Skip to main content
Version: 2.0

List generation presets

GET 
/v2/generation_presets

Organizations often struggle to fine-tune query responses and maintain consistency across different use cases. Vectara creates and maintains predefined generation presets for our users which provides a flexible and powerful way to utilize generation parameters. Each preset includes a complete Velocity template for the prompt along with other generation parameters. Presets are typically associated with a single LLM.

The List Generation Presets API lets you view the generation presets used for query requests. Generation presets group several properties that configure generation for a request. These presets provide more flexibility in how generation parameters are configured, enabling more fine-tuned control over query responses.

This includes the prompt_template, the Large Language Model (LLM), and other generation settings like max_tokens and temperature. Users specify a generation preset in their query or chat requests using the generation_preset_name field.

Generation presets object​

The generation_presets object contains the name, description, llm_name, prompt_template, and other fields make up the preset.

If your account has access to a preset, then enabled is set to true. A preset can also be set as a default.

Example list generation presets response​

{
  "generation_presets": [
    {
      "name": "vectara-summary-ext-24-05-med-omni",
      "description": "Generate summary with controllable citations, Uses GPT-4o with 2,048 max tokens",
      "llm_name": "gpt-4o",
      "prompt_template": "[\n {\"role\": \"system\", \"content\": \"Follow these detailed step-by-step.",
      "max_used_search_results": 25,
      "max_tokens": 2048,
      "temperature": 0,
      "frequency_penalty": 0,
      "presence_penalty": 0,
      "enabled": true,
      "default": false
    }
    // More presets appear here
    ]
  }

Request​

Responses​

List of Generation Presets.

Last updated on by Paul Wozniczka