AI studying companion, powered by LlamaCloud
Last updated Jan 21, 2026
23
Stars
6
Forks
0
Issues
0
Stars/day
Attention Score
21
Topics
Language breakdown
No language data available.
▸ Files
click to expand
README
Study-Llama
Study-Llama is a demo application for organizing, extracting, and searching study notes using LlamaAgents, LlamaClassify, and LlamaExtract. It features a Go-based web frontend and a Python backend for advanced note processing.
Overview
- Frontend (Go):
- Backend (Python):
How It Works
- User Interaction:
- Frontend-to-Backend Communication:
- Backend Processing:
classifyandextract/workflow.py to classify notes (with LlamaClassify) and extract structured information (with LlamaExtract).
- Database Operations:
- Files:
Uploaded files and their metadata (name, category, owner) are stored.
- Classification Rules:
Custom rules for categorizing notes are stored and retrieved from the database.
- Vector Search:
Extracted summaries and FAQs are indexed for semantic search and retrieval.
Features
- Upload and categorize study notes.
- Extract structured information from notes.
- Search notes with metadata filters.
- User authentication and access control.
- Modern web UI with Go templates.
Project Structure
- frontend/
main.go: Web server entry point.
- handlers/: HTTP request handlers.
- auth/, files/, rules/: Business logic and DB operations.
- static/: Images and assets.
- templates/: HTML templates.
- src/
study_llama/: Python backend modules for classification, extraction, search, and database logic.
Set Up
Clone this repository:
git clone https://github.com/run-llama/study-llama
cd /study-llamaDeploy the LlamaAgent:
In order for the classify-extract and vector search agent workflows to receive and process queries, they needs to be deployed to the cloud (or at least accessible through a public endpoint). The easiest way to do so is to use llamactl and deploy the agent workflow as a LlamaAgent:
uv tool install -U llamactl
llamactl auth # authenticate
llamactl deployments create # create a deployment from the current repositoryIn order for the LlamaAgent to work, you will need the following environment variables in a .env file (llamactl manages environments autonomously):
OPENAIAPIKEYto interact with GPT-4.1 for email generationLLAMACLOUDAPIKEYandLLAMACLOUDPROJECTIDto get predictions from LlamaClassify and LlamaExtractPOSTGRESCONNECTIONSTRINGto connect to the Postgres database with the uploaded files and the classification rules (you can use Neon, Supabase, Prisma or a self-hosted Postgres instance)QDRANTAPIKEYandQDRANTHOST, to upload sumamries and question/answers to perform vector search and retrieval (you can use Qdrant Cloud or a self-hosted Qdrant instance).
Once the agent is deployed, build the Docker image for the frontend (needed to interact with the LlamaAgents we just created), and deploy it through services like Dokploy or Coolify.
docker build . -t your-username/study-llama:prod
docker login ghcr.io # (uncomment if you wish to use the GitHub container registry)
docker push your-username/study-llama:prodThe frontend service uses a few env variables:
LLAMACLOUDAPIKEY,FILESAPIENDPOINT(which will presumably behttps://api.cloud.llamaindex.ai/deployments/study-llama/workflows/classify-and-extract/run) andSEARCHAPI_ENDPOINT(which will presumably behttps://api.cloud.llamaindex.ai/deployments/study-llama/workflows/search/run), the API key and the API endpoints to interact with your deployed LlamaAgentPOSTGRESCONNECTIONSTRINGto connect to the Postgres database with the uploaded files, the classification rules and the user auth (you can use Neon, Supabase, Prisma or a self-hosted Postgres instance, but it has to be the same as for the LlamaAgent)CACHETABLEandRATELIMITING_TABLE, the table names for the SQLite database taking care of caching and rate limiting.
