DSPySignatureChatGenerator

Generate structured chat responses using DSPy signatures and modules inside a Haystack pipeline.

Key Features

• Uses DSPy signatures to define structured input/output contracts for generation.
• Supports three DSPy module types: Predict, ChainOfThought, and ReAct.
• Accepts and returns messages in ChatMessage format, making it compatible with standard Haystack chat pipelines.
• Supports any LiteLLM-compatible model via the model parameter (for example, openai/gpt-5-mini, anthropic/claude-sonnet-4-5).
• Supports custom input_mapping to connect signature fields to pipeline inputs from upstream components.
• API keys are read automatically from environment variables by DSPy and LiteLLM (for example, OPENAI_API_KEY).

Configuration

1. Drag the DSPySignatureChatGenerator component onto the canvas from the Component Library.
2. Click on the component to open the configuration panel.
3. On the General tab:
   1. Set the model to the LiteLLM model identifier you want to use (for example, openai/gpt-5-mini).
   2. Set the signature to define your input/output structure (for example, "question -> answer" or a fully qualified class name).
   3. Set the module_type to Predict, ChainOfThought, or ReAct.
   4. Make sure the API key for the model provider is set as an environment variable (for example, OPENAI_API_KEY). For instructions, see Create Secrets.
4. Go to the Advanced tab to configure generation_kwargs, input_mapping, and pipeline_inputs.

Connections

DSPySignatureChatGenerator receives a list of ChatMessage objects. It extracts the last user message and passes its text to the DSPy module. It outputs a list of reply ChatMessage objects you can connect to AnswerBuilder or other downstream components.

Source Code

To check this component's source code, open chat_generator.py in the Haystack Core Integrations repository.

Usage Examples

Basic Configuration

  DSPySignatureChatGenerator:
    type: haystack_integrations.components.generators.dspy.chat.chat_generator.DSPySignatureChatGenerator
    init_parameters:
      signature: "question -> answer"
      model: openai/gpt-5-mini
      module_type: ChainOfThought
      output_field: answer
      generation_kwargs:
        max_tokens: 1024

Using the Component in a Pipeline

components:
  prompt_builder:
    type: haystack.components.builders.chat_prompt_builder.ChatPromptBuilder
    init_parameters:
      required_variables: "*"
      template:
      - role: user
        content: "Answer the following question: {{ question }}"

  llm:
    type: haystack_integrations.components.generators.dspy.chat.chat_generator.DSPySignatureChatGenerator
    init_parameters:
      signature: "question -> answer"
      model: openai/gpt-5-mini
      module_type: ChainOfThought
      output_field: answer

  answer_builder:
    type: deepset_cloud_custom_nodes.augmenters.deepset_answer_builder.DeepsetAnswerBuilder
    init_parameters:
      reference_pattern: acm

connections:
- sender: prompt_builder.prompt
  receiver: llm.messages
- sender: llm.replies
  receiver: answer_builder.replies

max_runs_per_component: 100

metadata: {}

inputs:
  query:
  - answer_builder.query
  - prompt_builder.question

outputs:
  answers: answer_builder.answers

Parameters

Inputs

Parameter	Type	Description
messages	List[ChatMessage]	A list of chat messages. The last user message is used as the primary input to the DSPy module.

Outputs

Parameter	Type	Description
replies	List[ChatMessage]	A list containing one reply ChatMessage with the value of the signature's output field.

Init Parameters

These are the parameters you can configure in Pipeline Builder:

Parameter	Type	Default	Description
signature	str | type[dspy.Signature]		The DSPy signature defining the input/output structure. Use a string like "question -> answer" or pass a fully qualified dspy.Signature subclass name.
model	str	openai/gpt-5-mini	LiteLLM model identifier (for example, openai/gpt-5-mini, anthropic/claude-sonnet-4-5).
api_base	Optional[str]	None	Optional base URL for the API. Useful for local or self-hosted models.
module_type	str	ChainOfThought	The DSPy module type: Predict, ChainOfThought, or ReAct.
output_field	str	answer	The signature output field whose value is used as the reply text.
generation_kwargs	Optional[Dict[str, Any]]	None	Additional generation parameters (for example, temperature, max_tokens).
module_kwargs	Optional[Dict[str, Any]]	None	Additional keyword arguments passed to the DSPy module constructor.
input_mapping	Optional[Dict[str, str]]	None	Maps signature input field names to run() kwarg names from upstream components.
pipeline_inputs	Optional[List[str]]	None	Signature input fields exposed as Haystack pipeline input sockets so upstream components can connect to them.

Run Method Parameters

These are the parameters you can configure for the component's run() method. This means you can pass these parameters at query time through the API, in Playground, or when running a job. For details, see Modify Pipeline Parameters at Query Time.

Parameter	Type	Default	Description
messages	List[ChatMessage]		A list of chat messages. The last user message is used as input.
generation_kwargs	Optional[Dict[str, Any]]	None	Runtime generation parameters that override init-time values.

Related Information

• Pipeline Components
• Integrations Overview