This repo will have the code for agent based hackathon using langchain tools. This agent will enable Natural Language Question Answer on structured and unstructured data. We will use COVID-19 Clinical Trials dataset from Kaggle
The intent of this repo is to provide starter code for developers in a Hackathon to write agent applications using Mosaic AI Agent Framework on Databricks Intelligent Platform. We will use LangChain library and notebooks.
The focus is to develop business logic and proof of concept instead of deployment patterns or scalability.
To create an agentic application that is consistent with host business objectives that leverages the following emerging technologies
- Vector Index Retriever ()
- AI/BI Genie Api ()
- Compound AI System using LangChain
- Tools
- AgentExecutor
- Mosaic AI Capabilities
- Model Serving
- Pay Per Token Foundation Models (List the models)
- Model Serving
- AI Gateway
- Required:
- Must include one retriever
- Must have Guardrails
- Must include five example queries with results
- Bonus
- Deployed endpoint get bonus points
- MLflow Evaluation based on examples gets bonus points
- Use of Genie Tool gets extra points
- Deploy the agent as a Lakehouse App will get bonus points
- Innovation and Problem-Solving
- Usability and User Experience
- Scalability and Technical Feasibility