Documentation review

[adamw]

• main page: an overview of features: api access, agent loop, so that ppl know what’s possible with the project
• quickstart: apart from deps, also include a starting code snippet
• openai/basics.html:
  • link is not a link
  • which openai features are available
  • sth above basic usage? maybe the title of the page should be "OpenAI API basics" or sth like that
  • Token limits on GPT-5 / reasoning models - should be a section
• openai/streaming.html
  • sections for fs2/ox
  • at least list the deps for other libs
  • whenever tapir, fs2 etc. is mentioned for the first time -> link to project
• openai/structured-outputs.html
  • the flow is weird and unclear: first its described how to derive schema via Tapir, then manually, then via Tapir, but differently (?), then again with Tapir?
  • it should be structured as: what problem does it solve - what is needed (json schema) - how to achieve it the easiest way - how to customise - how to do manually
• openai/compatible-apis.html
  • list ollama, grok, openrouter in the page title
  • better sectioning (top-level, not nested)
• claude/basics.html
  • page title could be better, but the features are nicely listed - we should have the same in openai
  • claude/messages.html - lacking information "what is this for"
  • claude/structured-outputs.html - can the information about different options for deriving structured outputs schemas be somehow common for openai & claude?
  • claude/tool-calling.html - why is there no tool calling section for openai? or maybe this is common for both?
  • claude/models-and-errors.html - seems like a catch-all page. Sync client should be the default one that we propose. Rework this or split into separate pages
• agents/tools.html
  • schema derivation could be shared with schema discussion for structured outputs?
• other/backends.html
  • the "other backends" seems relevant only for agent loop?
  • effect systems is only for openai?