v0.2.0

What's Changed

• ci: credit pull requests on the release page and check their titles by @bpinheiroms in #69
• build(docker): build images with the Rust toolchain CI uses by @bpinheiroms in #70
• fix: keep private linked code out of wider reviews and limit /ocr review to members by @bpinheiroms in #71
• fix(web): render the nuqs adapter inside the router by @bpinheiroms in #74
• fix(archive): purge merged pull requests' frozen evidence as soon as the merge is confirmed by @bpinheiroms in #73
• fix: post distinct findings on commented lines and sanitize published review text by @bpinheiroms in #72
• fix(review): count only reviewable files toward the file limit by @bpinheiroms in #75
• feat(archive): leave generated files and lockfiles out of review evidence by @bpinheiroms in #76
• fix(ci): send policy 2 evidence in the runner smoke and run it on archive changes by @bpinheiroms in #77
• ci: build the smoke images unoptimized on every core by @bpinheiroms in #79
• fix(review): redact credentials in linear time by @bpinheiroms in #78

New Contributors

• @dependabot[bot] made their first contribution in #66

Full Changelog: v0.1.1...v0.2.0

v0.1.1

What's Changed

• fix(docker): take distribution security updates and patch pulldown-cmark by @bpinheiroms in #68

Full Changelog: v0.1.0...v0.1.1

v0.1.0

Added

• First public release of OpenCodeReview: AI code review for GitHub pull requests that runs locally in Docker.
• Reviews with the Claude Code CLI or the Codex CLI, signed in with your own Claude or ChatGPT account. Claude
  reviews use Claude Opus 5.5 at medium effort by default; the top bar picks each provider's model and effort.
• Several accounts with failover. Pick the account new reviews start on from the top bar, which shows how much of
  each plan is left and when it resets. At a usage limit, a review continues on the next account of the same
  provider, then the other provider.
• Codebase indexing: file metadata, a package and import graph and lightweight symbols. Reviews get the callers,
  dependencies and tests of the changed code, plus your convention files (AGENTS.md, CLAUDE.md, REVIEW.md
  and others), including committed agent folders such as .claude and .agents.
• Evidence-checked findings. Each finding quotes the exact code, anchors to a changed line, carries a severity
  (high, medium or low) and names its trigger; a defect whose trigger cannot be named is asked as a question. A
  point that fails validation is dropped with a warning, and such a review is never treated as clean.
• Full, incremental, base-sync and stacked pull request reviews. Large pull requests are split into chunks
  reviewed in parallel and checked by a final pass.
• Publication on GitHub: a collapsed pull request summary, a result comment and one non-blocking review with
  inline comments. Lower-priority suggestions (tests, docs, naming) are posted only when Settings → PR selection →
  Review comments → Post suggestions is on. OpenCodeReview never approves a pull request.
• A Reviews list grouped into Needs you, AI reviewing, Completed, Drafts and skipped, and Closed and merged, with a
  details panel for each run.
• A Usage page with the tokens reviews used over 7, 30 or 90 days, by day, pull request, repository, model and
  account, and a picture of it to share.
• Editable, versioned review skills.
• Guided onboarding at http://localhost:1421.
• Data in Postgres. Several server instances can share one database (docker-compose.scale.yml).
• Prebuilt images for amd64 and arm64 at ghcr.io/bridglabs/opencodereview-server, -runner and -web
  (docker-compose.images.yml).
• Source-available project files: PolyForm Small Business 1.0.0 license, NOTICE, contributing guide, code of
  conduct, security policy, issue and pull request templates.