HuggingFaceLocalGenerator

Generate text using models from Hugging Face that run locally.

text2text-generation deprecated

The text2text-generation task is deprecated and may be removed in a future release. In transformers v5+, text2text-generation is no longer available as a valid pipeline task. If you currently use task="text2text-generation", replace it with task="text-generation" and ensure the selected model is compatible. To use the older behavior, pin transformers<5.

Key Features

• Runs LLMs locally using Hugging Face's transformers pipeline — no external API required.
• Supports decoder models like GPT and Qwen with the text-generation task.
• Configurable generation parameters such as max_new_tokens, temperature, and top_p.
• Supports streaming responses and configurable stop words.
• Automatically selects the best available device or accepts a custom device setting.

Configuration

Authentication

For remote models, this component uses the HF_API_TOKEN environment variable. Connect deepset AI Platform with Hugging Face first. For details, see Use Hugging Face Models.

1. Drag the HuggingFaceLocalGenerator component onto the canvas from the Component Library.
2. Click the component to open the configuration panel.
3. On the General tab:
   1. Enter the Hugging Face model name or local path, such as Qwen/Qwen3-0.6B.
4. Go to the Advanced tab to configure the task, device, token, generation kwargs, Hugging Face pipeline kwargs, stop words, and streaming callback.

Connections

HuggingFaceLocalGenerator receives a prompt string — typically from PromptBuilder. It outputs replies (a list of generated strings). Connect its replies output to AnswerBuilder for formatting answers.

Usage Example

This query pipeline uses HuggingFaceLocalGenerator for local text generation:

components:
  bm25_retriever:
    type: haystack_integrations.components.retrievers.opensearch.bm25_retriever.OpenSearchBM25Retriever
    init_parameters:
      document_store:
        type: haystack_integrations.document_stores.opensearch.document_store.OpenSearchDocumentStore
        init_parameters:
          hosts:
          index: 'default'
          max_chunk_bytes: 104857600
          embedding_dim: 768
          return_embedding: false
          method:
          mappings:
          settings:
          create_index: true
          http_auth:
          use_ssl:
          verify_certs:
          timeout:
      top_k: 10
      fuzziness: 0

  PromptBuilder:
    type: haystack.components.builders.prompt_builder.PromptBuilder
    init_parameters:
      template: |
        Given the following information, answer the question.

        Context:
        {% for document in documents %}
        {{ document.content }}
        {% endfor %}

        Question: {{ query }}
      required_variables:
      variables:

  HuggingFaceLocalGenerator:
    type: haystack.components.generators.hugging_face_local.HuggingFaceLocalGenerator
    init_parameters:
      model: Qwen/Qwen3-0.6B
      task: text-generation
      device:
      token:
        type: env_var
        env_vars:
          - HF_API_TOKEN
          - HF_TOKEN
        strict: false
      generation_kwargs:
        max_new_tokens: 100
        temperature: 0.9
      huggingface_pipeline_kwargs:
      stop_words:
      streaming_callback:

  AnswerBuilder:
    type: haystack.components.builders.answer_builder.AnswerBuilder
    init_parameters:
      pattern:
      reference_pattern:

connections:
  - sender: bm25_retriever.documents
    receiver: PromptBuilder.documents
  - sender: PromptBuilder.prompt
    receiver: HuggingFaceLocalGenerator.prompt
  - sender: HuggingFaceLocalGenerator.replies
    receiver: AnswerBuilder.replies
  - sender: bm25_retriever.documents
    receiver: AnswerBuilder.documents

inputs:
  query:
    - bm25_retriever.query
    - PromptBuilder.query
    - AnswerBuilder.query

outputs:
  answers: AnswerBuilder.answers

max_runs_per_component: 100

metadata: {}

Parameters

Inputs

Parameter	Type	Default	Description
prompt	str		A string representing the prompt.
streaming_callback	Optional[StreamingCallbackT]	None	A callback function that is called when a new token is received from the stream.
generation_kwargs	Optional[Dict[str, Any]]	None	Additional keyword arguments for text generation.

Outputs

Parameter	Type	Default	Description
replies	List[str]		A list of strings representing the generated replies.

Init Parameters

These are the parameters you can configure in Pipeline Builder:

Parameter	Type	Default	Description
model	str	Qwen/Qwen3-0.6B	The Hugging Face text generation model name or path.
task	Optional[Literal['text-generation', 'text2text-generation']]	text-generation	The task for the Hugging Face pipeline. The default is text-generation, supported by decoder models like GPT and Qwen. The text2text-generation task (encoder-decoder models like T5) is deprecated and not available in transformers v5+. If not specified, the component infers the task from the model name.
device	Optional[ComponentDevice]	None	The device for loading the model. If None, automatically selects the default device. If a device or device map is specified in huggingface_pipeline_kwargs, it overrides this parameter.
token	Optional[Secret]	Secret.from_env_var(['HF_API_TOKEN', 'HF_TOKEN'], strict=False)	The token to use as HTTP bearer authorization for remote files. If the token is specified in huggingface_pipeline_kwargs, this parameter is ignored.
generation_kwargs	Optional[Dict[str, Any]]	None	A dictionary with keyword arguments to customize text generation: max_length, max_new_tokens, temperature, top_k, top_p. See Hugging Face documentation.
huggingface_pipeline_kwargs	Optional[Dict[str, Any]]	None	Dictionary with keyword arguments to initialize the Hugging Face pipeline. These override model, task, device, and token init parameters. See Hugging Face documentation.
stop_words	Optional[List[str]]	None	If the model generates a stop word, the generation stops. If you provide this parameter, don't specify stopping_criteria in generation_kwargs.
streaming_callback	Optional[StreamingCallbackT]	None	An optional callable for handling streaming responses.

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
prompt	str		A string representing the prompt.
streaming_callback	Optional[StreamingCallbackT]	None	A callback function that is called when a new token is received from the stream.
generation_kwargs	Optional[Dict[str, Any]]	None	Additional keyword arguments for text generation.