https://github.com/developersdigest/internet-enabled-llms
https://github.com/developersdigest/internet-enabled-llms
Last synced: 7 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/developersdigest/internet-enabled-llms
- Owner: developersdigest
- Created: 2024-10-30T16:26:50.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2024-10-30T23:59:11.000Z (over 1 year ago)
- Last Synced: 2025-04-12T09:08:40.532Z (11 months ago)
- Language: TypeScript
- Size: 252 KB
- Stars: 8
- Watchers: 1
- Forks: 2
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Internet-Enabled LLMs
This API provides a powerful and flexible solution for web data extraction and LLM-powered question answering. It uses multiple integrations, including Firecrawl, OpenAI, Not Diamond, and Serper, to collect, process, and provide relevant information from web pages, enhancing user queries with contextual data.
## Getting Started
### Installation
1. Clone the repository:
```sh
git clone https://github.com/developersdigest/internet-enabled-llms.git
cd internet-enabled-llms
```
2. Install dependencies:
```sh
bun install
```
3. Create a `.env` file in the root directory and add your API keys:
```env
FIRECRAWL_API_KEY=your_firecrawl_api_key
NOTDIAMOND_API_KEY=your_notdiamond_api_key
OPENAI_API_KEY=your_openai_api_key
ANTHROPIC_API_KEY=your_anthropic_api_key
GOOGLE_API_KEY=your_google_api_key
SERPER_API_KEY=your_serper_api_key
```
### Running the Server
To start the server, run:
```sh
bun dev
```
This will start the server in development mode using Next.js.
## Environment Variables
The API requires the following environment variables:
- `FIRECRAWL_API_KEY` - Firecrawl API key for web scraping.
- `NOTDIAMOND_API_KEY` - NotDiamond API key for LLM handling.
- `OPENAI_API_KEY` - OpenAI API key for embeddings and LLMs.
- `ANTHROPIC_API_KEY` - Anthropic API key for LLMs.
- `GOOGLE_API_KEY` - Google API key for LLMs.
- `SERPER_API_KEY` - Serper API key for web search.
## Endpoints
### POST Request
- **URL**: `/api/llm/`
- **Method**: `POST`
- **Description**: Processes a user message, searches for relevant pages, scrapes the content, optionally uses embeddings, and then provides an LLM-powered response.
#### Request Body
The request body should be in JSON format with the following fields:
- `message` (string, required): The user query.
- `pagesToCrawl` (number, optional): Number of pages to crawl for context (default is 3, max is 10).
- `skipEmbeddings` (boolean, optional): If `true`, embeddings are skipped and raw content is processed instead.
Example:
```json
{
"message": "When did chatgpt canvas come out?",
"pagesToCrawl": 5,
"skipEmbeddings": false
}
```
#### Response
The response is in JSON format and includes:
- `answer` (string): The generated response from the LLM.
- `selectedModel` (object): Details about the model used.
- `crawlInfo` (object): Information about the crawling and embedding process.
- `sources` (array): A list of sources that were used to generate the response.
Example Response:
```json
{
"answer": "ChatGPT Canvas came out on October 3, 2024.",
"selectedModel": {
"provider": "openai",
"model": "gpt-4o-mini"
},
"crawlInfo": {
"requestedPages": 5,
"actualPagesCrawled": 4,
"embeddingsUsed": true
},
"sources": [
{
"title": "ChatGPT Canvas Release Date",
"url": "https://example.com/chatgpt-canvas-release"
}
]
}
```
## Detailed Flow
1. **Request Parsing**: Extract the user query and validate the request.
2. **Web Search & Filtering**: Search using Serper and filter out unsupported domains (e.g., social media).
3. **Scraping & Content Processing**: Scrape relevant content from filtered URLs and optionally create embeddings.
4. **LLM Query Handling**: Generate an answer using one of the LLMs (OpenAI, Anthropic, or Gemini).
5. **Return the Response**: Send the answer along with source and model information.
## License
This project is licensed under the MIT License.