mirror of
https://github.com/BerriAI/litellm.git
synced 2026-09-14 23:21:35 +00:00
Adds the pieces needed for the 3-stage public rollout: - red 'not ready' label paired with 'ready for review'; the review gate reconciles the pair on every run and reconsider flips it on reopen - lite mode (AGENT_SHIN_MODE=lite): the review gate never closes and instead posts a one-time 'closes start in 7 days' notice linking the policy blog post (AGENT_SHIN_POLICY_URL); labels still reconcile - stricter reconsider for closed PRs: requires a Greptile confidence score of at least 4/5 (missing score fails with a comment pointing at @greptileai) and bypasses the linked-issue short-circuit so the QA evidence rubric always applies; a pass reopens and tags ready - reconsider on an OPEN PR now runs the review gate so the label pair flips immediately instead of waiting for the daily sweep - PR rubric tightened: bug fixes need before AND after evidence; command proof from a custom script must include the script source (collapsible section or linked gist) - default triage model bumped to gpt-5.6-luna |
||
|---|---|---|
| .. | ||
| actions/detect-backend-changes | ||
| codeql | ||
| ISSUE_TEMPLATE | ||
| observatory | ||
| screenshots | ||
| scripts | ||
| workflows | ||
| dependabot.yaml | ||
| deploy-on-aws.png | ||
| deploy-on-gcp.png | ||
| deploy-to-aws.png | ||
| FUNDING.yml | ||
| pull_request_template.md | ||
| template.yaml | ||