How anupbaranwal/learn-buddy-ai works
Project type and purpose
• Learn-Buddy-AI is a Python-based, full-stack (CLI) multi-agent generative AI system whose purpose is to automatically produce personalized study materials (summaries, flashcards, quizzes) for any topic. It’s built as a Kaggle intensive capstone project and runs as a command-line application.
File structure
Top-level:
• requirements.txt – dependency list
• src/capstone/config/config.py & topics.json – configuration constants and topic definitions
• src/capstone/agents/ – each file defines one agent in the workflow
– aggregator_agent.py
– controller_agent.py
– flash_card_agent.py
– parallel_agent.py
– quiz_agent.py
– root_agent.py
– summary_agent.py
Patterns:
• Modular “agent” classes under a single package
• A controller agent orchestrates the overall flow
• A parallel agent to spawn summary/flashcard/quiz agents concurrently
• An aggregator agent to merge outputs
README for additional insights
• Workflow: User → Search Agent → Parallel Agents (Summary, Flashcard, Quiz, Weakness) → Aggregator → Final Study Pack
• No persistent storage yet; CLI only
• Future enhancements: database, adaptive pathways
a. Main components:
1. User Interface (CLI Entrypoint)
2. ControllerAgent (Orchestrator)
3. SearchAgent / RootAgent (Topic fetcher)
4. ParallelAgent (Concurrency manager)
5. Worker Agents:
– SummaryAgent
– FlashCardAgent
– QuizAgent
– [WeaknessAgent placeholder]
6. AggregatorAgent (Combines outputs)
7. Config/Data Store: config.py, topics.json
8. External AI Service (OpenAI / Gemini, etc.)
b. Relationships & Data Flow:
• CLI → ControllerAgent: accepts subject/topic input
• ControllerAgent → Root/SearchAgent: fetch raw content
• ControllerAgent → ParallelAgent: spin off worker agents simultaneously
• ParallelAgent → Summary/Flashcard/Quiz Agents: each calls external AI API
• Worker Agents → return generated content → AggregatorAgent
• AggregatorAgent → produces final Study Pack → CLI displays
c. Architectural Patterns & Principles:
• Modular multi-agent pattern: single-responsibility agents
• Orchestrator + worker concurrency (ParallelAgent)
• Config-driven design (topics.json, config.py)
• Future extensibility: plug in new agents (e.g., recommendation_agent)
d. Key Technologies & Frameworks:
• Python 3.x
• OpenAI/Gemini API (via requests or openai client)
• JSON for topic definitions
• Standard concurrency modules (asyncio or threading inside ParallelAgent)
Diagramming Recommendations
1. User → ControllerAgent
2. ControllerAgent → RootAgent → (parallel fork) → Summary/Flashcard/Quiz Agents
3. Worker Agents → AggregatorAgent → CLI output
• Blue = orchestration (ControllerAgent, ParallelAgent)
• Green = worker agents (SummaryAgent, FlashCardAgent, QuizAgent, AggregatorAgent)
• Orange = external dependencies (AI API)
• Gray = config/data