Expose retrieval mode in query-based responses for easier debugging #14

New Issue

sam · 2026-04-03T11:27:12Z

sam commented

2026-04-03 11:27:12 +00:00

Summary

Expose whether a query-based tool used semantic/vector search or the Postgres full-text fallback path in its response payloads.

Why

llm/todo.md lists this as a useful follow-up for debugging search behavior after the semantic fallback work. Right now fallback appears to be implemented, but callers cannot easily tell which retrieval mode was used for a given response.

Affected tools

Consider this for the query-based tools that use shared retrieval logic:

search_thoughts
recall_context
get_project_context when query is provided
summarize_thoughts when query is provided
semantic neighbors in related_thoughts

Scope ideas

Add a small machine-readable field such as retrieval_mode: semantic|text
Preserve backward compatibility for existing clients
Keep the field consistent across all query-driven outputs
Update README and any relevant tests

Acceptance criteria

Query responses clearly indicate whether vector search or full-text fallback was used
The field is consistent across the supported tools
Documentation and tests cover the new output shape

Source

Nice follow-up from llm/todo.md: expose the retrieval mode in responses for easier debugging of semantic vs text fallback.

## Summary Expose whether a query-based tool used semantic/vector search or the Postgres full-text fallback path in its response payloads. ## Why `llm/todo.md` lists this as a useful follow-up for debugging search behavior after the semantic fallback work. Right now fallback appears to be implemented, but callers cannot easily tell which retrieval mode was used for a given response. ## Affected tools Consider this for the query-based tools that use shared retrieval logic: - `search_thoughts` - `recall_context` - `get_project_context` when `query` is provided - `summarize_thoughts` when `query` is provided - semantic neighbors in `related_thoughts` ## Scope ideas - Add a small machine-readable field such as `retrieval_mode: semantic|text` - Preserve backward compatibility for existing clients - Keep the field consistent across all query-driven outputs - Update README and any relevant tests ## Acceptance criteria - Query responses clearly indicate whether vector search or full-text fallback was used - The field is consistent across the supported tools - Documentation and tests cover the new output shape ## Source Nice follow-up from `llm/todo.md`: expose the retrieval mode in responses for easier debugging of semantic vs text fallback.

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: wdevs/amcs#14