judge

sigmap judge tells you whether an answer appears to be supported by the context you actually supplied.

sigmap judge --response response.txt --context .context/query-context.md
sigmap judge --response response.txt --context .context/query-context.md --json
sigmap judge --response response.txt --context .context/query-context.md --learn

What it reports

• Groundedness — overlap score between the answer and the source context
• Support level — pass/fail style verdict
• Unsupported symbols — tokens and claims that look weakly supported

This is a traceability check, not a truth oracle. It helps answer: "Did this response come from the code I provided, or did the model drift?"

Typical output

Groundedness       : 0.78
Support level      : pass
Unsupported symbols: none

Best input pairing

Use judge right after ask:

1. sigmap ask "explain auth flow"
2. Save the model answer to response.txt
3. sigmap judge --response response.txt --context .context/query-context.md

Opt-in learning

With --learn, judge can apply a small local boost or penalty to the files referenced in the context headings:

• strongly grounded result → small boost
• weakly grounded result → small penalty
• middle band → no change

This learning is local-only and stored in .context/weights.json.

When to use it

• reviewing AI-generated explanations
• checking whether a debugging suggestion is really grounded in the shown files
• grading prompt/response pairs in demos or release benchmarks
• feeding the learning engine carefully instead of manually every time