Foggy - your personalized coding tutor

Goal

With AI coding assistants and tools in the picture, the way a developer works has changed. This tool accounts for the shift in the way developers work.

1. Syntax: Important to understand the syntax of a language or a framework, but not necessarily memorize it.
2. Language Concepts: Crucial for us to understand all the basic concepts of a language or a framework. Helps write better prompts.
3. Methods and method signatures: Important to know the capabilities supported by the framework, library or language. Method signatures not essential.
4. Design Patterns: Essential to know the common design patterns of a language to help solve a problem in a cleaner, more structured manner.
5. Project Based Learning: The approach is to build projects to learn, even if it's as small as a hello world program.

Features

Plan Mode (Completed ✓)

Generate personalized learning plans based on your goals and prerequisites.

Teach Mode (In Progress 🚧)

Interactive teaching sessions with adaptive content delivery.

Evaluate Mode (In Progress 🚧)

Assess your progress and understanding of concepts.

Installation

# Clone the repository
git clone https://github.com/yourusername/foggy.git
cd foggy

# Install dependencies
poetry install

# Setup environment variables
cp .env-template .env
# Add your GOOGLE_API_KEY, GEMINI_MODEL, and TAVILY_API_KEY to .env

Usage

# Generate a learning plan
poetry run foggy plan

# Interactive teaching (coming soon)
poetry run foggy teach

# Evaluate progress (coming soon)
poetry run foggy evaluate

Plan Flow - User Journey

The plan mode creates an adaptive learning plan through an interactive conversation. Here's how it works:

1. Todo Generation

Foggy analyzes your learning goal and generates an initial set of todos/topics to cover. This breaks down your learning objective into manageable chunks.

2. Web Search for Current Information

The system uses web search tools to gather up-to-date information about the topics, ensuring the learning plan reflects current best practices and resources.

3. Understanding Your Knowledge Level

Through interactive questions, Foggy assesses your current knowledge level and prerequisites to tailor the learning path to your needs.

4. Completing Remaining Todos

Foggy systematically works through the remaining todos, structuring content for each topic in your learning plan.

5. Final Learning Plan

You receive a comprehensive, personalized learning plan saved to a markdown file, complete with concepts, examples, and project ideas.

For example: Example learning plan

LangGraph Architecture

Foggy uses LangGraph for orchestrating the planning workflow. The flow diagram below shows how different nodes interact in the planning flow:

Flow Overview

1. Welcome Message Node: Greets the user and explains the planning process.
2. HumanGoal Node: Captures the user's learning goal.
3. AI: TodoListGenerator: LLM generates initial task list based on the goal. 4..AINode (Planner Agent): Core LLM that orchestrates the planning process.
4. HumanNode: Human in the loop for the whole planning process.
5. ToolNode: Provides access to:
   • Search (web_search)
   • Todo create/read operations
6. Write Plan: Updates and saves the learning plan to a markdown file.

State Management

The graph uses a PlanState model with:

• messages: Annotated conversation history with add_messages
• todo: List of Task objects (name, isFinished)
• finished: Boolean flag indicating plan completion

Project Structure

foggy/
├── foggy/
│   ├── cli/              # Click-based CLI
│   ├── langgraph/        # LangGraph orchestration
│   │   ├── models.py     # Pydantic state models
│   │   ├── tools.py      # LangGraph tools
│   │   └── graph.py      # Graph definition
│   ├── prompts.py        # Prompt templates
│   └── conversation/     # Response templates
├── docs/                 # Documentation and diagrams
├── tests/                # Test suite
└── pyproject.toml        # Poetry dependencies