A thought I keep coming back to:
If agent workflows are software, we should be able to replay, fork, and test them like software.
Built "langgraph-replay" to experiment with that idea.
• Deterministic replay
• Checkpoint forking
• Zero LLM/API calls during replay
• SQLite-backed execution history
curious how others are handling agent debugging and regression testing.
https://t.co/WFCxpwkMwD
A thought I keep coming back to:
If agent workflows are software, we should be able to replay, fork, and test them like software.
Built "langgraph-replay" to experiment with that idea.
• Deterministic replay
• Checkpoint forking
• Zero LLM/API calls during replay
• SQLite-backed execution history
curious how others are handling agent debugging and regression testing.
https://t.co/WFCxpwkMwD
Thanks @Kautukkundan and team for setting up MLX Community event at Mumbai
Literally made my first local llm based project within 15 min.
MLX framework is lit 🔥 didn't thought it would be that easy and fast
And ofcourse thanks @Parth_Shethji26 for sharing your macbook 😅