Skip to content

Release 0.25.2: Claude 5.5 models work with an Anthropic key - #36

Merged
mattbaconz merged 2 commits into
mainfrom
fix/anthropic-thinking-blocks
Oct 21, 2026
Merged

mattbaconz merged 2 commits into
mainfrom
fix/anthropic-thinking-blocks

Conversation

@mattbaconz

Copy link
Copy Markdown
Contributor

With an Anthropic key, 0.25.1 failed goals with anthropic response missing content[0].text whenever the model thought before answering. Claude Sonnet 5.5, Opus 5.5 and Haiku 5.5 think by default, so a reply can open with a thinking block, and the provider read only content[0].text. Sonnet 5.5 and Opus 5.5 are Phonton's default Anthropic models for the standard and frontier tiers.

Changes

  • Join every text block of a Messages reply and skip thinking blocks.
  • A reply with no text fails with the API's stop_reason instead of a missing-field error.
  • Regression test anthropic_text_skips_leading_thinking_block.
  • Release 0.25.2: version in phonton-cli/Cargo.toml, Cargo.lock, package.json and the README badge, plus the CHANGELOG entry and release-notes/v0.25.2.md.

Verified on Windows

  • cargo fmt --check, cargo clippy --workspace --all-targets -D warnings, and the release build pass.
  • cargo test --workspace --no-fail-fast: every suite passes except phonton-verify browser_check_passes_on_valid_html, which needs the Playwright install that CI performs (Cannot find module 'playwright' locally).
  • phonton doctor, phonton doctor --provider against Anthropic, extensions list/doctor --json, mcp list --json, scripts/benchmark-plan.ps1 -ReleaseBinary, npm run test:npm-wrapper (reports 0.25.2) and npm pack --dry-run pass.
  • Live API: claude-sonnet-5-5 answered a coding prompt with a thinking block followed by the text. On 0.25.1 an Aider Polyglot JavaScript goal failed at a model call with this error; on this branch the same goal got text back from every model call, and the Rust exercise luhn-from ended review-ready on Sonnet 5.5 with its tests passing.

Claude Sonnet 5.5, Opus 5.5 and Haiku 5.5 think by default, so a reply can
open with a thinking block. The provider read only content[0].text and
failed those goals. Join the text blocks instead, and report stop_reason
when a reply has no text.
@chatgpt-codex-connector

Copy link
Copy Markdown

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review 🔄 Running since 2026-10-21T16:38:09.533484Z e168fc4 PR opened
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@mattbaconz
mattbaconz merged commit 9367970 into main Oct 21, 2026
2 checks passed
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