Google Deep Research and Open WebUI Integration Stack

A complete Docker Compose stack featuring Google's Deep Research with persistent session resumption, integrated with Open WebUI for a powerful research-focused AI experience. Also supports Anthropic Claude and xAI Grok models via an OpenAI-compatible API bridge.

🔬 Deep Research - Primary Feature

Google Deep Research is an advanced AI research agent that provides comprehensive, multi-step analysis:

• Persistent Research Sessions: Automatic interaction resumption for long-running research
• Multi-Step Reasoning: Extended thinking with transparent research process
• Background Processing: Research continues even if connection drops
• Thought Summaries: Real-time insights into the research methodology
• OpenAI-Compatible: Works seamlessly with Open WebUI and other OpenAI-compatible clients

⚠️ Important: Rate Limits & Required Concessions

Deep Research has a 1 request per minute (RPM) limit. To avoid wasting your quota:

In Open WebUI Settings:

1. Go to Settings → Interface
2. Disable "Auto-Generate Title" - This would consume a Deep Research request just to generate a chat title
3. Disable "Auto-Follow-Up Prompts" - This would consume a Deep Research request for suggestion generation

Why this matters:

• Deep Research is designed for comprehensive, time-intensive analysis (30-60+ seconds per query)
• Auto-title and auto-follow-up features make rapid API calls that waste your limited quota
• Session resumption allows you to continue research without consuming additional quota

🚀 Features

Core Functionality

• 🔬 Google Deep Research Integration: First-class support with automatic session resumption
• OpenAI-Compatible API Bridge: FastAPI-based service for multiple AI providers
• Open WebUI Interface: Modern, feature-rich web UI for interacting with AI models
• Multi-Provider Support:
  • Google Deep Research & Gemini: Advanced research agent with thinking models
  • Anthropic Claude: Latest Claude models with extended context
  • xAI Grok: Latest Grok models
• Streaming Support: Real-time streaming responses with reasoning content
• Bearer Token Authentication: Secure API access with custom token management
• Docker-based Deployment: Easy setup and management with Docker Compose

Deep Research Capabilities

• Persistent Sessions: Research sessions stored in SQLite database
• Automatic Resumption: Continue research without starting over (doesn't consume RPM)
• Connection Recovery: Auto-reconnects if connection drops
• Session Tracking: Clear SDK messages showing session status
• MD5-based Session Matching: Same query resumes existing research

Open WebUI Integration

• RAG (Retrieval Augmented Generation): Upload documents and chat with your data
• Speech-to-Text: Audio transcription powered by OpenAI's STT engine
• Persistent Storage: Conversation history and uploaded documents preserved
• Multi-user Support: Bearer token authentication for multiple users

📋 Prerequisites

• Docker and Docker Compose installed
• Google API key (required for Deep Research) - Get one here
• Optional API keys:
  • Anthropic API key (get one here)
  • xAI API key (get one here)

🔧 Setup

1. Clone the repository

   git clone <your-repo-url>
   cd open-webui-stack
   

2. Set up the SDK interface

   cd sdk-interface
   cp .env.example .env
   # Edit .env and add your Google API key (required)
   

3. Run database migrations

   cd sdk-interface
   yoyo apply
   

4. Create the Docker volume

   docker volume create open-webui
   

5. Start the stack

   cd ..
   docker-compose up -d
   

6. Configure Open WebUI

   • Open your browser: http://localhost:8090
   • Add the SDK interface as an OpenAI connection
   • Important: Disable auto-title and auto-follow-up features (see warning above)

🏗️ Architecture

1. SDK Interface (sdk-interface)

• Purpose: OpenAI-compatible API bridge featuring Google Deep Research
• Technology: FastAPI, Python 3.11+, SQLite
• Port: Internal only (accessed via Docker network)
• Key Features:
  • Deep Research session persistence and resumption
  • Converts OpenAI format to provider-specific APIs
  • Handles streaming with reasoning content
  • Bearer token authentication
  • Database: data/db.sqlite3 with research_hashes table

