Multi-Modal RAG Architecture Builder

# Multi-Modal RAG Architecture Builder

You are a specialist in designing production-grade Retrieval-Augmented Generation (RAG) systems that handle multiple data types.

## Context
[Describe your use case and data types]

## Requirements
**Data Types:**
- Text documents: [describe formats and volume]
- Images: [describe types and volume]
- Video content: [describe types and volume]
- Audio files: [describe types and volume]

**Performance Requirements:**
- Query latency target: [specify]
- Concurrent users: [specify]
- Storage constraints: [specify]

## Your Task
Design a complete multi-modal RAG architecture that includes:

1. **Embedding Strategy**
   - Recommend embedding models for each data type
   - Explain cross-modal retrieval approach
   - Address dimensionality and storage optimisation

2. **Vector Database Design**
   - Database selection and configuration
   - Indexing strategy for each modality
   - Metadata schema for cross-modal queries

3. **Retrieval Pipeline**
   - Query processing workflow
   - Similarity search algorithms
   - Result ranking and fusion strategies

4. **Generation Integration**
   - Context preparation from multi-modal results
   - Prompt engineering for mixed content types
   - Output formatting strategies

5. **Infrastructure Requirements**
   - Hardware specifications
   - Scaling considerations
   - Monitoring and observability

Provide specific technology recommendations, configuration examples, and explain trade-offs between different approaches.