给 Astra 补上记忆、检索、工具与追踪的 7 个仓库

Lummox @Lummox_eth原推发布于 2026.10.05

在 X 上打开原推

GPT 6 Astra still needs a f…cking system-block around it 7 open GitHub repos fixing the parts raw intelligence cannot FORGETS EVERYTHING 01 mem0ai/mem0 ▸ http://github.com/mem0ai/mem0 → persistent memory that lets an agent keep useful context across sessions CONTEXT CHANGES 02 getzep/graphiti ▸ http://github.com/getzep/graphiti → temporal graphs remember what is true now, what used to be true RETRIEVES THE WRONG THING 03 qdrant/qdrant ▸ http://github.com/qdrant/qdrant → multivector, filtering, and hybrid search give agents a stronger retrieval layer CAN THINK BUT CANNOT ACT 04 ComposioHQ/composio ▸ http://github.com/ComposioHQ/composio → connects agents to 1000+ toolkits with auth, sessions, triggers 05 modelcontextprotocol/servers ▸ http://github.com/modelcontextprotocol/servers → reference MCP servers show how models can get controlled access to tools FAILS AND NOBODY KNOWS WHY 06 langfuse/langfuse ▸ http://github.com/langfuse/langfuse → traces model calls, retrieval, tool actions, latency, usage, and evaluations ONE AGENT BECOMES THE BOTTLENECK 07 crewAIInc/crewAI ▸ http://github.com/crewAIInc/crewAI → orchestrates multiple specialized agents when one giant agent should not own the task the actual system: memory → changing context → retrieval → tools → execution → traces → coordination GPT 6 Astra can make the brain stronger GitHub is where you build the nervous system around it if I rebuilt an agent tomorrow, I would design the failure path before the happy path

这条是引用着另一条写的：x.com/lummox_eth/status/2100288151897981255