Use Milvus with SambaNova
SambaNova is an innovative AI technology platform that accelerates the deployment of advanced AI and deep learning capabilities. Designed for enterprise use, it empowers organizations to leverage generative AI for enhanced performance and efficiency. By providing cutting-edge solutions like the SambaNova Suite and DataScale, the platform enables businesses to extract valuable insights from their data, driving operational improvements and fostering new opportunities in the AI landscape.
SambaNova AI Starter Kits are a collection of open-source resources designed to assist developers and enterprises in deploying AI-driven applications with SambaNova. These kits provide practical examples and guides that facilitate the implementation of various AI use cases, making it easier for users to leverage SambaNova’s advanced technology.
This tutorial leverages Milvus integration in SambaNova AI Starter Kits to build an Enterprise Knowledge Retrieval system, similar to RAG(Retrieval-Augmented Generation), for retrieval and answering based on the enterprise private documents.
This tutorial is mainly referred to the SambaNova AI Starter Kits official guide. If you find that this tutorial has outdated parts, you can prioritize following the official guide and create an issue to us.
Prerequisites
We recommend using Python >= 3.10 and < 3.12.
Visit the SambaNova Cloud to get an SambaNova API key.
Clone the repository
$ git clone https://github.com/sambanova/ai-starter-kit.git
$ d ai-starter-kit/enterprise_knowledge_retriever
Change the vector store type
Change the vector store by setting db_type='milvus' in the create_vector_store() and load_vdb() functions in src/document_retrieval.py.
...
vectorstore = self.vectordb.create_vector_store(
..., db_type='milvus'
)
...
vectorstore = self.vectordb.load_vdb(..., db_type='milvus', ...)
Install dependencies
Install the required dependencies by running the following command:
python3 -m venv enterprise_knowledge_env
source enterprise_knowledge_env/bin/activate
pip install -r requirements.txt
Start the application
Use the following command to start the application:
$ streamlit run streamlit/app.py --browser.gatherUsageStats false
After that, you see the user interface in your browser:
http://localhost:8501/
After set your SambaNova API key in the UI, you can play around with the UI and ask questions about your documents.
For further details, please refer to the Enterprise Knowledge Retrieval of SambaNova AI Starter Kits official documentation.