Nice Prompt REST API

Nice prompt is a REST API offers an intuitive way for developers and users to enhance their prompts using advanced language models. This API supports both OpenAI and Ollama, providing versatile options for generating and improving prompt responses.

showcase.mov

Features

• Generate dynamic responses with OpenAI's language models.
• Utilize Ollama for improving your prompts locally.
• Easy-to-use API endpoints tailored for prompt generation and improvement.

Getting Started

Begin by obtaining an API key from OpenAI or if you want to use Ollama, setting up your local Ollama server following the Ollama documentation.

Prerequisites

• An OpenAI API key, available through OpenAI.
• Local Ollama server setup for those preferring or requiring local computations.

Using the API

There are two primary endpoints for interaction:

Endpoint	Description
/api/v1/openai	For generating content with OpenAI.
/api/v1/ollama	For generating content with local Ollama server.

OpenAI

Send a POST request to /api/v1/openai with the following JSON payload and api key in headers as "x-api-key":

Parameter	Required	Default	Description
prompt	Yes	N/A	The prompt you want to generate a response for.
model	No	gpt-3.5-turbo	The model to use for generation.
temperature	No	0.7	Controls the randomness of the output.
max_tokens	No	150	The maximum length of the output.
top_p	No	1	Controls the diversity of the output.
frequency_penalty	No	0	Penalizes frequently used tokens.
presence_penalty	No	0	Penalizes new tokens.
x-api-key	Yes	N/A	Your OpenAI API key, in headers.

{
  "prompt": "How to learn code?", // Required
  "model" : "gpt-3.5-turbo",  // Default
  "temperature": 0.7,
  "max_tokens": 150,
  "top_p": 1,
  "frequency_penalty": 0,
  "presence_penalty": 0,
}

Here is response;

{
    "prompt": "What are the most effective strategies and resources for beginners to learn coding, especially for someone with no prior programming experience?"
}

Local LLMS with Ollama

For the Nice Prompt API to communicate with your local Ollama server, you can use localtunnel to expose your local server. Here's how to set it up:

1. Install localtunnel globally on your machine:

npm install -g localtunnel

Start your local Ollama server on the default port (or configure it as needed).

Use localtunnel to expose your local server:

lt --port 11434

This command will generate a publicly accessible URL (e.g., https://yoursubdomain.loca.lt) that proxies all requests to your local Ollama server.

When using the Nice Prompt API /api/v1/ollama endpoint, include the generated localtunnel URL in the request header as X-LocalTunnel-Url.

This setup allows your API to direct requests to your local Ollama server through the localtunnel URL.

Ensure your local Ollama server is running and exposed via localtunnel. Then, send a POST request to /api/v1/ollama with the following JSON payload and your localtunnel URL in the X-LocalTunnel-Url header:

Parameter	Required	Default	Description
prompt	Yes	N/A	The prompt you want to generate a response for.
model	No	llama2	The model to use for generation.
temperature	No	0.5	Controls the randomness of the output.
max_tokens	No	100	The maximum length of the output.
top_p	No	1	Controls the diversity of the output.
frequency_penalty	No	0	Penalizes frequently used tokens.
presence_penalty	No	0	Penalizes new tokens.

{
  "prompt": "How to learn code?", // Required
  "model": "llama2", // Default
  "temperature": 0.5,
  "max_tokens": 100,
  "top_p": 1,
  "frequency_penalty": 0,
  "presence_penalty": 0
}

Here is response;

{
    "prompt": "How do I effectively learn programming languages \n and develop practical coding skills to tackle real-world problems and projects? What are the best resources, strategies, and practices for acquiring knowledge and proficiency in software development?"
}

Note: The Ollama server must be started locally for this endpoint to function. Follow the setup instructions in the Ollama documentation.