Problem: JARVIS has an embeddings stack for memory but cannot answer questions over your own documents.
Where:
- Embeddings:
jarvis/core/embeddings.py, jarvis/core/semantic_memory.py
- New plugin:
jarvis/plugins/rag_tools.py
Approach: Add index_folder(path, glob) - chunk text files (md/txt/pdf/code), embed, store in a separate table/collection. Add ask_docs(query) that retrieves top-k chunks and injects them into context. Show indexed sources in the HUD.
Acceptance:
index_folder ingests a folder of .md/.txt
- Asking a question retrieves relevant chunks and answers with citations (file + chunk)
- Falls back to keyword search if embeddings unavailable
Difficulty: large. Flagship feature.
Problem: JARVIS has an embeddings stack for memory but cannot answer questions over your own documents.
Where:
jarvis/core/embeddings.py,jarvis/core/semantic_memory.pyjarvis/plugins/rag_tools.pyApproach: Add
index_folder(path, glob)- chunk text files (md/txt/pdf/code), embed, store in a separate table/collection. Addask_docs(query)that retrieves top-k chunks and injects them into context. Show indexed sources in the HUD.Acceptance:
index_folderingests a folder of .md/.txtDifficulty: large. Flagship feature.