pr-agent · Issues· 53 open

Replace SUPPORT_REASONING_EFFORT_MODELS with litellm's model metadata plus the config override

GPT-5.6-Terra silently fails

GitHub Action exits 0 when an automatic command fails

extract_relevant_lines_str breaks the review block when the extracted lines contain a code fence

[Feature]: add an opt-in fail-fast policy for automatic command sequences

Large pull requests can yield incomplete provider change sets

Azure DevOps renames lose the original path and content

Azure persistent_finding_state never resolves findings because the head SHA is empty

Easier per-review model and reasoning-effort selection: guidance on an acceptable approach

Support replying directly to the trigger comment for main commands (e.g., /review, /improve)

feat(status): acknowledge a pull request the moment automatic commands start

docs: the Google Tag Manager container 404s and its loader never renders; four dead outbound links

The default focus_only_on_problems instruction forbids the import a suggestion itself introduces

docs: github.md examples use `num_code_suggestions`, but the real config key is `num_code_suggestions_per_chunk` - documented overrides are silently ignored

Is there any plan for supporting code graph?