Croway opened a new pull request, #30451:
URL: https://github.com/apache/camel/pull/30451

   [CAMEL-27247](https://issues.apache.org/jira/browse/CAMEL-27247)
   
   ## Description
   
   `CamelSpringAiChatFinishReason` holds the finish reason as the provider 
names it (`stop`, `end_turn`, `STOP`, `max_tokens`, `content_filter`, ...), so 
a route that checks for a truncated or filtered answer depends on the provider. 
Spring AI 2.1.0 classifies the finish reason the same way for every provider, 
and keeps the reasoning of a reply as message parts.
   
   - New header `CamelSpringAiChatFinishReasonCategory`: `COMPLETED`, 
`TRUNCATED`, `TOOL_CALL`, `FILTERED`, `OTHER` or `UNSPECIFIED`. A refusal of 
the SafeGuard advisor is `FILTERED`.
   - New header `CamelSpringAiChatReasoning`: the text of the reasoning parts 
of the reply (Anthropic and Bedrock thinking, Gemini thoughts, the 
`reasoning_content` of DeepSeek and OpenAI-compatible servers), or the summary 
of a part that has only a summary. It is set only when the reply has reasoning, 
and removed otherwise, so a header of an earlier call does not stay.
   - Both are also in the `CamelSpringAiChatResponseMetadata` map.
   - Documentation in "Response Headers and Metadata", with a `choice()` on the 
category.
   
   ## Tests
   
   `SpringAiChatFinishReasonReasoningTest` (4): category and reasoning of a 
completed reply, a truncated reply without reasoning that removes an earlier 
reasoning header, a reasoning part with only a summary, and the SafeGuard 
refusal as `FILTERED`. The module has 49 unit tests, all passing.
   
   _Claude Code on behalf of Croway_


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to