An interactive Streamlit app that integrates a vector database (FAISS) into an MLOps pipeline for semantic search on textual data. This project demonstrates how modern vector-based techniques can be used within real-world retrieval workflows, complete with monitoring, inference, and model management.
- Manual or file-based document ingestion
- Semantic embedding generation (Sentence Transformers)
- FAISS index creation and vector-based similarity search
- Interactive semantic query interface
- Real-time query monitoring and performance metrics
- Model state export and management interface
| MLOps Phase | Implementation in Code |
|---|---|
| Data Ingestion | Upload via text form or .txt file |
| Feature Engineering | Embedding generation via SentenceTransformer |
| Model Deployment | FAISS indexing with IndexFlatIP |
| Inference | Real-time semantic search |
| Monitoring | Query logs, response times, result count, hashes |
| Model Management | Reset, export, and control pipeline state |
- Clone the repository:
git clone https://github.com/your-username/vector-db-mlops.git
cd vector-db-mlops- Install the dependencies:
pip install -r requirements.txt- Run the application:
streamlit run main.py- Python ≥ 3.9
- faiss-cpu
- sentence-transformers
- streamlit
- plotly
- pandas, numpy
Dashboard view with key metrics and model status
This project is part of a broader study exploring how vector databases can be effectively integrated into MLOps pipelines.