Cameron
52 lines
1.6 KiB
Markdown
52 lines
1.6 KiB
Markdown
# AI Answers for SearXNG
|
|
|
|
**Does not block result loading time.**
|
|
|
|
A SearXNG plugin that generates an AI answer using search results as RAG grounding context. Supports Google Gemini and OpenAI-compatible providers (OpenRouter, Ollama, OpenAI API etc.).
|
|
|
|
Features token by token UI updates as response is recieved.
|
|
|
|
## Installation
|
|
|
|
Place `ai_answers.py` into the `searx/plugins` directory of your instance (or mount it in a container) and enable it in `settings.yml`:
|
|
|
|
```yaml
|
|
plugins:
|
|
searx.plugins.ai_answers.SXNGPlugin:
|
|
active: true
|
|
```
|
|
|
|
## Configuration
|
|
|
|
Set the following environment variables:
|
|
|
|
### General
|
|
|
|
- `LLM_PROVIDER`: `openrouter` (default) or `gemini`. (openrouter for all OpenAI APIs)
|
|
- `RESPONSE_MAX_TOKENS`: Defaults to `500`.
|
|
- `RESPONSE_TEMPERATURE`: Defaults to `0.2`.
|
|
|
|
### OpenRouter / OpenAI / Ollama
|
|
(for any OpenAI compatible API, will revise naming clarity in update soon)
|
|
- `OPENROUTER_API_KEY`: Your API key.
|
|
- `OPENROUTER_MODEL`: Defaults to `google/gemma-3-27b-it:free`.
|
|
- `OPENROUTER_BASE_URL`: Defaults to `openrouter.ai`. (Change to `localhost:11434` for Ollama, or base url of target OpenAI-compatible API).
|
|
|
|
### Google Gemini
|
|
|
|
- `GEMINI_API_KEY`: Your Google AI API key.
|
|
- `GEMINI_MODEL`: Defaults to `gemma-3-27b-it`.
|
|
|
|
## How It Works
|
|
|
|
After search completes, the plugin extracts the top 6 results as context. A client-side script calls the stream endpoint with a signed token. The LLM response streams back. Token by token rendering is soon.
|
|
|
|
## Ollama (Local)
|
|
|
|
```
|
|
LLM_PROVIDER=openrouter
|
|
OPENROUTER_API_KEY=ollama
|
|
OPENROUTER_MODEL=gemma3:27b
|
|
OPENROUTER_BASE_URL=localhost:11434
|
|
```
|