We are excited to announce our partnership with CambioML today!
CambioML(https://www.cambioml.com/pykoi) removes the hurdles for ML scientists looking to finetune LLMs. Its open-source library, pykoi (https://github.com/CambioML/pykoi), allows you to easily collect labeling demonstration data and user feedback, train RLHF on that data, and compare different models to each other.
Through this partnership, Epsilla is now integrated with pykoi as a vector store vendor and delivers the knowledge retrieval component.
Let’s watch a demo to see how to create a chatbot that can conduct Q&A with your personalized data!
Demo
In this demo, we uploaded the book “The Mom Test”, which discusses how to conduct user interviews. We then posed questions related to the book to the chatbot and provided feedback on the quality of the answers.
Super cool, huh? Below is step-by-step guidance on how to run the demo yourself.
Installation
Step 1. Install Pykoi
- conda create -n test python=3.10 -y conda activate test pip install pykoi
Step 2. Install Epsilla vector DB through docker image
- docker pull epsilla/vectordb docker run --pull=always -d -p 8888:8888 epsilla/vectordb
Try the retrieval-qa demo
Pykoi offers a declarative Python interface that allows developers to compose applications with components such as chatbots, RLHF feedback dashboards, and more. Download the demo code at https://github.com/CambioML/pykoi/blob/main/example/retrieval_qa/retrieval_qa_demo.py
Now, let’s try to run the demo. Before proceeding, remember to insert your OpenAI key here:
- os.environ["OPENAI_API_KEY"] = "Put your OpenAI key here"
Run the demo:
- python retrieval_qa_demo.py -vectordb="epsilla"
Open your browser and access the UI at http://127.0.0.1:5000. Now, you can upload your own files, ask the chatbot personalized questions, and provide feedback regarding the quality of the chatbot’s responses!