Skip to content

docs: fix errors and fill gaps found in a full docs review - #398

Merged
yuanhao merged 3 commits into
mainfrom
docs/review-fixes
Oct 30, 2026
Merged

yuanhao merged 3 commits into
mainfrom
docs/review-fixes

Conversation

@yuanhao

@yuanhao yuanhao commented Oct 30, 2026

Copy link
Copy Markdown
Collaborator

From two read-only reviews of every docs page against src/ (concepts + architecture + llms.txt; getting-started + guides + providers + reference + READMEs + skill). No broken links were found; the preset table, provider details and defaults matched the code.

Errors fixed

  • prompt() while a run is in flight waits for it (no panic); is_streaming() semantics.
  • DecisionError::Refused ends a fallback chain (and is listed).
  • prompt-caching: the usage sample compiles; Sonnet 5 figures use Sonnet 5's rates ($1.00 vs $0.19).
  • The extensions spend bucket in pricing.md, api.md and the skill.
  • tool_output_max_lines_overrides default includes shared_state.
  • openai-compat "add a provider" example compiles (OpenAiCompat is #[non_exhaustive]).
  • Bedrock has no default base URL.
  • MCP: refused redirects name origins only (0.26.1).
  • StreamConfig is #[non_exhaustive].
  • Examples use the gpt_5_5() preset (1M / 64K), not the generic constructor (128K / 4,096).
  • Smaller: SharedState capacity error type, add_priced signature, architecture module tree and provider wiring, agent-loop input-filter link.
  • Stale versions: README.zh-CN pin, clef-worker pins, workers/rutis README version notes, rutis version history 0.7–0.10.

Gaps filled

  • MCP guide: 0.26.1 URL redaction (HTTP error reaching <origin>, [redacted], prefer with_header, match McpError::Http).
  • ModelConfig::presets() / preset() / generic().
  • API reference: extension builders (Agent + SubAgentTool), extension_spend(), with_mcp_server_http_transport, with_execution_limits, extension re-exports.
  • agent-loop: dropping a _with_sender future cancels the run; field rows for max_tokens, temperature, extensions, tree_extensions, max_stop_continues, run_label; extensions.md finish on a dropped prompt.
  • callbacks: synchronous InputFilter.
  • sub-agents: spend section; wasm32: clef*() / gpt_6_luna() read no env there; Go and Bun plugins in READMEs.
  • Skill feature map: preset lookup, Bedrock/Azure, decision gate/guard, spend_handle, GASP (names compiled via skill_test's prose_names).
  • llms.txt lists all pages (8 were missing).

Checks: skill_test and testing_guide_test pass; every new or changed Rust snippet was compile-checked in a scratch test against the crate (removed). The Mistral pricing sentences are left to #396.

🤖 Generated with Claude Code

https://claude.ai/code/session_01T7iq5hpndSiHQcnAsywKuG

yuanhao and others added 3 commits October 30, 2026 12:15
Errors: prompt() during a run waits (no panic); DecisionError::Refused
ends a fallback chain; prompt-caching sample and Sonnet 5 figures;
extension spend bucket in pricing/api/skill; tool-output override
defaults; non-compiling openai-compat example; Bedrock has no default
URL; refused redirects name origins only (0.26.1); StreamConfig is
non_exhaustive; presets over generic constructors in examples; stale
version notes and pins (zh-CN README, clef-worker, workers/rutis
READMEs, rutis version history 0.7-0.10).

Gaps: 0.26.1 URL redaction in the MCP guide; ModelConfig::preset /
generic; extension builders and re-exports in the API reference;
dropping a _with_sender future cancels the run; AgentLoopConfig field
rows; synchronous InputFilter; sub-agent spend; wasm32 env-less presets;
Go and Bun plugins; skill feature-map rows; llms.txt lists every page.

New/changed snippets compile-checked against the crate.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01T7iq5hpndSiHQcnAsywKuG
…n/notify wording, Go and Bun everywhere

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01T7iq5hpndSiHQcnAsywKuG
….md and CLAUDE.md

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01T7iq5hpndSiHQcnAsywKuG
@yuanhao
yuanhao merged commit 08c8c99 into main Oct 30, 2026
16 checks passed
@yuanhao yuanhao mentioned this pull request Oct 30, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant