Search Results Summarizer

AI-powered search results summary generator that works transparently with existing search engine instances (SearXNG and 4get) using OpenAI-compatible AI providers.

demo.mp4

Installation

Option 1: Using the Docker Image

Method A: With Docker Run

1. Create a configuration directory:

   mkdir -p ./search-results-summarizer/config
   cd ./search-results-summarizer

2. Run the container:

   docker run -d --name search-results-summarizer \
     -p 3000:3000 \
     -v "./config:/config" \
     --restart unless-stopped \
     ghcr.io/chandujr/search-results-summarizer:latest

3. Configure the service:

   # Edit the configuration files
   nano ./config/config.yaml
   nano ./config/.env

4. Restart the container to apply changes:

   docker restart search-results-summarizer

Method B: With Docker Compose

1. Create a project directory and docker-compose.yml:

   mkdir -p ./search-results-summarizer/config
   cd ./search-results-summarizer
   cat > docker-compose.yml << EOF
   services:
     search-results-summarizer:
       image: ghcr.io/chandujr/search-results-summarizer:latest
       container_name: search-results-summarizer
       ports:
         - "3000:3000"
       volumes:
         - ./config:/config
       restart: unless-stopped
   EOF

2. Start the service:

   docker-compose up -d

3. Configure the service:

   # Edit the configuration files
   nano ./config/config.yaml
   nano ./config/.env

4. Restart the service to apply changes:

   docker-compose restart

Option 2: Using the Repository

1. Clone the repository:

   git clone https://github.com/chandujr/search-results-summarizer.git
   cd search-results-summarizer

2. Copy the configuration templates:

   cp config/config.yaml.default config/config.yaml
   cp config/.env.default config/.env

3. Edit your configuration:

   nano config/config.yaml
   nano config/.env

4. Build and run with Docker Compose:

   docker-compose up -d --build

Configuration

In config/.env, set your API key based on your provider:

# For OpenRouter
OPENROUTER_API_KEY=your_api_key_here

In config/config.yaml, configure:

Search engine

• ENGINE_NAME: "searxng" or "4get"
• ENGINE_URL: URL of your search engine instance (see networking note below)
  • Works with both locally installed and public instances

AI configuration

• SUMMARIZER_LLM_URL: URL of the AI provider to use for summarization
• SUMMARIZER_MODEL_ID: AI model to use for summarization
• CLASSIFIER_LLM_URL: URL of the AI provider to use for deciding whether to show summary in smart mode
• CLASSIFIER_MODEL_ID: AI model to use for deciding whether to show summary in smart mode
  • Should be a model that supports function/tool calling
• MAX_TOKENS: Maximum tokens for AI responses (default: 750)

Performance

• SUMMARY_MODE: "auto" (automatic), "manual" (button-triggered), or "smart" (AI decides when to summarize)
• MAX_RESULTS_FOR_SUMMARY: Number of results to summarize (default: 7)

Connection

• MODIFY_CSP_HEADERS: Set to true if using public search engine instances that block external scripts (default: false)
• TRUST_PROXY: Set to true when running behind a reverse proxy - needed for proper rate limiting (default: false)
• PROXY_IP_RANGE: IP range of trusted proxy when TRUST_PROXY is enabled (default: "10.0.0.0/8" for Render)
• RATE_LIMIT_MS: Rate limit in milliseconds between requests (default: 1000)

Summarization Filters (only in "auto" mode)

• MIN_KEYWORD_COUNT: Minimum number of keywords required (default: 3)
• MIN_RESULT_COUNT: Minimum search results required (default: 3)
• EXCLUDE_WORDS: Words that prevent summarization
• EXCLUDE_OVERRIDES: Words that override the exclude list and force summarization

Note: The "smart" mode bypasses these filters and uses AI to determine if summarization is needed.

Docker Networking Note

You may not be able to use localhost for URLs since from within a Docker container, localhost refers to the container itself, not the host machine.

To connect to your installed services (search engine or Ollama) running on your host machine:

1. Use the Docker network gateway IP:

   # Find your Docker network name
   docker network ls | grep search-results-summarizer
   
   # Inspect the network to find the gateway IP
   docker network inspect <network_name> | grep Gateway
   # Example result: "Gateway": "172.19.0.1"
   # Set ENGINE_URL to: http://172.19.0.1:8081 (for search engine)

2. Use your host machine's IP address:

   # Find your host IP
   ip route get 1.1.1.1 | awk '{print $7}'
   # Example result: 192.168.1.100
   # Set ENGINE_URL to: http://192.168.1.100:8081 (for search engine)

Usage

1. Access the service at http://localhost:3000 (or your configured port) to verify it's working
2. Visit the service once, then you can set it as your default search engine through your browser's settings

Most modern browsers allow you to add this as a search engine.

Privacy

• Summaries are generated by sending search query and results to your selected AI provider
• No data is stored by this proxy
• Consider privacy implications before use

License

AGPL-3.0