v1.2.4

1.2.4 (2026-03-08)

Bug Fixes

• embeddings: make incremental generation content-hash based and project-safe (8a2f025)

  • use content hashes as the source of truth for incremental freshness instead of relying on mtimes
  • distinguish full scans from partial file updates so stale embeddings are only pruned when safe
  • prune deleted file and document embeddings during full scans to keep review context current
  • scope project structure embeddings and retrieval by project path to avoid cross-project collisions
  • harden retrieval and summary invalidation so code review uses up-to-date project context

v1.2.3

1.2.3 (2026-03-08)

Bug Fixes

• enhance chunk processing with error handling (946e84b)

  • Updated reviewLargePRInChunks to use Promise.allSettled for better error handling during chunk processing.
  • Implemented logic to preserve successful chunk results even when some chunks fail, ensuring comprehensive results are returned.

• ensure markdown output is shown when --output-file option is set to markdown (42f830d)

• normalize reviewFiles result shape (4da4693)

  • Flatten file-level results when aggregating multi-file reviews so output formatters receive consistent entries.
  • Also wrap skipped single-file results with file metadata.

• preserve partial context retrieval results (dbd1c37)

  • Handle context retrieval failures per source so one rejected lookup does not discard successful PR comments, guidelines, code examples, or custom document results.

v1.2.2

1.2.2 (2026-03-04)

Bug Fixes

• actions: remove eval from CI action command execution (472b81f)

  Harden composite actions by building CLI arguments as arrays and invoking node directly, eliminating eval-based command construction.

• add line numbers to code content for improved review context (362a0c4)

  • Implemented addLineNumbers function to prepend line numbers to source code, enhancing clarity during code reviews.
  • Updated rag-analyzer.js and prompt-cache.js to utilize line numbers in displayed content for better reference.

• custom-docs: use canonical cache-manager custom chunk APIs (7958578)

  Remove now-unused alias methods and update custom document processor/tests to rely on the existing custom chunk cache interfaces.

• git: harden diff execution and avoid branch checkout side effects (5cc7a53)

  Use safely escaped git diff invocation and create tracking branches without switching the working tree during branch preparation.

• rag: prevent parse fallback crash on non-string LLM responses (2ff1f16)

  Serialize fallback raw responses safely so malformed analysis payloads no longer throw during reconstruction.

• runtime: harden cleanup paths and status query fallbacks (993eb22)

  Ensure temporary DB connections and progress timers are always cleaned up, guard missing review output paths in the PR comment action, and use table query APIs for last-analysis lookups.

v1.2.1

1.2.1 (2026-02-11)

Bug Fixes

• ensure Dockerfiles are analyzed during PR reviews (948eccf)
• ensure Makefiles are analyzed during PR reviews (bb4463e)

v1.2.0

1.2.0 (2026-02-07)

Features

• add support for prompt caching (0f92f56)

  • Introduced cache-ttl option for LLM prompts, allowing users to specify a cache duration of 5m (default) or 1h (extended).
  • Updated README and action.yml to reflect the new option.
  • Enhanced prompt caching mechanism in the LLM module to optimize token usage.
  • Adjusted related code to support the new caching feature across various components.

v1.1.1

1.1.1 (2026-02-03)

Bug Fixes

• update prompt to make LLM only report actionable code issues (b0b8a7f)

v1.1.0

1.1.0 (2026-02-02)

Features

• add automatic rejection guidelines for non-code fix suggestions (35e49ad)

v1.0.0

1.0.0 (2026-02-01)

Initial release

• v1 release of codecritique