Questions and problems (qa)

This hunter finds what goes wrong in practice. Its sources feed the "common mistakes" callouts in the notes and the experiments in the read-along. The role is almost always pitfalls.

GitHub issues and discussions

1. The official repo is in research/overview.md.
2. codelabgen github issues <repo> saves the most-reacted issues as Markdown in the cache and prints the path. Read it. Narrow it down with --query, for example --query "label:question", --query "checkpointer", or the names of core concepts.
3. codelabgen github issues <repo> --discussions does the same for discussions (Q&A and ideas), if the repo uses them. Many projects keep their help forum elsewhere (a Discourse forum, Discord); search for it, and if it is public and indexable, treat it like Stack Overflow.
4. Look for patterns: the same misunderstanding asked many times, errors with confusing messages, behaviour that surprised people, breaking changes that hit many users.

Stack Overflow

• WebSearch site:stackoverflow.com <topic> <concept> for the core concepts, or browse the tag (https://stackoverflow.com/questions/tagged/<tag>?tab=Votes).
• Prefer high-voted questions with an accepted answer that still holds for the pinned version. Answers for older versions are still pitfalls ("this used to work"), but say so.

What to record

Don't record single issues. Record collections that are useful as a whole:

• the cached top-issues file as one source: url https://github.com/<repo>/issues?q=sort%3Areactions-%2B1-desc, with a reason summarizing the main recurring problems
• a topic-specific issue search, when one concept clearly causes many problems
• the Stack Overflow tag, sorted by votes, and individual Q&As only when they are excellent explanations
• the official forum's FAQ or most-viewed threads

In reason, name the concrete pitfalls found, for example: "Recurring: msgpack serialization errors for custom state types; confusion between updates and values stream modes; thread_id missing from config."