Skip to content

The output fixing parser isn't working with Gemini 2.5 flash. #1015

@carlows

Description

@carlows

Describe the bug
The output fixing parser isn't working with Gemini 2.5 flash.

As you described in the readme, google models require a different format:

messages = [
  { role: "system", content: "You are a helpful assistant." },
  { role: "user", content: "What's the weather like today?" }
  # Google Gemini and Google VertexAI expect messages in a different format:
  # { role: "user", parts: [{ text: "why is the sky blue?" }]}
]

However, the OutputFixingParser does not take that into account: https://github.com/patterns-ai-core/langchainrb/blob/main/lib/langchain/output_parsers/output_fixing_parser.rb#L50-L55

To Reproduce
Using the output fixing parser:

parser = Langchain::OutputParsers::StructuredOutputParser.from_json_schema('some_schema.json'))
Langchain::OutputParsers::OutputFixingParser.from_llm(llm:, parser:)

Raises an error when using the model gemini 2.5-flash.

Expected behavior
The output fixing parser should work with google/gemini models

Terminal commands & output
Commands you used and the terminal output.

Screenshots
If applicable, add screenshots to help explain your problem.

Desktop (please complete the following information):

  • OS: [e.g. OS X, Linux, Ubuntu, Windows]
  • Ruby version [3.3]
  • Langchain.rb version [0.19.5]

Additional context

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions