OpenWebUI Integration¶

Connect OpenWebUI to stdapi.ai as an OpenAI-compatible backend. Access Amazon Bedrock models through OpenWebUI's chat interface with no code changes required.

About OpenWebUI¶

🔗 Links: Website | GitHub | Documentation | Discord

OpenWebUI is a feature-rich, self-hosted web UI for AI models:

• 40,000+ GitHub stars - Popular open-source AI web interface
• Feature-complete - Matches ChatGPT's interface capabilities
• Multi-modal - Chat, voice input/output, image generation, and document RAG
• Extensible - Plugin system, custom functions, and community tools
• Privacy-focused - Self-hosted with no external dependencies

Why OpenWebUI + stdapi.ai?¶

Key Benefits:

• Familiar interface - Continue using OpenWebUI as before
• Enterprise models - Access Anthropic Claude, Amazon Nova, and more
• Full feature set - Chat, voice, image generation, and RAG all work
• Privacy - Your data stays in your AWS environment
• Cost efficient - AWS pricing without OpenAI lock-in
• Multi-modal - Combine text, voice, images, and documents

Prerequisites¶

🚀 Quick Start Configuration¶

OpenWebUI is configured entirely through environment variables. Below is a step-by-step guide to enable each AI capability.

Step 1: Core Connection — Chat Completions¶

Start by establishing the base connection to stdapi.ai. This configuration is required and enables conversational AI features.

# Core OpenAI API connection
OPENAI_API_BASE_URL=https://YOUR_SERVER_URL/v1
OPENAI_API_KEY=YOUR_STDAPI_KEY

# Default model for background tasks
TASK_MODEL_EXTERNAL=amazon.nova-micro-v1:0

Available Models:

All Amazon Bedrock chat models are automatically available in OpenWebUI once you configure the connection. This includes:

• Anthropic Claude — All Claude model variants (Sonnet, Haiku, Opus)
• Amazon Nova — All Nova family models (Pro, Lite, Micro)
• Meta Llama — Llama 3 models (if enabled in your AWS region)
• Mistral AI — Mistral and Mixtral models (if enabled in your AWS region)
• And more — Any other Bedrock models available in your configured AWS region

Step 2: Voice Input — Speech to Text¶

Enable hands-free interaction by allowing users to speak their queries instead of typing.

# Enable speech-to-text
AUDIO_STT_ENGINE=openai
AUDIO_STT_OPENAI_API_BASE_URL=https://YOUR_SERVER_URL/v1
AUDIO_STT_MODEL=amazon.transcribe

Use Cases:

• Accessibility: Make your AI interface accessible to users with typing difficulties
• Mobile Use: Perfect for on-the-go interactions on smartphones and tablets
• Multitasking: Interact with AI while performing other tasks
• Natural Interaction: Speak naturally instead of formulating written queries

Step 3: Voice Output — Text to Speech¶

Transform text responses into natural-sounding audio for a fully voice-enabled experience.

# Enable text-to-speech
AUDIO_TTS_ENGINE=openai
AUDIO_TTS_OPENAI_API_BASE_URL=https://YOUR_SERVER_URL/v1
AUDIO_TTS_MODEL=amazon.polly-standard

Voice Options:

Use Cases:

• Accessibility: Support visually impaired users with audio output
• Learning: Improve comprehension through audio and text together
• Productivity: Listen to responses while working on other tasks
• Content Creation: Generate audio content directly from AI responses

Step 4: Visual Creativity — Image Generation¶

Bring creative ideas to life by generating images directly from text descriptions within your chat interface.

# Enable image generation
ENABLE_IMAGE_GENERATION=True
IMAGE_GENERATION_ENGINE=openai
IMAGES_OPENAI_API_BASE_URL=https://YOUR_SERVER_URL/v1
IMAGE_GENERATION_MODEL=amazon.nova-canvas-v1:0

Available Image Models:

While the example shows amazon.nova-canvas-v1:0, you can use any image generation model available in Amazon Bedrock:

• Amazon Nova Canvas — amazon.nova-canvas-v1:0 (recommended, fast and high-quality)
• Stability AI — Stable Diffusion models (if enabled in your AWS region)
• Other providers — Any Bedrock image generation model

Simply change the IMAGE_GENERATION_MODEL environment variable to your preferred model ID.

