A full-stack application that collects relevant information from search APIs (e.g., PubMed), creates latent embedding vector using embedding models, and provides a user-friendly interface for users to search and explore the information, then create readable report utilizing LLMs.

• Extract expert knowledge and store it in a database.
• Retrieve expert knowlege efficiently.
• Generate a reliable summary of the expert knowledge.
• Generate a readable marketing brochure out of it.
• Provide a user-friendly interface for users to search and explore the information.

./services
    ./backend
    ./frontend
./tfconfig
./scripts

make env-ui

make env-api

A full-stack application with frontend and backend services.

• Svelte
• User interface to get input from users and display search results.

• FastAPI + Docker
• Collects information from search APIs (e.g., PubMed).
• Creates index using Llama-index and embedding models.
• RAG (Retrieval Augmented Generation) approach and llms (e.g., Llama2-7B) to generate expert-knowledge include summary.
• Provides a user-friendly interface for users to search and explore the information.
• Creates readable report out of it.

• Azure Static Web Apps
• Azure App Service
• Azure MongoDB (vector database)
• Azure ML studio (Language model and embedding model)
• Azure Container Registry

make run-ui

make run-api

You might need to set the following environment variable to avoid the error message OMP: Error #15: Initializing libiomp5.dylib, but found libiomp5.dylib already initialized.:

export KMP_DUPLICATE_LIB_OK=TRUE