langchain_ollama

Ollama embedding model integration.

Ollama large language models.

Ollama embeddings models.

Ollama large language models.

Ollama chat models.

Input Flow (LangChain -> Ollama)

_convert_messages_to_ollama_messages():

• Transforms LangChain messages to ollama.Message format
• Extracts text content, images (base64), and tool calls

_chat_params():

• Combines messages with model parameters (temperature, top_p, etc.)
• Attaches tools if provided
• Configures reasoning/thinking mode via think parameter
• Sets output format (raw, JSON, or JSON schema)

Output Flow (Ollama -> LangChain)

1. Ollama Response

Stream dictionary chunks containing:

• message: Dict with role, content, tool_calls, thinking
• done: Boolean indicating completion
• done_reason: Reason for completion (stop, length, load)
• Token counts/timing metadata

2. Response Processing (_iterate_over_stream())

• Extracts content from message.content
• Parses tool calls into ToolCalls
• Separates reasoning content when reasoning=True (stored in additional_kwargs)
• Builds usage metadata from token counts

3. LangChain Output (ChatGenerationChunk -> AIMessage)

• Streaming: Yields ChatGenerationChunk with AIMessageChunk content
• Non-streaming: Returns ChatResult with complete AIMessage
• Tool calls attached to AIMessage.tool_calls
• Reasoning content in AIMessage.additional_kwargs['reasoning_content']

Ollama chat model integration.

ollama pull gpt-oss:20b
pip install -U langchain-ollama

Key init args — completion params: model: str Name of Ollama model to use. reasoning: bool | None Controls the reasoning/thinking mode for supported models.

    - `True`: Enables reasoning mode. The model's reasoning process will be
        captured and returned separately in the `additional_kwargs` of the
        response message, under `reasoning_content`. The main response
        content will not include the reasoning tags.
    - `False`: Disables reasoning mode. The model will not perform any reasoning,
        and the response will not include any reasoning content.
    - `None` (Default): The model will use its default reasoning behavior. Note
        however, if the model's default behavior *is* to perform reasoning, think tags
        (`<think>` and `</think>`) will be present within the main response content
        unless you set `reasoning` to `True`.
temperature: float
    Sampling temperature. Ranges from `0.0` to `1.0`.
num_predict: int | None
    Max number of tokens to generate.

See full list of supported init args and their descriptions in the params section.