51 lines
812 B
Markdown
51 lines
812 B
Markdown
|
|
# Vector Database POC
|
|
|
|
## What it is
|
|
|
|
- Development environment using Devcontainers
|
|
|
|
- Stores document embeddings in ChromaDB
|
|
|
|
- Retrieves relevant documents for a given query
|
|
|
|
- Generates answers using Ollama
|
|
|
|
## Getting started
|
|
|
|
Download some models
|
|
|
|
```bash
|
|
ollama pull mxbai-embed-large # Used for embeddings
|
|
|
|
ollama pull gemma3 # Used as LLM
|
|
```
|
|
|
|
Download python depedencies
|
|
|
|
```bash
|
|
pip install -r requirements.txt
|
|
```
|
|
|
|
## Run
|
|
|
|
First you need to seed the database with a few documents
|
|
|
|
```bash
|
|
python script.py seed
|
|
```
|
|
|
|
And then you can do your search
|
|
|
|
```bash
|
|
python script.py search --query "How does AI change industries?"
|
|
```
|
|
|
|
## Roadmap
|
|
|
|
- Streaming responses (for real-time chat)
|
|
|
|
- Metadata filtering (e.g., retrieving only recent docs)
|
|
|
|
- Improved ranking (reranking results for better accuracy)
|