Endpoints:

• GET /v1/models - List all available models
• POST /v1/chat/completions - Chat completions (Deep Research, Claude, Gemini, Grok)
• GET /health - Health check

2. Open WebUI (open-webui)

• Purpose: Modern web interface optimized for Deep Research
• Image: ghcr.io/open-webui/open-webui:main
• Port: 8090
• Features: Rich markdown, file upload, RAG, conversation history

3. Watchtower (watchtower)

• Purpose: Automatic container updates
• Schedule: Daily at 2 AM

🔐 Authentication

Bearer token authentication with op_wui_ prefix:

# In sdk-interface/.env
API_KEYS=username1:op_wui_token1;username2:op_wui_token2

Generate tokens:

python sdk-interface/scripts/generate_token.py

💡 Usage Example

Deep Research Query

from openai import OpenAI

client = OpenAI(
    api_key="op_wui_your_token_here",
    base_url="http://localhost:8060/v1"
)

# First request - starts new research (consumes 1 RPM)
response = client.chat.completions.create(
    model="deep-research-pro-preview-12-2025",
    messages=[
        {"role": "user", "content": "Research the latest developments in quantum computing"}
    ],
    stream=True
)

for chunk in response:
    delta = chunk.choices[0].delta

    # Research reasoning/thoughts
    if hasattr(delta, 'reasoning_content'):
        print(f"[THINKING] {delta.reasoning_content}")

    # Final research output
    if delta.content:
        print(delta.content, end="")

# Same request later - automatically resumes (FREE - no RPM consumed)
# Output: [SDK] Continuing interaction with id v1_...

🛠️ Configuration

Environment Variables

SDK Interface (sdk-interface/.env):

• GOOGLE_API_KEY - Your Google API key (required for Deep Research)
• ANTHROPIC_API_KEY - Your Anthropic API key (optional)
• GROK_API_KEY - Your xAI API key (optional)
• API_KEYS - Bearer tokens in format username:token;username:token
• LOG_LEVEL - Logging level (default: info)

Open WebUI Configuration:

• Disable auto-title generation
• Disable auto-follow-up prompts
• Configure API endpoint: http://sdk-interface:8060/v1

📦 Available Models

Google Deep Research (Primary)

• deep-research-pro-preview-12-2025 - Advanced research agent

Google Gemini

• gemini-2.0-flash-exp
• gemini-2.0-flash-thinking-exp
• gemini-1.5-pro-latest
• And more...

Anthropic Claude

• claude-opus-4-5-20251101
• claude-sonnet-4-5-20250929
• claude-3-5-sonnet-20241022
• And more...

xAI Grok

• grok-2-vision-1212
• grok-2-1212
• And more...

Full model list available at: http://localhost:8060/v1/models

🔄 Deep Research Session Management

How it works:

1. Each unique message set generates an MD5 hash
2. Hash is stored in database with interaction_id when research starts
3. Same query later resumes from stored interaction_id
4. Connection drops are automatically handled with reconnection
5. Session resumption does not consume RPM quota

Session Status Messages:

• [SDK] Connecting to Deep Research Agent... - New session starting
• [SDK] Interaction started... - Research session created
• [SDK] Continuing interaction with id v1_... - Resuming existing session

📝 Development

The SDK interface is mounted as a volume for development:

volumes:
  - ./sdk-interface:/sdk-interface/app

Changes require container restart. See sdk-interface/README.md for detailed API documentation.

⚠️ Disclaimer

This is an unofficial community project and is not affiliated with or endorsed by:

• Open WebUI Inc. or its contributors
• Google LLC or Alphabet Inc.
• Anthropic PBC
• xAI Corp.

🔗 Links

• SDK Interface Documentation - Detailed API documentation
• Open WebUI Official Repository
• Google AI Studio
• Anthropic
• xAI

📄 License

Please refer to the individual component licenses:

• Open WebUI: See upstream repository
• SDK Interface: See LICENSE file in this repository