Implementation of agentic GraphRAG feedback as described above is now discussed in this issue.
Should be implemented as a separate project using the MCP endpoint.
Requires GraphRAG features (#4) first.
MCP agent should ask a question, write the answer to disk into a tracked folder, manage follow-up-questions, keep the expansion/traversal balanced (also partially descibed in #4), and extend the formatted answer with links between questions and answers.
(The resulting interlinked Markdown file tree will be perfect for visualization and exploration in Obsidian).
Following counterfactual questions is also an interesting avenue to explore, encouraging the AI to think outside the box; "what if...?"
Implementation of agentic GraphRAG feedback as described above is now discussed in this issue.
Should be implemented as a separate project using the MCP endpoint.
Requires GraphRAG features (#4) first.
MCP agent should ask a question, write the answer to disk into a tracked folder, manage follow-up-questions, keep the expansion/traversal balanced (also partially descibed in #4), and extend the formatted answer with links between questions and answers.
(The resulting interlinked Markdown file tree will be perfect for visualization and exploration in Obsidian).
Following counterfactual questions is also an interesting avenue to explore, encouraging the AI to think outside the box; "what if...?"