LLM policy
This page explains what is automated vs evidence-backed in SolvedStack fix pages. If you use these pages during a production incident, treat them as a runbook starter: verify in your environment.
What LLMs may generate
- Short “TL;DR” summaries and fix framing.
- Structured sections (why-prod, when-not, verify) derived from issue/PR evidence.
- Cluster and topic tagging for navigation and internal linking.
What SolvedStack tries to keep evidence-linked
- GitHub issue links and PR/commit references.
- Error strings / signatures extracted from the thread or patch context.
- Version claims (affected/fixed) when a release can be inferred from the repo metadata.
What you should verify before deploying
- Reproduction on your exact version + dependency graph (FastAPI/Pydantic/Starlette-style breakages are common).
- Behavioral changes and backward compatibility (especially for auth, OpenAPI, retries, and idempotency).
- Rollback plan / safe pin if upgrade isn’t safe mid-incident.
Corrections
If a page is wrong or unsafe, email [email protected] with the URL and your environment details.