Archon
Archon copied to clipboard
Feature Request: Automatic summarization & key‑point extraction for knowledge items
Why it matters:
- Rapidly scanning large knowledge bases is difficult; concise summaries reduce time‑to‑understand and help identify relevant sources.
- Summaries could improve RAG workflows by providing a preview of the content and guiding agents to relevant chunks.
- The feature aligns with Archon’s mission of managing context and tasks for AI coding assistants
Proposed implementation:
- When a document is uploaded or a site is crawled, call the existing LLM provider to generate a summary (configurable length) and key bullet‑points.
- Store summaries in the database alongside embeddings.
- Display the summary on the knowledge item card in the UI, with a button to regenerate/re‑summarize if needed.
- Provide a toggle to enable/disable automatic summarization in settings (to manage API costs).
Acceptance criteria:
- Users see a summary and key‑points for newly created knowledge items.
- A “Regenerate Summary” action is available and uses the selected LLM provider.
- Summaries can be disabled per project or globally.
- Summaries are included in search or can be excluded from queries.
Hi> Why it matters:
- Rapidly scanning large knowledge bases is difficult; concise summaries reduce time‑to‑understand and help identify relevant sources.
- Summaries could improve RAG workflows by providing a preview of the content and guiding agents to relevant chunks.
- The feature aligns with Archon’s mission of managing context and tasks for AI coding assistants
Proposed implementation:
- When a document is uploaded or a site is crawled, call the existing LLM provider to generate a summary (configurable length) and key bullet‑points.
- Store summaries in the database alongside embeddings.
- Display the summary on the knowledge item card in the UI, with a button to regenerate/re‑summarize if needed.
- Provide a toggle to enable/disable automatic summarization in settings (to manage API costs).
Acceptance criteria:
- Users see a summary and key‑points for newly created knowledge items.
- A “Regenerate Summary” action is available and uses the selected LLM provider.
- Summaries can be disabled per project or globally.
- Summaries are included in search or can be excluded from queries.
I appreciate this suggestion! Adding it to the backlog in our board.