fix: RubyLLM adapter not capturing token counts from Gemini #37
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
- Create Gemini-specific mapper to handle Google Gemini API response format - Extract token counts from usageMetadata (promptTokenCount, candidatesTokenCount) - Support explicit token passing via meta hash (takes priority over response) - Route "gemini" and "google" providers to the new Gemini mapper - Add comprehensive tests for Gemini token extraction and meta override - Fixes issue where RubyLLM adapter could not capture tokens from Gemini API responses
The Gemini mapper now checks for token counts at the top level of the response hash (input_tokens, output_tokens) which is where RubyLLM places them after normalizing the Gemini API response. Token extraction priority: 1. Explicit meta (highest priority) 2. Top-level response (RubyLLM format) 3. usageMetadata (raw Gemini API format) Closes #25
Instead of patching each provider mapper to handle RubyLLM's normalized format, add a single RubyLLM mapper that handles all providers consistently. The mapper routing now auto-detects RubyLLM format by checking for top-level input_tokens/output_tokens fields. This approach: - Works for all providers (OpenAI, Anthropic, Gemini, etc.) - Keeps provider-specific mappers focused on raw API formats - Preserves the original provider name in the interaction record Reverts the band-aid fix to the Gemini mapper since the RubyLLM mapper now handles this case properly.
BREAKING CHANGE: TraceBook now only supports RubyLLM normalized responses. Provider-specific mappers (OpenAI, Anthropic, Gemini, Ollama) have been removed. RubyLLM abstracts all providers into a consistent Message format with top-level input_tokens/output_tokens fields. This simplifies the codebase significantly - one mapper handles all providers. Removed: - lib/tracebook/mappers/openai.rb - lib/tracebook/mappers/anthropic.rb - lib/tracebook/mappers/gemini.rb - lib/tracebook/mappers/ollama.rb - lib/tracebook/mappers/base.rb - Provider-specific tests Updated: - lib/tracebook/mappers.rb - single normalize method for RubyLLM format - lib/tracebook/adapters/active_agent.rb - use unified Mappers.normalize
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Summary
BREAKING CHANGE: Removes provider-specific mappers and uses RubyLLM format exclusively.
RubyLLM abstracts all LLM providers into a consistent
Messageformat with top-levelinput_tokens/output_tokensfields. Since we only support RubyLLM, there's no need for provider-specific mappers.Removed (-850 lines):
lib/tracebook/mappers/openai.rblib/tracebook/mappers/anthropic.rblib/tracebook/mappers/gemini.rblib/tracebook/mappers/ollama.rblib/tracebook/mappers/base.rbSimplified:
lib/tracebook/mappers.rb- singlenormalizemethod for RubyLLM formatlib/tracebook/adapters/active_agent.rb- use unifiedMappers.normalizeTest plan
Closes #25