AISDKv5 chunk type "reasoning-signature" not supported

Error message

AISDKv5 chunk type "reasoning-signature" not supported

What it means

The AI SDK v5 wire format has no representation for 'reasoning-signature' chunks (and 'redacted-reasoning'), so convertMastraChunkToAISDKv5 throws instead of dropping provider signature data silently. This surfaces during streaming conversion when a model emits reasoning signature payloads.

Source

Thrown at packages/core/src/stream/aisdk/v5/transform.ts:474

        // Cast needed: Mastra's LanguageModelUsage has optional properties, V2 has required-but-nullable
        totalUsage: chunk.payload.output.usage as LanguageModelV2Usage,
      };
    }
    case 'reasoning-start':
      return {
        type: 'reasoning-start',
        id: chunk.payload.id,
        providerMetadata: chunk.payload.providerMetadata,
      };
    case 'reasoning-delta':
      return {
        type: 'reasoning-delta',
        id: chunk.payload.id,
        text: chunk.payload.text,
        providerMetadata: chunk.payload.providerMetadata,
      };
    case 'reasoning-signature':
      throw new Error('AISDKv5 chunk type "reasoning-signature" not supported');
    case 'redacted-reasoning':
      throw new Error('AISDKv5 chunk type "redacted-reasoning" not supported');

    case 'reasoning-end':
      return {
        type: 'reasoning-end',
        id: chunk.payload.id,
        providerMetadata: chunk.payload.providerMetadata,
      };
    case 'source':
      if (chunk.payload.sourceType === 'url') {
        return {
          type: 'source',
          sourceType: 'url',
          id: chunk.payload.id,
          url: chunk.payload.url!,
          title: chunk.payload.title,
          providerMetadata: chunk.payload.providerMetadata,

View on GitHub (pinned to 75dd419e61)

Solutions

Upgrade @mastra/core to a version where reasoning-signature chunks are handled/supported in the v5 conversion
Disable reasoning/signature output for the model if you don't need it (e.g. omit reasoning options)
Filter 'reasoning-signature' chunks out in stream middleware before conversion
Catch the error per-stream and fall back to a non-v5 (native Mastra) stream consumer

Example fix

// before
const stream = agent.stream(...).toAISDKv5Stream(); // crashes on signature chunks
// after
const stream = agent.stream(...).fullStream.pipe(
  filter(chunk => chunk.type !== 'reasoning-signature' && chunk.type !== 'redacted-reasoning')
); // then convert

Defensive patterns

Strategy: try-catch

Type guard

function isUnsupportedInV5(chunk: MastraChunk): boolean {
  return chunk.type === 'reasoning-signature' || chunk.type === 'redacted-reasoning';
}

Try / catch

try {
  result = convertMastraChunkToAISDKv5(chunk);
} catch (e) {
  if (String(e).includes('reasoning-signature')) {
    logger.debug('Dropping reasoning-signature chunk (unsupported in AISDK v5)');
    return null; // skip chunk
  }
  throw e;
}

Prevention

Filter out reasoning-signature/redacted-reasoning chunks before v5 conversion
Upgrade @mastra/core to a release that supports these chunk types if you need signatures
Disable provider reasoning signature output when the consumer doesn't use it
Test streaming with your reasoning-enabled models before deploying upgrades

When it happens

Trigger: Streaming from a model that emits reasoning-signature chunks (e.g. OpenAI o-series/Anthropic extended thinking signature payloads) through the AISDKv5 conversion path in an environment where that chunk type is not supported.

Common situations: Using reasoning-enabled models with the v5 compat stream; upgrading a provider package that started emitting signature chunks; replaying stored traces containing signature chunks.

Related errors

AI-assisted analysis of mastra-ai/mastra@75dd419e61 (2026-08-30). Data as JSON: /api/errors/ee7120657ee1c639. Report an issue: GitHub.