RegexTextExtractor

Extract text from chat messages or strings using a regex pattern.

Key Features

• Parses input text or ChatMessage objects using a regular expression you provide
• Supports capture groups to extract a specific part of a match
• Returns the entire match if no capture group is defined
• Can search through all messages or just the last message in a list of ChatMessage objects
• Returns an empty string when no match is found (configurable)
• Useful in query pipelines for extracting structured data from LLM responses

Configuration

1. Drag the RegexTextExtractor 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 regular expression pattern to use for extraction. Include a capture group to extract specific text from a match. For example, <issue url="([^"]+)"> captures the URL from the tag.
4. Go to the Advanced tab to configure return_empty_on_no_match.

Connections

RegexTextExtractor accepts either a string or a list of ChatMessage objects as input. It outputs the matched text as a string (captured_text). It is typically used in query pipelines to extract structured data from a ChatGenerator's output, and it sends the extracted text to components such as PromptBuilder, AnswerBuilder, or Retrievers.

Usage Example

This query pipeline uses a ChatGenerator to analyze text and produce structured output, then uses RegexTextExtractor to extract a specific issue URL from the response:

components:
  ChatPromptBuilder:
    type: haystack.components.builders.chat_prompt_builder.ChatPromptBuilder
    init_parameters:
      template:
        - _content:
            - text: |
                You are a helpful assistant that identifies issues in text.
                When you find an issue, format your response as:
                <issue url="https://github.com/example/repo/issues/123">Description of the issue</issue>
                
                Analyze the following text and identify any issues:
                {{ query }}
          _role: user
      required_variables:
      variables:

  OpenAIChatGenerator:
    type: haystack.components.generators.chat.openai.OpenAIChatGenerator
    init_parameters:
      model: gpt-4o-mini
      generation_kwargs:
        temperature: 0.3

  RegexTextExtractor:
    type: haystack.components.extractors.regex_text_extractor.RegexTextExtractor
    init_parameters:
      regex_pattern: '<issue url="([^"]+)">'
      return_empty_on_no_match: true

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

connections:
  - sender: ChatPromptBuilder.prompt
    receiver: OpenAIChatGenerator.messages
  - sender: OpenAIChatGenerator.replies
    receiver: RegexTextExtractor.text_or_messages
  - sender: RegexTextExtractor.captured_text
    receiver: AnswerBuilder.replies

inputs:
  query:
    - ChatPromptBuilder.query
    - AnswerBuilder.query

outputs:
  answers: AnswerBuilder.answers

In this example:

1. ChatPromptBuilder creates a prompt asking the LLM to identify issues and format them with a specific XML-like tag.
2. OpenAIChatGenerator generates a response containing the structured output.
3. RegexTextExtractor uses the pattern <issue url="([^"]+)"> to extract the URL from the response. The capture group ([^"]+) matches any characters except quotes inside the url attribute.
4. AnswerBuilder formats the extracted URL as the final answer.

Parameters

Inputs

Parameter	Type	Default	Description
text_or_messages	Union[str, List[ChatMessage]]		Either a string or a list of ChatMessage objects to search through.

Outputs

Parameter	Type	Default	Description
captured_text	str		The matched text if a match is found. Empty string if no match is found and return_empty_on_no_match=False.

Init Parameters

These are the parameters you can configure in Pipeline Builder:

Parameter	Type	Default	Description
regex_pattern	str		The regular expression pattern used to extract text. The pattern should include a capture group to extract the desired text. Example: '<issue url="(.+)">' captures the URL from the tag.
return_empty_on_no_match	bool	True	If True, returns an empty dictionary when no match is found. If False, returns {"captured_text": ""}.

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
text_or_messages	Union[str, List[ChatMessage]]		Either a string or a list of ChatMessage objects to search through.