-
Notifications
You must be signed in to change notification settings - Fork 2
Open
Labels
enhancementNew feature or requestNew feature or request
Description
Summary
Support "pinned" source documents that get injected at high priority into agent context, separate from vector search results. (Probably into the system prompt.)
Context
One use case: vocabulary files. Enterprise deployments often have highly domain-specific terminology and acronyms. When these terms are present in a query, high-priority context injection will work better to give the LLM necessary definitions rather than relying on vector search.
Proposed Features
- Pinned documents: Always included in agent context at high priority
- YAML front matter support for source documents:
pinned: true- mark document for priority injectionchunking_strategy- control how document is chunked (or not)priority- relative priority for context ordering
- Default chunking behavior: "don't chunk" for reference docs (similar to API documentation)
Use Cases
- Domain-specific acronym glossaries
- Internal terminology definitions
- Critical runbook sections that should always be available
- API reference documentation
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
enhancementNew feature or requestNew feature or request