🩹 fix: Drop Null/Empty Reasoning Blocks in the v1 Bedrock Converter

[jomspk]

Problem

#248 guarded the native-Bedrock converter (convertAIMessageToConverseMessage) against reasoning blocks whose reasoningText.text is null/empty, but the v1 converter (convertFromV1ToChatBedrockConverseMessage, taken when response_metadata.output_version === 'v1') still serializes reasoning unguarded in both of its branches:

} else if (block.type === 'reasoning') {
  const reasoning = block as { reasoning: string };
  assistantMsg.content?.push({
    reasoningContent: { reasoningText: { text: reasoning.reasoning } }, // ← no null/empty check
  } as BedrockContentBlock);
} else if (block.type === 'reasoning_content') {
  assistantMsg.content?.push({
    reasoningContent: langchainReasoningBlockToBedrockReasoningBlock(block), // ← no isSerializableBedrockReasoningBlock guard
  } as BedrockContentBlock);
}

A v1 reasoning block whose reasoning is null/empty — e.g. produced by a model that omits reasoning text by default (thinking.display: "omitted", the default on Claude Opus 4.7+ / Sonnet 5) — reaches Bedrock as reasoningText: { text: null } and the whole request is rejected:

1 validation error detected: Value at 'messages.N.member.content.M.member.reasoningContent.reasoningText.text'
failed to satisfy constraint: Member must not be null

We hit exactly this in a LibreChat deployment: Bedrock anthropic.claude-sonnet-5 + MCP tools fails on the tool-call replay turn with the error above. It is the same failure class #248 fixed — #248's "Scope / remaining gap" section covers the non-streaming @langchain/aws delegation but not this v1 path in the same file, which is still unguarded on main / 3.2.63.

Fix

Apply #248's drop-the-unsendable-block approach to the v1 converter:

• reasoning (v1 standard block): drop when reasoning is null/empty.
• reasoning_content: gate on the existing isSerializableBedrockReasoningBlock before converting, same as the non-v1 path.
• Add the BEDROCK_EMPTY_TEXT_PLACEHOLDER fallback (from 🩹 fix: Drop Foreign Reasoning Blocks in Anthropic Message Converter #243) when the drop leaves a turn with no content blocks and no tool calls — mirroring the non-v1 converter, so the drop can't trade the null-text 400 for an empty-content 400.

Blocks carrying real text (or redacted content) are converted unchanged.

Change Type

• Bug fix (non-breaking change which fixes an issue)

Testing

Test-first: added a convertToConverseMessages — v1 reasoning serialization describe block to message_inputs.test.ts (5 cases; 4 fail without the fix, reproducing reasoningText: { text: undefined } passing through):

• v1 reasoning block with missing reasoning → dropped; sibling text + tool calls survive
• v1 reasoning block with empty reasoning → dropped
• v1 signature-only reasoning_content block → dropped
• v1 turn emptied by the drop → placeholder emitted, not empty content
• v1 reasoning / reasoning_content blocks carrying real text → still converted (not over-dropped)

Test Configuration:

• npx jest src/llm/bedrock: 9 suites / 106 passed, 4 skipped, 0 failed
• Wider npx jest src/llm src/messages: only failures are 4 pre-existing env-dependent suites (invoke, google, anthropic, vertexai) that fail identically on a clean main checkout without this change
• npx tsc --noEmit, npx eslint, npx prettier --check on changed files: clean
• Verification is unit-level; no live Bedrock run was performed for this PR (the production error above is from a real deployment, and the raw-Converse repro in 🩹 fix: Drop Bedrock Reasoning Blocks With Null/Empty Text #248 already demonstrates Bedrock's rejection of null-text reasoning blocks)

Checklist

• My code adheres to this project's style guidelines
• I have performed a self-review of my own code
• I have commented in any complex areas of my code
• My changes do not introduce new warnings
• I have written tests demonstrating that my changes are effective or that my feature works
• Local unit tests pass with my changes

🤖 Generated with Claude Code

[danny-avila]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 541bbe3822

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[danny-avila]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d0c84a0939

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[danny-avila]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 9b7ca42e95

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[danny-avila]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: c390f0240a

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[danny-avila]

@codex review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. You're on a roll.

Reviewed commit: 3c00ae3f8b

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".