10 open-source projects that show what influenced an AI agent's decision
The author lists 10 open-source observability projects, including Langfuse, Phoenix and Ragas, to answer which piece of context actually changed the answer.
GPT-6 Astra context stack is getting f...cking illegal 10 open-source GitHub projects for showing an AI agent WHAT actually influenced its decision 01 Langfuse ▸ https://github.com/langfuse/langfuse → traces + prompts + context 02 Arize Phoenix ▸ https://github.com/Arize-ai/phoenix → inspect retrieval + tool calls + traces X-RAY THE CONTEXT 03 Ragas ▸ https://github.com/explodinggradients/ragas → measure whether retrieved context actually helped 04 TruLens ▸ https://github.com/truera/trulens → evaluate context relevance + groundedness 05 DeepEval ▸ https://github.com/confident-ai/deepeval → test the reasoning pipeline TRACE THE KNOWLEDGE 06 Graphiti ▸ https://github.com/getzep/graphiti → track relationships + changing facts 07 Cognee ▸ https://github.com/topoteretes/cognee → data → connected agent knowledge WATCH THE AGENT 08 AgentOps ▸ https://github.com/AgentOps-AI/agentops → replay agent sessions 09 OpenLLMetry ▸ https://github.com/traceloop/openllmetry → trace tools + models + vector databases 10 OpenLIT ▸ https://github.com/openlit/openlit → telemetry across the AI stack the x-ray: input → retrieve → memory → tools → reasoning → decision → verify → trace the interesting question isn't: “what did the AI answer?” it's: which context changed the answer? which memory mattered? which tool call moved the decision? which file caused the failure? once agents start touching hundreds of files and tools seeing the causal path becomes almost as important as the output itself ⭣
This post quotes another post: x.com/promixyz/status/2102424036416348494