How jurjans/m365-graph-rag-samples works
This repository is a set of illustrative patterns for a private Microsoft 365 document-answering assistant, not an integrated application. Its intended workflow is to authenticate to Graph, search SharePoint content, prepare retrieval-friendly chunks, and generate a cited answer grounded in supplied sources. The samples show the individual calls to MSAL, Microsoft Graph, and Anthropic; they do not implement the full retrieval-to-answer orchestration. The documented search-to-chunk-to-answer path is therefore shown as unverified.
Main entry points: graph_search.py SharePoint Search · rag.py Grounded Answer · auth.py Graph Auth · chunk.py Text Chunking
Components 8
Graph Access
- Graph Auth
src/auth.py - MSAL
Content Retrieval
- SharePoint Search
src/graph_search.py - Microsoft Graph
- Text Chunking
src/chunk.py
Grounded Answering
- Grounded Answer
src/rag.py - Anthropic Claude
Other components
- Questioner
Connections 8
- Questioner → SharePoint Search: asks query
src/graph_search.py - SharePoint Search → Graph Auth: gets token
src/graph_search.py - Graph Auth → MSAL: acquires token
src/auth.py - SharePoint Search → Microsoft Graph: searches documents
src/graph_search.py - SharePoint Search → Text Chunking: prepares hits
- Text Chunking → Grounded Answer: supplies context
- Grounded Answer → Anthropic Claude: requests answer
src/rag.py - Grounded Answer → Questioner: returns cited answer
src/rag.py