Skip to content

Latest commit

 

History

History
67 lines (52 loc) · 4.02 KB

test.mdx

File metadata and controls

67 lines (52 loc) · 4.02 KB
Raw

Generate Text

The Generate Text node in Magick allows you to generate natural language text using AI language models. It provides a flexible and powerful way to create dynamic text content within your spells.

Inputs

  • flow: The input flow that triggers the text generation.
  • prompt (string, default: ""): The prompt or initial text to guide the language model in generating the text.
  • system (string, default: ""): A system message that sets the context or role for the language model.
  • messages (array, default: []): An array of message objects representing the conversation history.
  • modelOverride (string, default: ""): Allows overriding the default language model with a specific model name.
  • stop (string, default: ""): A string or list of strings to stop the text generation when encountered.
  • temperature (float, default: 0.5): Controls the randomness and creativity of the generated text. Higher values result in more diverse outputs.
  • maxRetries (integer, default: 3): The maximum number of retries if the text generation fails.
  • top_p (integer, default: 1): Controls the diversity of the generated text by limiting the cumulative probability of token selection.
  • seed (integer, default: 42): A seed value for reproducible results.
  • maxTokens (integer, default: 256): The maximum number of tokens (words or subwords) to generate in the output text.

Outputs

  • response (string): The generated text based on the provided inputs.
  • completionResponse (object): The complete response object returned by the language model, containing additional metadata.
  • done: The output flow indicating the completion of the text generation.
  • onStream: The output flow triggered for each token or chunk of text generated during streaming.
  • stream (string): The stream of generated text, emitted incrementally during streaming.

Configuration

  • modelProviders (array, default: []): An array of available language model providers.
  • modelProvider (string, default: "openai"): The selected language model provider.
  • models (array, default: []): An array of available language models for the selected provider.
  • model (string, default: "gpt-3.5-turbo"): The selected language model for text generation.
  • customBaseUrl (string, default: ""): A custom base URL for the language model API, if applicable.
  • hiddenProperties (string, default: ["hiddenProperties", "modelProvider", "model", "models", "customBaseUrl"]): Properties to be hidden from the node's configuration interface.

Usage

  1. Connect the desired input flow to the flow input of the Generate Text node.
  2. Set the prompt input to provide the initial text or question to guide the language model.
  3. Optionally, set the system input to establish the context or role for the language model.
  4. If needed, provide an array of messages representing the conversation history.
  5. Adjust the temperature, top_p, and maxTokens inputs to control the randomness, diversity, and length of the generated text.
  6. Connect the done output to the next node in your spell to continue the flow after text generation.
  7. Retrieve the generated text from the response output.

Example

Here's an example of how to use the Generate Text node in a Magick spell:

1. Start
2. Generate Text
   - prompt: "Once upon a time, in a magical land far away..."
   - temperature: 0.7
   - maxTokens: 100
3. Log
   - message: "[Generated Text]"
4. End

In this example, the Generate Text node is used to generate a fairy tale-like story. The prompt input sets the beginning of the story, and the temperature and maxTokens inputs control the creativity and length of the generated text. The generated text is then logged using the Log node.

Best Practices

  • Provide clear and specific prompts to guide the language model towards the desired output.
  • Experiment with different temperature and top_p values to find the right balance between creativity and coherence.
  • Use the stop input to define specific strings or patterns that should terminate the text generation.