How to Use:

1. Type an image description in the chat
2. Ask the AI to generate an image (e.g., "Create an image of...")
3. The image appears directly in the conversation
4. Download or refine with follow-up prompts

Example Prompts:

• "Generate a photorealistic image of a modern office with plants and natural lighting"
• "Create a minimalist logo for a tech startup focused on sustainability"
• "Design a watercolor illustration of a sunset over mountains"
• "Make a futuristic concept art of a smart city with flying vehicles"

Use Cases:

• Marketing Materials: Generate visuals for campaigns and presentations
• Prototyping: Quick mockups for design concepts
• Content Creation: Illustrations for blog posts and articles
• Education: Visual aids for teaching and learning
• Brainstorming: Rapid visualization of ideas

Step 5: Intelligent Documents — RAG Embeddings¶

Unlock the full power of Retrieval-Augmented Generation (RAG) to chat with your documents using semantic search.

# Enable RAG with embeddings
RAG_EMBEDDING_ENGINE=openai
RAG_OPENAI_API_BASE_URL=https://YOUR_SERVER_URL/v1
RAG_EMBEDDING_MODEL=amazon.titan-embed-text-v2:0

Available Embedding Models:

The example uses amazon.titan-embed-text-v2:0, but you can choose from multiple embedding models available in Amazon Bedrock:

• Amazon Titan Embed Text v2 — amazon.titan-embed-text-v2:0 (recommended, optimized for RAG, 8192 dimensions)
• Amazon Titan Embed Text v1 — amazon.titan-embed-text-v1 (legacy, 1536 dimensions)
• Cohere Embed — Cohere embedding models (if enabled in your AWS region)
• Other providers — Any Bedrock embedding model

Choose your embedding model based on:

• Dimension size — Higher dimensions (like Titan v2's 8192) capture more nuance but use more storage
• Performance — Some models are faster than others
• Language support — Ensure your chosen model supports your documents' languages

How It Works:

1. Upload Documents: Add PDFs, Word docs, text files, or web content
2. Automatic Indexing: OpenWebUI creates vector embeddings using Amazon Titan
3. Semantic Search: Ask questions in natural language
4. Contextual Answers: AI retrieves relevant passages and generates accurate responses

What You Can Do:

• Research: Analyze multiple research papers and extract insights
• Documentation: Search technical documentation using natural language
• Legal: Review contracts and legal documents with AI assistance
• Business Intelligence: Query reports, presentations, and business documents
• Learning: Study textbooks and educational materials interactively

Supported File Types:

• PDF documents
• Microsoft Word (.docx)
• Plain text (.txt)
• Markdown (.md)
• Web pages (via URL)

🔧 Deployment Methods¶

Choose the deployment method that best fits your infrastructure.

Docker Compose (Recommended)¶

The easiest way to deploy OpenWebUI with stdapi.ai using Docker.

version: '3.8'

services:
  openwebui:
    image: ghcr.io/open-webui/open-webui:main
    container_name: openwebui
    ports:
      - "3000:8080"
    environment:
      # Core stdapi.ai connection
      - OPENAI_API_BASE_URL=https://YOUR_SERVER_URL/v1
      - OPENAI_API_KEY=${STDAPI_KEY}
      - TASK_MODEL_EXTERNAL=amazon.nova-micro-v1:0

      # Speech to Text
      - AUDIO_STT_ENGINE=openai
      - AUDIO_STT_OPENAI_API_BASE_URL=https://YOUR_SERVER_URL/v1
      - AUDIO_STT_MODEL=amazon.transcribe

      # Text to Speech
      - AUDIO_TTS_ENGINE=openai
      - AUDIO_TTS_OPENAI_API_BASE_URL=https://YOUR_SERVER_URL/v1
      - AUDIO_TTS_MODEL=amazon.polly-standard

      # Image Generation
      - ENABLE_IMAGE_GENERATION=True
      - IMAGE_GENERATION_ENGINE=openai
      - IMAGES_OPENAI_API_BASE_URL=https://YOUR_SERVER_URL/v1
      - IMAGE_GENERATION_MODEL=amazon.nova-canvas-v1:0

      # RAG Embeddings
      - RAG_EMBEDDING_ENGINE=openai
      - RAG_OPENAI_API_BASE_URL=https://YOUR_SERVER_URL/v1
      - RAG_EMBEDDING_MODEL=amazon.titan-embed-text-v2:0

    volumes:
      - openwebui_data:/app/backend/data
    restart: unless-stopped

volumes:
  openwebui_data:

Deploy:

# Create .env file with your API key
echo "STDAPI_KEY=your_api_key_here" > .env

# Start OpenWebUI
docker-compose up -d

# View logs
docker-compose logs -f

Kubernetes / Helm¶

Deploy OpenWebUI in Kubernetes with a ConfigMap for environment variables.

