OpenAI

Raycast Unblock provides an OpenAI feature. You can use it to generate text using the OpenAI API.

Usage

1. Set AI.default to openai in your configuration file.
2. Set AI.OpenAI.api_key to your OpenAI API key in your configuration file.
3. (optional) Set AI.OpenAI.default to the default model you want to use in your configuration file.
4. (optional) You can add your custom models to the AI.OpenAI.Models in your configuration file.

Custom Models

You can add your custom models to the AI.OpenAI.Models in your configuration file.

The definition of a custom model is like this:

[AI.OpenAI.Models.model_name]

You shouldn't use the dot in the model name. It will be parsed as a section. For example, GPT-3.5 should be GPT3_5 or other names without a dot.

Other parameters you can see in the example below.

Configuration

• temperature: The temperature of the model. Optional
• max_tokens: The maximum tokens of the model. Optional
• api_key: Your OpenAI API key.
• default: The default model to use. Optional
• Models: The custom models to use. Optional
  • Models.<model>: The model name.
    • base_url: The base URL of the model. Optional ^v0.5.0-beta.2
    • api_key: The API key of the model. Optional ^v0.5.0-beta.2
    • real_id: The real id of the model. Optional ^v0.5.0-beta.2
    • provider_name: The provider name of the model. Optional ^v0.5.0-beta.2
    • id: The model id. It should be unique.
    • model: The model name.
    • name: The model name.
    • description: The model description.
    • speed: The speed of the model.
    • intelligence: The intelligence of the model.
    • context: The context of the model.
    • status: The status of the model. Optional
    • capabilities: The capabilities of the model. Optional
      • image_generation: The image generation capability. (Need DALL supported) Optional
      • web_search: The web search capability. (Function Call) Optional
      • vision: The vision capability. (Need model supported) Optional
• Endpoints: Custom endpoints Optional ^v0.7.0-beta.1
  • base_url: The base URL of the endpoint. Optional
  • api_key: The API key of the endpoint. Optional
  • tag: The tag of the endpoint. It will be displayed in the model list. Optional
  • display_models: The models to display in the model list. Once you set this parameter, the model list will be displayed in the order you set. Optional
  • addon_models: The models to add to the model list. Optional
  • banned_models: The models to ban from the model list. Optional

Q: When to use the Endpoints parameter, and what is the difference between Models and Endpoints?

Answer

When you use the Endpoints parameter, you can add multiple endpoints in your configuration file, and Ray.Pro will automatically get the model list from these endpoints.

You can use the display_models parameter to control the displayed model list, the addon_models parameter to add additional models to the model list, and the banned_models parameter to ban the displayed models.

Also, Ray.Pro will automatically match the model information based on the Model ID, so you don't need to add the model information to the configuration file.

When you use the Models parameter, you need to manually add the model information to the configuration file. This is suitable for when you only need to use a single model.

Tips: When to use the base_url parameter?

You can use the base_url parameter when you want to use a different base URL for the model. For example, you can use it to use a different endpoint for the model.

Also, api_key and real_id parameters are used for the same purpose. You can use them when you want to use a different API key or real id for the model.

real_id is the real id of the model. It's used to request the model from the api. If you don't provide it, the id parameter will be used as the real id.

Example for real_id, api_key, and base_url

• Model1
  • id: something-endpoint-gpt-4-0125-preview
  • real_id: gpt-4-0125-preview
  • base_url: https://something-endpoint.com/v1
  • Actually, the request will be sent to https://something-endpoint.com/v1/chat/completions, and model id will be gpt-4-0125-preview.
• Model2
  • id: another-endpoint-gpt-4-0125-preview
  • real_id: gpt-4-0125-preview
  • base_url: https://another-endpoint.com/v1
  • Actually, the request will be sent to https://another-endpoint.com/v1/chat/completions, and model id will be gpt-4-0125-preview.

Example

[AI.OpenAI]
api_key = '<your api key>'
default = ''
# temperature = 0.5
# max_tokens = 100

[AI.OpenAI.Models.GPT4]
id = "gpt-4-0125-preview"
model = "gpt-4-0125-preview"
name = "GPT-4 (Preview)"
description = "GPT-4 is OpenAI’s most capable model with broad general knowledge, allowing it to follow complex instructions and solve difficult problems.\n"
speed = 3
intelligence = 3
context = 8
status = "beta"

[AI.OpenAI.Models.GPT4.Capabilities]
image_generation = "dall-e-2" # Write generation model.
web_search = true # The premise is that the model needs to support Function Call. Or if the model defaults to having network access and cannot be turned off, you need to set this parameter to fixed.
vision = true # The model needs to support vision capability.