fix(anthropic): parse non-streaming responses (H-04)

[PierrunoYT]

Summary

• parse non-streaming Anthropic Messages API JSON through dedicated typed response, content-block, and usage representations
• return provider-neutral text, thinking, and tool-use messages for print mode and compaction while leaving SSE streaming parsing unchanged
• propagate input, output, cache-read, and cache-creation usage and reject Anthropic error envelopes
• add regression tests for text, tool use, usage, errors, and retained streaming behavior

Testing

• cargo test --locked llm::anthropic::tests -- --test-threads=1 (10 passed)
• cargo test --locked --all-targets -- --test-threads=1 (227 passed)
• git diff --check upstream/main...HEAD

Fixes #18

[coderabbitai]

📝 Walkthrough

Walkthrough

Anthropic completion handling now uses separate parsers for streaming SSE events and non-streaming JSON responses. Non-streaming parsing extracts text, tool use, thinking, usage, cache tokens, and API errors, with expanded tests.

Changes

Anthropic response parsing

Layer / File(s)	Summary
Non-streaming response parser src/llm/anthropic.rs	Adds JSON response types and parsing for text, tool-use, thinking, usage, cache tokens, and Anthropic API errors, with corresponding tests.
Dedicated parser wiring src/llm/anthropic.rs	Routes complete through the non-streaming parser and stream_complete through the streaming parser; updates streaming parser tests.

Estimated code review effort: 4 (Complex) | ~45 minutes

Suggested reviewers: euxaristia

Poem

A rabbit parsed the JSON bright,
While SSE hopped along its flight.
Text and tools joined usage too,
Cache tokens counted in the queue.
API errors now speak clear—
“No empty answers hiding here!”

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly states the main change: fixing Anthropic non-streaming response parsing.
Description check	✅ Passed	The description matches the code changes and tests around Anthropic non-streaming parsing.
Linked Issues check	✅ Passed	The changes implement the dedicated non-streaming parser and usage handling requested in #18.
Out of Scope Changes check	✅ Passed	The diff stays focused on Anthropic parsing and regression tests, with no clear unrelated changes.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[PierrunoYT]

@coderabbitai review

[coderabbitai]

✅ Action performed

Review finished.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.

[coderabbitai]

🧹 Nitpick comments (1)

src/llm/anthropic.rs (1)

550-564: 🩺 Stability & Availability | 🔵 Trivial | ⚡ Quick win

Default missing Anthropic non-streaming token counts to 0 for consistency.

Anthropic non-streaming usage.input_tokens/usage.output_tokens are normally present, but unlike the streaming parser, this complete-response path turns a missing/invalid usage object or count into a hard parse error after content has already been parsed. Default these token counts to 0 like Usage::default() to keep empty usage from discarding otherwise-usable responses.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/llm/anthropic.rs` around lines 550 - 564, Update the non-streaming
Anthropic response parsing around the usage extraction in the
response-construction path to tolerate missing or invalid usage data. Default
both input_tokens and output_tokens to 0, including when the usage object is
absent, while preserving the parsed role and content instead of returning an
error.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In `@src/llm/anthropic.rs`:
- Around line 550-564: Update the non-streaming Anthropic response parsing
around the usage extraction in the response-construction path to tolerate
missing or invalid usage data. Default both input_tokens and output_tokens to 0,
including when the usage object is absent, while preserving the parsed role and
content instead of returning an error.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: c4737792-4120-465f-b90c-81c9448b7e1a

📥 Commits

Reviewing files that changed from the base of the PR and between 0304f80 and faeb723.

📒 Files selected for processing (1)

• src/llm/anthropic.rs