[KongchangAI]
Product

terraphim-grep

An open-source local search and summarization tool that uses a knowledge-graph synonym thesaurus to parse queries and rank local matches. It can invoke a low-cost LLM to synthesize summaries of retrieved passages, and supports a search-only mode that avoids network calls entirely.

Timeline (last 90 days)

Oct 19

The terraphim-grep tool resolves query terms through a curated knowledge-graph synonym thesaurus and ranks local matches; the --answer parameter requests a generated answer when the sufficiency check requires it, --force-rlm forces the synthesis path, and --search-only mode completely disables synthesis and skips building the LLM client.

Unverified50%
Oct 19

In a validation run, the raw snippet content was 505 characters, the evidence with source locations was 1108 characters, and the synthesized answer text was 440 characters, so the frontier model received about 40% of the original context.

Unverified50%
Oct 19

If the full answer object with all ten citations (1918 characters) were forwarded as-is to the frontier model, the savings would disappear; only the synthesized result plus the citations needed for verification should be passed.

Unverified50%

All Facts (3)

Source Articles