Permalink
Comparing changes
Choose two branches to see what’s changed or to start a new pull request.
If you need to, you can also or
learn more about diff comparisons.
Open a pull request
Create a new pull request by comparing changes across two branches. If you need to, you can also .
Learn more about diff comparisons here.
...
- 5 commits
- 10 files changed
- 3 contributors
Commits on Mar 4, 2026
-
-
Fix binary attachment reads on Windows
Use File.binread for path attachments to avoid text-mode truncation on Windows and add a regression spec ensuring path attachments are read in binary mode.
-
Fix schema JSON parsing for intermediate tool-call responses (#650)
## Summary Fixes #649 When both `with_schema` and `with_tool` are used together, the `complete` method unconditionally parses any string `response.content` as JSON whenever a schema is set — including intermediate responses that also contain tool calls. On the next API call, the parsed Hash gets serialized as `{ type: "text", text: <Hash> }`, which providers like Anthropic reject with: ``` text.text: Input should be a valid string ``` ### Root cause In `lib/ruby_llm/chat.rb`, the JSON parsing condition: ```ruby if @Schema && response.content.is_a?(String) ``` does not check whether the response is an intermediate tool-call response, so it eagerly parses content that should remain as a plain string. ### Fix Add a `!response.tool_call?` guard so JSON parsing is only applied to the final (non-tool-call) response: ```ruby if @Schema && response.content.is_a?(String) && !response.tool_call? ``` ### Test Added a regression test that mocks the provider to return an intermediate tool-call response with JSON-like text content, verifying that: - The intermediate tool-call message keeps its content as a `String` - The final non-tool-call response has its content parsed into a `Hash` --------- Co-authored-by: Claude Opus 4.6 <[email protected]> Co-authored-by: Carmine Paolino <[email protected]>
3 people authoredMar 4, 2026 -
Populate Gemini cached token usage
Map usageMetadata.cachedContentTokenCount to cached_tokens for Gemini chat responses and streaming chunks, and add provider specs covering both paths.
-
Loading
This comparison is taking too long to generate.
Unfortunately it looks like we can’t render this comparison for you right now. It might be too big, or there might be something weird with your repository.
You can try running this command locally to see the comparison on your machine:
git diff 1.13.0...1.13.1