fix(aws): Handle thinking with tools, use CountTokens API for Haiku 4.5 #698
  Add this suggestion to a batch that can be applied as a single commit.
  This suggestion is invalid because no changes were made to the code.
  Suggestions cannot be applied while the pull request is closed.
  Suggestions cannot be applied while viewing a subset of changes.
  Only one suggestion per line can be applied in a batch.
  Add this suggestion to a batch that can be applied as a single commit.
  Applying suggestions on deleted lines is not supported.
  You must change the existing code in this line in order to create a valid suggestion.
  Outdated suggestions cannot be applied.
  This suggestion has been applied or marked resolved.
  Suggestions cannot be applied from pending reviews.
  Suggestions cannot be applied on multi-line comments.
  Suggestions cannot be applied while the pull request is queued to merge.
  Suggestion cannot be applied right now. Please check back later.
  
    
  
    
Couple of updates to ChatBedrock/ChatBedrockConverse for Haiku 4.5 support:
get_num_tokens_from_messages().bind_tools()andwith_structured_output(), as Haiku (and Sonnet) 4.5 still do not support this.test_structured_output_tool_choice_not_supportedConverse integration test has not been updated to include Haiku 4.5, which appears to behave very inconsistently with a structured output schema compared to previous Claude thinking models. Sonnet 4.5 and older models produce a tool call 100% of the time if the query is directly applicable to the schema. However, on Haiku 4.5, the chance of this happening was observed to be <5%.