Table of Contents

• Configuring RAG
• SWIRL RAG Process:
  • Enterprise RAG Support
  • Preloaded RAG Configurations
  • API-Based RAG Processing
  • Configuring Timeout Behavior

Retrieval Augmented Generation (RAG) Configuration
Community Edition

Please note: we've renamed our products! SWIRL AI Connect is now SWIRL AI Search 🔎 and SWIRL AI Co-Pilot is now SWIRL AI Search Assistant 🤖

---

SWIRL supports Real-Time Retrieval Augmented Generation (RAG) out of the box, using result snippets and/or the full text of fetched result pages.

Configuring RAG

1. Install SWIRL as noted in the Quick Start Guide, including the latest version of the Galaxy UI.

2. Add your OpenAI API key or Azure OpenAI credentials to the .env file:

    OPENAI_API_KEY='your-key-here'
   

   Check out OpenAI's YouTube video if you don't have an OpenAI API Key.

    AZURE_OPENAI_KEY=<your-key>
    AZURE_OPENAI_ENDPOINT=<your-azure-openai-endpoint-url>
    AZURE_MODEL=<your-azure-openai-model>
   

   SWIRL AI Search Community Edition supports RAG only with OpenAI and Azure OpenAI.
   The Enterprise Edition supports additional providers.

3. For PRODUCTION use, update static/api/config/default:

   Change this:

    "webSocketConfig": {
        "url": "ws://<yourhost>:<your-port>/chatgpt-data"
    }
   

   …to this:

    "webSocketConfig": {
        "url": "wss://<yourhost>:<your-port>/chatgpt-data"
    }
   

   Using ws:// locally is fine, but it is not secure for production!

4. Configure SearchProviders for RAG:

   Modify the page_fetch_config_json parameter for each SearchProvider:

    "page_fetch_config_json": {
        "cache": "false",
        "headers": {
            "User-Agent": "Swirlbot/1.0 (+http://swirl.today)"
        },
        "timeout": 10
    }
   

   • Adjust timeout if needed.
   • Change User-Agent if required.
   • Authorize SWIRL to fetch pages from internal applications.

   To override the timeout via the Galaxy UI, use:

    http://localhost:8000/galaxy/?q=gig%20economics&rag=true&rag_timeout=90
   

5. Restart SWIRL:

    python swirl.py restart
   

6. Run a Search in the Galaxy UI:

   • Open http://localhost:8000/galaxy/.
   • Ensure the "Generate AI Response" switch is off.
   • Search for:

     http://localhost:8000/galaxy/?q=SWIRL+AI+Connect
     

   Results appear:
   

7. Manually Select Results for RAG (Optional):

   • Click "Select Items" to enable manual selection.
   • Pre-selected results indicate SWIRL's best matches for RAG.
   • Check/uncheck results, sort, or filter.
   • Click "Generate AI Response".
   • A spinner appears; results follow within seconds.

   

8. Verify citations under the RAG response.

   To cancel a RAG process, toggle "Generate AI Summary" off.

   By default, SWIRL's RAG uses the first 10 selected results (auto or manual).
   To adjust this, set SWIRL_RAG_MAX_TO_CONSIDER in .env, as noted in the AI Search Guide.

SWIRL RAG Process:

SWIRL's RAG workflow is shown above, and explained below:

1. Search - SWIRL sends the users query to one or more SearchProviders, then aggregates and normalizes the results
2. Re-Rank - The last step of the Search workflow re-ranks the aggregated, normalized results to find the most relevant results across all sources
3. Review - SWIRL supports presenting re-ranked results for review and optional adjustment, prior to executing the rest of the workflow
4. Fetch - This involves following the result link and downloading the full text or data set of the result
5. Read - SWIRL extracts text from 1,500 file formats then identifies the most relevant passages while chunking it for the next step
6. Prompt - SWIRL binds the chunked text to the appropriate prompt, connects to the configured LLM, sends the prompt and waits patiently for the response
7. Package - Finally, SWIRL compiles the results, prompt and response and returns it in a neat JSON package, ready for visualization in SWIRL's Galaxy Search UI

Enterprise RAG Support

• The Community Edition can fetch publicly accessible sources.
• For RAG with enterprise services (e.g., Microsoft 365, ServiceNow, Salesforce, Atlassian) with OAuth2 and SSO, contact us for SWIRL Enterprise.

Preloaded RAG Configurations

The following SearchProviders come pre-configured for RAG:

✅ European PMC
✅ Google Web
✅ Google News
✅ LinkedIn
✅ SWIRL Documentation

API-Based RAG Processing

RAG processing is available via a single API call:

?qs=metasearch&rag=true

For details, see the Developer Guide.

Configuring Timeout Behavior

• The default timeout is 60 seconds.
• To modify the timeout and error message, update:

"webSocketConfig": {
    "url": "ws://localhost:8000/chatgpt-data",
    "timeout": 60000,
    "timeoutText": "Timeout: No response from Generative AI."
}