OpenInterpreter
RAG
Is your feature request related to a problem? Please describe.
Open Interpreter could be significantly enhanced with the integration of RAG (Retrieval-Augmented Generation) and memory. For RAG, feeding it sources like project documentation, Python docs, or YouTube transcripts would allow it to retrieve relevant information during interactions, boosting accuracy. Memory, on the other hand, would enable the model to remember past mistakes or successes, like the latest API approach for Azure or OpenAI, preventing the reuse of outdated methods.
Any approach for this in pipeline yet?
Describe the solution you'd like
Rag/memory
Describe alternatives you've considered
No response
Additional context
No response
+1 on adding RAG capabilities to Open Interpreter.
Otherwise, the only way to pass specific "knowledge" (i.e. a PDF manual of a CLI tool; a link to the man page of a tool; etc.) is basically to manage the Assistant sys prompt, memory and RAG on a another frontend tool (i.e. Ollama WebUI, BoltAI, Jan, etc.) that points to Open Interpreter server.
Maybe the team is thinking about including RAG features on the native MacOS app? Finger Xssed...
This can be done by building a script that imports the Open Interpreter python library.
It can also be done for the CLI by building a profile with custom tools/instructions.
Adding RAG to the core project is not current in scope