apiVersion: v1
kind: ConfigMap
metadata:
  name: openwebui-config
data:
  OPENAI_API_BASE_URL: "https://YOUR_SERVER_URL/v1"
  TASK_MODEL_EXTERNAL: "amazon.nova-micro-v1:0"
  AUDIO_STT_ENGINE: "openai"
  AUDIO_STT_OPENAI_API_BASE_URL: "https://YOUR_SERVER_URL/v1"
  AUDIO_STT_MODEL: "amazon.transcribe"
  AUDIO_TTS_ENGINE: "openai"
  AUDIO_TTS_OPENAI_API_BASE_URL: "https://YOUR_SERVER_URL/v1"
  AUDIO_TTS_MODEL: "amazon.polly-standard"
  ENABLE_IMAGE_GENERATION: "True"
  IMAGE_GENERATION_ENGINE: "openai"
  IMAGES_OPENAI_API_BASE_URL: "https://YOUR_SERVER_URL/v1"
  IMAGE_GENERATION_MODEL: "amazon.nova-canvas-v1:0"
  RAG_EMBEDDING_ENGINE: "openai"
  RAG_OPENAI_API_BASE_URL: "https://YOUR_SERVER_URL/v1"
  RAG_EMBEDDING_MODEL: "amazon.titan-embed-text-v2:0"
---
apiVersion: v1
kind: Secret
metadata:
  name: openwebui-secret
type: Opaque
stringData:
  OPENAI_API_KEY: "your_stdapi_key_here"

Environment File (.env)¶

For bare metal or systemd deployments, use an environment file.

# Core connection
OPENAI_API_BASE_URL=https://YOUR_SERVER_URL/v1
OPENAI_API_KEY=your_stdapi_key_here
TASK_MODEL_EXTERNAL=amazon.nova-micro-v1:0

# Speech to Text
AUDIO_STT_ENGINE=openai
AUDIO_STT_OPENAI_API_BASE_URL=https://YOUR_SERVER_URL/v1
AUDIO_STT_MODEL=amazon.transcribe

# Text to Speech
AUDIO_TTS_ENGINE=openai
AUDIO_TTS_OPENAI_API_BASE_URL=https://YOUR_SERVER_URL/v1
AUDIO_TTS_MODEL=amazon.polly-standard

# Image Generation
ENABLE_IMAGE_GENERATION=True
IMAGE_GENERATION_ENGINE=openai
IMAGES_OPENAI_API_BASE_URL=https://YOUR_SERVER_URL/v1
IMAGE_GENERATION_MODEL=amazon.nova-canvas-v1:0

# RAG Embeddings
RAG_EMBEDDING_ENGINE=openai
RAG_OPENAI_API_BASE_URL=https://YOUR_SERVER_URL/v1
RAG_EMBEDDING_MODEL=amazon.titan-embed-text-v2:0

🎯 What You Can Do Now¶

Once configured and running, your OpenWebUI + stdapi.ai integration unlocks powerful AI capabilities:

💬 Intelligent Conversations¶

• Multi-Turn Dialogues: Natural conversations with context awareness across messages
• Model Switching: Change models mid-conversation based on task complexity
• System Prompts: Customize AI behavior with custom system instructions
• Chat History: Full conversation history with search and export

🎤 Voice Interactions¶

• Voice-to-Voice: Speak your question and listen to the response—true hands-free AI
• Language Flexibility: Switch between languages seamlessly
• Mobile Friendly: Perfect experience on smartphones and tablets
• Meeting Mode: Use voice for brainstorming sessions and meetings

🎨 Creative Content¶

• Image Generation: Create visuals directly in your chat with natural language
• Iterative Design: Refine images with follow-up prompts
• Style Control: Specify artistic styles, lighting, composition, and mood
• Download & Share: Export generated images for use in your projects

📚 Document Intelligence¶

• Knowledge Retrieval: Upload documents and get instant answers from their content
• Multi-Document Analysis: Compare and synthesize information across multiple files
• Contextual Citations: See which documents informed each answer
• Continuous Learning: Build a growing knowledge base over time

🔧 Advanced Features¶

• Function Calling: Extend AI capabilities with custom functions and tools
• Code Execution: Run code directly in the chat interface
• Web Search: Enable real-time web search for current information
• Plugins: Extend functionality with OpenWebUI's plugin ecosystem

📊 Model Selection Guide¶

Choose the right model for your use case to optimize performance and cost. These are suggestions only—many more models are available through Amazon Bedrock depending on your AWS region configuration.

Example Recommendations by Use Case:

🔒 Security Best Practices¶

🚀 Next Steps & Resources¶

Getting Started¶

1. Deploy: Use the Docker Compose example above to get running in minutes
2. Configure Models: Add your preferred models in OpenWebUI's settings
3. Test Features: Try chat, voice, images, and document upload
4. Customize: Adjust system prompts and model parameters to your needs
5. Scale: Monitor usage and scale your infrastructure as needed

Learn More¶

Community & Support¶

⚠️ Important Considerations¶