10 GitHub repos to build a full AI system around Astra
The author lists ten GitHub repos, from LangGraph for the agent loop to E2B sandboxes and Langfuse observability, arguing the model is only one block and the advantage comes from the architecture around it.
GPT 6 Astra is making AI agents f…cking illegal 10 GitHub repos I’d put around Astra to turn one model into an entire AI system 01 LangGraph ▸ https://github.com/langchain-ai/langgraph → control the agent loop 02 PydanticAI ▸ https://github.com/pydantic/pydantic-ai → structured agents + typed outputs GIVE IT MEMORY 03 Mem0 ▸ https://github.com/mem0ai/mem0 04 Graphiti ▸ https://github.com/getzep/graphiti → temporal knowledge graph GIVE IT TOOLS 05 Composio ▸ https://github.com/ComposioHQ/composio 06 Browser Use ▸ https://github.com/browser-use/browser-use → let agents operate the web LET IT EXECUTE 07 E2B ▸ https://github.com/e2b-dev/E2B → isolated cloud sandboxes 08 Daytona ▸ https://github.com/daytonaio/daytona → disposable development environments WATCH EVERYTHING 09 Langfuse ▸ https://github.com/langfuse/langfuse 10 DeepEval ▸ https://github.com/confident-ai/deepeval the architecture: GPT 6 Astra → context → memory → tools → execution → evaluation → correction → memory that’s the part people keep missing the model is only one block the real advantage comes from everything you build around it ⭣
This post quotes another post: x.com/promixyz/status/2102424036416348494