Repository navigation
Integrate ordinary Sarvam action execution and preserve reporting failures - #55
Merged
Merged
Conversation
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Connect native Sarvam actions to ordinary SQLite approval checks and lossless transcript reporting, including read-only requests, safe refusals, actual rejection, duplicates and incomplete/error paths. Preserve raw model output and all historical runners.
The frozen 24-episode setup completes with all current dates/states correct and all four approved writes successful. Raw reporting is 21/24; strict tasks are 19/24 because of three status-label mismatches and two safely rejected closed-booking attempts. The original gate fails and the larger comparison remains unstarted. A separate offline receipt-status projector corrects the three status fields on those same histories; this is application behavior after inspection, not a new model score or memory improvement.
Validation: 113 focused tests, Ruff and diff checks; all 277 artifact hashes and clean-export offline replay of every request, SQLite state transition, usage charge and separate projection. Live phase: 61 calls, no errors/retries, estimated INR3.364035. Six synthetic setup situations, no independent native-language review or general accuracy claim.
Evidence and detailed miss review:
benchmarks/audits/sarvam-integrated-ordinary-2026-10-11/REPORT.md. Three independent commits retain implementation, receipt-status follow-up and research evidence.