Tool call transparency fix #21
Open
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
tool_calls and tool_call_id are needed to correctly inform the model about historic tool calls. Also fixed a related error message.
For tool call requests by the model, accept their typical parameters rather than throwing an error. The goal of this fix is limited to preventing undue errors: impact of tool calls on token counting is only a guess which needs to revised.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.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.
Enables model-initiated function calls by allowing tool call related parameters to pass though the helper.
Aligns common error conditions with OpenAI documentation which allows tool related messages to have certain combination of arguments.
Limitation: tool related messages may have an impact on token counts. The goal of this fix is limited to enabling tool calls; token counting may not be accurate.