fix(anthropic): cover retry stream recreation

[rosetta-livekit-bot]

Summary

• create a fresh Anthropic streaming request for each retry attempt
• avoid re-awaiting the same coroutine after a transient stream creation failure
• add a no-network regression test for the retry path

Fixes #5805.

Tests

• uv run pytest tests\test_plugin_anthropic.py -q
• uv run ruff check livekit-plugins\livekit-plugins-anthropic\livekit\plugins\anthropic\llm.py tests\test_plugin_anthropic.py
• python -m py_compile livekit-plugins\livekit-plugins-anthropic\livekit\plugins\anthropic\llm.py tests\test_plugin_anthropic.py
• git diff --check

[changeset-bot]

🦋 Changeset detected

Latest commit: 9f8f651

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 32 packages

Name	Type
@livekit/agents-plugin-anthropic	Patch
@livekit/agents	Patch
@livekit/agents-plugin-anam	Patch
@livekit/agents-plugin-assemblyai	Patch
@livekit/agents-plugin-baseten	Patch
@livekit/agents-plugin-bey	Patch
@livekit/agents-plugin-cartesia	Patch
@livekit/agents-plugin-cerebras	Patch
@livekit/agents-plugin-deepgram	Patch
@livekit/agents-plugin-elevenlabs	Patch
@livekit/agents-plugin-fishaudio	Patch
@livekit/agents-plugin-google	Patch
@livekit/agents-plugin-hedra	Patch
@livekit/agents-plugin-hume	Patch
@livekit/agents-plugin-inworld	Patch
@livekit/agents-plugin-lemonslice	Patch
@livekit/agents-plugin-liveavatar	Patch
@livekit/agents-plugin-livekit	Patch
@livekit/agents-plugin-minimax	Patch
@livekit/agents-plugin-mistral	Patch
@livekit/agents-plugin-mistralai	Patch
@livekit/agents-plugin-neuphonic	Patch
@livekit/agents-plugin-openai	Patch
@livekit/agents-plugin-phonic	Patch
@livekit/agents-plugin-resemble	Patch
@livekit/agents-plugin-rime	Patch
@livekit/agents-plugin-runway	Patch
@livekit/agents-plugin-sarvam	Patch
@livekit/agents-plugin-silero	Patch
@livekit/agents-plugin-trugen	Patch
@livekit/agents-plugin-xai	Patch
@livekit/agents-plugins-test	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[devin-ai-integration]

Devin Review found 1 new potential issue.

View 12 additional findings in Devin Review.

[devin-ai-integration]

🟡 Empty text emitted and retryable set to false after thinking block is completely stripped

When a text delta consists entirely of a <thinking>...</thinking> block (or when the </thinking> tag appears at the very end of a delta leaving nothing after it), the remaining text after filtering is "". The code falls through to emit this empty string as a content chunk and sets retryable = false on line 367. This has two consequences: (1) an empty content delta is needlessly queued to the consumer, and (2) the stream becomes non-retryable despite no meaningful content being delivered, so subsequent API errors will not trigger a retry.

Trace for text delta containing only a thinking block

For text = "<thinking>hidden</thinking>":

1. thinkingStart = 0, preThinking = "" (not emitted because of the if (preThinking) guard)
2. text = "hidden</thinking>" after slicing past <thinking>
3. #ignoringCoT = true
4. thinkingEnd = 6, text = "" after slicing past </thinking>
5. #ignoringCoT = false
6. Falls through to line 363, emits { content: "" } and sets retryable = false

Suggested change

	if (this.#ignoringCoT) {
	continue;
	}
	this.queue.put({
	id: this.#requestId,
	delta: { role: 'assistant', content: text },
	});
	retryable = false;
	if (this.#ignoringCoT) {
	continue;
	}
	if (!text) {
	continue;
	}
	this.queue.put({
	id: this.#requestId,
	delta: { role: 'assistant', content: text },
	});
	retryable = false;

---

Was this helpful? React with 👍 or 👎 to provide feedback.