Spaces:
Running
Running
File size: 3,499 Bytes
760c69b 066b132 760c69b 632def0 bb2ac09 3cd1290 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 |
---
title: LLMSearchEngine
emoji: 🏆
colorFrom: gray
colorTo: purple
sdk: docker
app_file: app.py
pinned: false
---
# LLM Search Engine
This is a Flask-based web application that uses a large language model (LLM) to generate search engine-like results, styled to resemble Google’s classic search results page. Instead of querying an external search API, it prompts an LLM to create titles, snippets, and URLs for a given query, delivering a paginated, familiar interface.
## Why We Built It
We created this app to explore how LLMs can mimic traditional search engines by generating results directly from their training data. It offers:
- A nostalgic, Google-like pagination design with clickable links.
- A proof-of-concept for LLM-driven search without real-time web access.
- A simple, self-contained alternative for queries within the model’s knowledge base.
## Features
- **Google-Styled Interface**: Search bar, result list, and pagination styled with Google’s colors and layout.
- **Generated Results**: Titles, snippets, and URLs are fully produced by the LLM.
- **Pagination**: Displays 10 results per page, up to 30 total results across 3 pages.
## Limitations
- **Static Knowledge**: Results are limited to the LLM’s training cutoff (e.g., pre-2025).
- **Generated Content**: URLs and snippets may not correspond to real web pages—use as a starting point.
- **No Real-Time Data**: Best for historical or established topics, not breaking news.
## Using It on Hugging Face Spaces
### Try the Demo
Deployed on Hugging Face Spaces, you can test it at [https://codelion-llmsearchengine.hf.space](https://codelion-llmsearchengine.hf.space):
1. Open the URL in your browser.
2. Type a query (e.g. "best Python libraries") in the search bar and press Enter or click "LLM Search".
3. Browse the paginated results, styled like Google, using "Previous" and "Next" links.
## Using It as an API
### API Endpoint
Your app doubles as an API when hosted on HF Spaces:
- **URL:** `https://codelion-llmsearchengine.hf.space/`
- **Method:** `GET`
- **Parameters:**
- `query`: The search query (e.g., `"best Python libraries"`).
- `page`: Page number (`1-3`, defaults to `1`).
### Example Request
```bash
curl "https://codelion-llmsearchengine.hf.space/?query=best+Python+libraries&page=1"
```
### Response
Returns **raw HTML** styled like a Google search results page.
---
## Integration
You can fetch results programmatically and render or parse the HTML:
```python
import requests
from urllib.parse import quote
query = "best Python libraries"
page = 1
url = f"https://codelion-llmsearchengine.hf.space/?query={quote(query)}&page={page}"
response = requests.get(url)
html_content = response.text # Render or process as needed
print(html_content)
```
---
## How It Works
1. **LLM Prompting**
- Queries trigger a prompt to the `"gemini-2.0-flash-lite"` model.
- Generates **30 results** in JSON format.
2. **Rendering**
- Flask converts results into a **Google-styled** HTML page.
- Includes a **search bar, results, and pagination**.
3. **Deployment**
- Runs via **Flask** and **Docker** on HF Spaces.
- Serves **dynamic pages** based on URL parameters.
---
## Setup Locally
### Install dependencies
```bash
pip install -r requirements.txt
```
### Set environment variables
```bash
export OPENAI_API_KEY="your-key"
export OPENAI_BASE_URL="your-url"
```
### Run the app
```bash
python app.py
```
Visit `http://localhost:5000`. |