code Analysis
8 files read · 3 roundsA local-first persistent memory layer for AI agents that stores session knowledge as markdown files with SQLite FTS5 indexing, exposes MCP tools for recall/record/lookup, and provides Python/TypeScript SDKs with transparent LLM interceptors for automatic memory injection.
Strengths
Well-structured monorepo with clean separation between the local daemon (SQLite + FTS5 + markdown files), MCP protocol layer, and multi-language SDKs. The interceptor pattern for transparent OpenAI/Anthropic memory injection and the 3-layer progressive query rewriting (context-aware → structural keywords → optional LLM) show thoughtful engineering. Atomic file writes, daemon lifecycle management with PID files, workspace scanning, and skill decay timers demonstrate production-grade concerns.
Weaknesses
The core search/recall engine and knowledge extraction logic were not fully readable due to file size, making it hard to verify the depth of the hybrid retrieval algorithms. The project is primarily a well-organized integration layer around SQLite FTS5 and markdown files rather than implementing novel retrieval algorithms — the 'hybrid search' is BM25 + vector embeddings, which is standard. The monorepo has significant code duplication across the _shared, awareness-memory, and claudecode directo
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
31
Contributors
2
Files
113
Active weeks
1
Repository
Language
Python
Stars
2
Forks
0
License
Apache-2.0