Ollama Python Library 0.4 Release Notes

Ollama Python library 0.4 enables developers to provide Python functions directly as tools for LLMs, removing the need to manually define JSON schemas. This update streamlines the integration of external capabilities—such as mathematical operations or API requests—into model workflows.

Direct Python Function Integration as Tools

Developers can now pass actual Python function references into the tools field of the ollama.chat method. This allows the model to request the execution of specific functions based on the user's prompt.

Implementation Workflow

Integrating a function as a tool follows a three-step process:

  1. Function Definition: Define a standard Python function. For optimal model performance, developers should annotate parameter and return value types and include Google-style docstrings to provide the model with necessary context.
  2. Passing the Tool: Include the function reference in the tools list when calling ollama.chat.
  3. Execution: Iterate through response.message.tool_calls to identify the requested function and execute it using the arguments provided by the model.

Support for External Libraries

This functionality extends to existing Python libraries and SDKs. For example, the request function from the requests library can be passed as a tool, allowing the model to trigger HTTP requests to fetch external web content.

Automated JSON Schema Generation

Ollama Python library 0.4 automates the creation of the tool definitions required by the model. Previously, these JSON schemas had to be provided manually.

The library now utilizes Pydantic and docstring parsing to automatically generate the JSON schema from the Python function's signature and documentation. This process extracts the function name, description, and parameter types (e.g., mapping Python int to JSON integer) to create a valid tool definition for the LLM.

Additional Library Enhancements

Beyond function calling improvements, the 0.4 release introduces the following updates:

  • Full Typing Support: The library now features comprehensive typing support, enabling direct object access while preserving backward compatibility with existing functionality.
  • Updated Examples: New and updated implementation examples are available on the official Ollama Python GitHub repository.

Sources

Related

  • Dispatch
    Ollama Python and JavaScript Libraries Release

    Ollama has released official Python and JavaScript libraries to simplify the integration of Ollama's REST API features into applications.

  • Dispatch
    Ollama Structured Outputs Release

    Ollama now supports structured outputs, allowing users to constrain model responses to a specific format defined by a JSON schema for increased reliability and consistency.

  • Project
    ollama/ollama-python

    A Python library that simplifies integration with Ollama, enabling developers to chat with, manage, and generate embeddings from LLMs locally or in the cloud.

  • Dispatch
    Ollama Tool Support Release

    Ollama has introduced tool calling support for models like Llama 3.1, allowing local LLMs to interact with external APIs, functions, and code interpreters.

  • Dispatch
    Ollama adds streaming tool calling support

    Ollama now supports streaming responses with tool calling, letting chat applications receive partial model output and invoke functions in real time.