Cloud-Ready RAG System

Create a detailed design document to replace ChromaDB with Upstash Vector Database. I have added the Upstash Vector credentials in the .env file.

Key Information about Upstash Vector:
- Built-in embedding model: mixedbread-ai/mxbai-embed-large-v1
- 1024 dimensions, 512 sequence length, MTEB score 64.68
- Automatic text vectorization (no need for external embedding API)
- Serverless and cloud-hosted
- REST API based with authentication tokens
- Cosine similarity for semantic search

Current Implementation Details:
- Using ChromaDB for local vector storage
- Using Ollama's mxbai-embed-large for embeddings
- Python-based RAG system with food data
- Manual embedding generation and upsert process

Requirements for Migration:
1. Replace ChromaDB client with Upstash Vector client
2. Remove manual embedding generation (Upstash handles this automatically)
3. Update upsert process to use raw text data instead of pre-computed embeddings
4. Modify query process to work with Upstash Vector API
5. Handle authentication and error management
6. Maintain the same RAG functionality and user experience

Please provide:
- Architecture comparison (before vs after)
- Detailed implementation plan
- Code structure changes required
- API differences and implications
- Error handling strategies
- Performance considerations
- Cost implications of cloud vs local
- Security considerations for API keys

Create a detailed plan to migrate from local Ollama LLM to Groq Cloud API. I have added the GROQ_API_KEY to the .env file.

Current Implementation:
- Using Ollama locally with llama3.2 model
- Direct HTTP requests to localhost:11434/api/generate
- Streaming disabled for simplicity
- Local inference with no API costs

Target Implementation:
- Groq Cloud API with llama-3.1-8b-instant model
- HTTP requests to Groq's API endpoints
- API key authentication required
- Cloud-based inference with usage-based pricing

Groq API Details:
- Model: "llama-3.1-8b-instant"
- Endpoint: Groq's REST API
- Authentication: Bearer token with API key
- Response format: Similar to OpenAI API
- Rate limits and usage tracking

Code Example Reference:
```python
from groq import Groq
client = Groq()
completion = client.chat.completions.create(
    model="llama-3.1-8b-instant",
    messages=[{"role": "user", "content": ""}],
    temperature=1,
    max_completion_tokens=1024,
    top_p=1,
    stream=True,
    stop=None
)
```

Please provide:
- Detailed migration steps
- Code changes required
- Error handling for API failures
- Rate limiting considerations
- Cost implications and usage monitoring
- Fallback strategies
- Testing approach
- Performance comparison expectations