132 smart memory management #142
Conversation
* initial work on split and upload for Supabase SQL redesign * Added changes to supabase calls due to new incoming formatting with the database * changed url metadata in py_ingest function from str to None * full implementatino of parallel embeddings + bulk upload to qdrant * format ONLY, nothing more * errors in split and upload and embeddings * fully working code for the new Supabase SQL refactor * changed the env variable name back to normal name, and also made sure that everything is working, this is final push for new supabase * some of the previous changes were erased when I Pushed and it was breaking, so I added it all back in and now it should be working fine * added url and baseurl to all metadatas * made minor adjustments getenv variables and baseurl to fix bugs * fixed bug with split and upload and fixed supabase table as well * all functions now work with new database changes * got rid of ingest_single_html function * minor fix to optional metadata properties --------- Co-authored-by: jkmin3 <[email protected]> Co-authored-by: Joshua K Min <[email protected]>
… with Document objects
* printing * Should fix all getTopContext errors, and use best practices in Flask API optional params
…t_webhooks branch
|
Closing this issue as the team decided to experiment with langgraph (https://python.langchain.com/docs/langgraph) The team decided to first log/capture all agent messages using langgraph, manage the agent using states and then work on smart memory management. The code will be in the 'langgraph' branch (https://github.com/UIUC-Chatbot/ai-ta-backend/tree/langgraph) |
The PlanAndExecute agent has a callback to
memorycallbacks.pyclass. The set callbacks have supabase update and insert commands. However, the supabase table ('docker_images') is not being populated.