Anthropic - remove redacted thinking handling #2864
Description
Stems from #2683
Redacted thinking was a Claude 3.7 behavior, removed in Claude 4 models.
Occasionally Claude’s internal reasoning will be flagged by our safety systems. When this occurs, we encrypt some or all of the thinking block and return it to you as a redacted_thinking block. redacted_thinking blocks are decrypted when passed back to the API, allowing Claude to continue its response without losing context.
Sources:
- https://platform.claude.com/docs/en/build-with-claude/extended-thinking
- https://docs.aws.amazon.com/bedrock/latest/userguide/claude-messages-thinking-encryption.html (here the removal is explicitly specified)
Handling redacted thinking makes our code more complex. Now that 3.7 models are retired, we should consider simplifying the code.