fix(chat): persona models (Fréd/M Corpus/Mwen) speak only as themselves - #360
Merged
Merged
Conversation
… no scaffold Fréd / M Corpus / Mwen (config extra_models[].system_prompt*) are personas the user talks TO. The server built their persona prompt correctly, then glued the frontend's generic "You are CODEC Deep Chat — a J.A.R.V.I.S.-class AI…" system message onto it and (Think on) appended the FINAL-ANSWER reasoning scaffold. On a 4B that produced two conflicting identities: it answered in the generic voice, confabulated an attachment it never read, and emitted a stray <Final Answer> tag the UI stripper (### FINAL ANSWER only) never removed. For a persona model: REPLACE the client system message with the persona prompt instead of concatenating, and skip _REASON_SCAFFOLD. Also strip <Final Answer> tag variants in the UI as defense-in-depth. Operator persona (everyday model) unchanged — it still gets the CODEC prompt + scaffold + skills. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Persona models are talked TO, not through CODEC. The server built the persona prompt, then glued the frontend's generic "You are CODEC Deep Chat — J.A.R.V.I.S.-class AI…" system message on top and appended the FINAL-ANSWER scaffold (Think on). On a 4B: generic voice, attachment confabulation, and a stray
<Final Answer>tag the UI stripper missed.Fix: for a persona model (
extra_models[].system_prompt*), REPLACE the client system message with the persona prompt (no concatenation) and skip_REASON_SCAFFOLD; strip<Final Answer>variants in the UI too. Operator/everyday model unchanged.Requires
pm2 restart codec-dashboardto take effect (Python route).🤖 Generated with Claude Code