WarpFix

About WarpFix

WarpFix is an autonomous CI repair agent that watches for failing workflow runs, diagnoses them, generates a patch, validates it in an isolated sandbox and opens a pull request — closing the loop that usually costs an engineer a context switch. The pipeline is explicit and observable: detect the workflow_run failure, parse the log, classify the error type, patch, validate against the full test suite in a container, score confidence, then ship a PR. The demo walks a real null-reference TypeError from detection to an opened PR in about twenty seconds with a 94% confidence score and 47 of 47 tests passing. The mechanism that makes repeat failures cheap is fingerprint learning: error patterns are hashed with SHA-256 and cached, so a fingerprint that has already been seen and fixed resolves from cache rather than re-running the full analysis — the demo shows a fingerprint matched 47 times with a 340 ms average resolution. Every patch is tested in an isolated container before any PR opens, and high-confidence fixes can be marked auto-merge ready. Beyond repair, WarpFix ships what it calls a CI Brain — per-test reliability scores, fingerprint history across repositories and owner hotspot mapping, computed from logs without an LLM in the path — and the platform advertises 21 specialised engines spanning CI repair, code review, security, observability, intelligence and quality. A six-agent pipeline handles the parse, classify, patch, validate, score and ship stages.

Key Features

Tags