AI Assistant Task Manager (Backend)

A FastAPI-powered backend with LangChain AI agents for intelligent task management.
Built with LangGraph, Groq LLM, and streaming responses via Server-Sent Events (SSE).

Frontend Repository: AI-Assistant-Task-Manager-Frontend

---

🚀 Features

• 🤖 AI-powered task management using LangChain agents
• 🛠️ 5 core tools: Add, List, Complete, Edit, Delete tasks
• ⚡ Real-time streaming responses via SSE
• 🧵 Thread-based memory for conversation context
• 🎯 Multi-agent architecture with tool orchestration
• 📝 Natural language processing - users interact conversationally
• 🔄 FastAPI for high-performance async endpoints

---

📦 Installation

Prerequisites

• Python 3.9+
• uv package manager (Install uv)
• Groq API Key (Get one here)

Install uv

# macOS and Linux
curl -LsSf https://astral.sh/uv/install.sh | sh

# Windows
powershell -c "irm https://astral.sh/uv/install.ps1 | iex"

Setup

# Clone repository
git clone https://github.com/babs12316/AI-Assistant-Task-Manager-Backend.git
cd AI-Assistant-Task-Manager-Backend

# Install dependencies
uv sync

# Create .env file
echo "GROQ_API_KEY=your_groq_api_key_here" > .env

Get Your Groq API Key

1. Visit Groq Console
2. Sign up or log in
3. Generate a new API key
4. Add it to your .env file

---

🎯 Usage

Start the Server

uv run uvicorn src.api:app --reload

Server will run at: http://localhost:8000

Health Check

curl http://localhost:8000/health

Response:

{"status": "healthy"}

---

📡 API Endpoints

POST /chat

Stream AI agent responses using Server-Sent Events.

Request:

curl -X POST http://localhost:8000/chat \
  -H "Content-Type: application/json" \
  -d '{
    "message": "Add gym at 6pm today",
    "thread_id": "user-123"
  }'

Response (SSE Stream):

data: ✅ Added **Gym** [9a8cca01] for today at **6 PM**.  
data: 
data: Anything else you'd like to add?  
data: [DONE]

Request Body:

{
  "message": "string",    // User's natural language input (required)
  "thread_id": "string"   // Unique conversation thread ID (required)
}

Thread ID:

• Use the same thread_id to maintain conversation context
• Each user should have a unique thread ID
• Example: "user-123", "session-abc-456", etc.

---

🛠️ Available Tools

The AI agent has access to these tools:

1. add_task

Add a new task to the list.

Example:

• User: "Add gym at 6pm today"
• Agent: add_task("Gym", "today", "18:00")
• Response: "✅ Added 'Gym' [abc123] for today at 6 PM"

2. list_tasks

List tasks for a specific day or all tasks.

Example:

• User: "Show me today's tasks"
• Agent: list_tasks("today")
• Response: Task list with IDs

3. complete_task

Mark a task as completed.

Example:

• User: "Complete gym"
• Agent: Finds task ID → complete_task("abc123")
• Response: "🎉 Awesome! Marked 'Gym' as complete!"

4. edit_task

Edit an existing task.

Example:

• User: "Change gym to 7pm"
• Agent: edit_task("abc123", updated_due_time="19:00")
• Response: "✅ Updated 'Gym': time → 19:00"

5. delete_task

Delete a task.

Example:

• User: "Delete gym"
• Agent: delete_task("abc123")
• Response: "🗑️ Deleted: Gym"

---

🧠 How the AI Agent Works

Agent Workflow

User: "Add gym at 6pm and show my tasks"
    ↓
Agent analyzes intent
    ↓
Step 1: Calls add_task("Gym", "today", "18:00")
    ↓
Step 2: Calls list_tasks("today")
    ↓
Streams response back to user via SSE
    ↓
User sees: "✅ Added Gym. 📋 Tasks: [abc123] Gym @ 6 PM"

Multi-Step Reasoning Example

User: "Complete gym"
  ↓
Agent: "I need to find the gym task ID first"
  ↓
Agent: Calls list_tasks() → Finds [abc123] Gym
  ↓
Agent: Calls complete_task("abc123")
  ↓
Agent: "✅ Nice work! 'Gym' is done!"

---

📝 Example Interactions

User Input	Agent Actions
Add gym at 6pm	add_task("Gym", "today", "18:00")
Show my tasks	list_tasks()
What's on my plate today?	list_tasks("today")
Complete gym	list_tasks() → complete_task("abc123")
Change lunch to 2pm	list_tasks() → edit_task(...)
Delete the gym task	list_tasks() → delete_task("abc123")
Add gym and call mom, then list	Multiple tool calls in sequence

---

📋 Response Examples

Adding a Task

curl -X POST http://localhost:8000/chat \
  -H "Content-Type: application/json" \
  -d '{"message": "Add gym at 6pm", "thread_id": "user-123"}'

Response:

data: ✅ Added **Gym** [9a8cca01] for today at **6 PM**.
data: 
data: Anything else you'd like to add?
data: [DONE]

Listing Tasks

curl -X POST http://localhost:8000/chat \
  -H "Content-Type: application/json" \
  -d '{"message": "Show my tasks", "thread_id": "user-123"}'

Response:

data: 📋 Tasks (2):
data: 
data: [9a8cca01] ○ Gym @ 6 PM
data: [f3b2c456] ○ Call Mom @ 3 PM
data: 
data: You have 2 tasks pending.
data: [DONE]

Completing a Task

curl -X POST http://localhost:8000/chat \
  -H "Content-Type: application/json" \
  -d '{"message": "Complete gym", "thread_id": "user-123"}'

Response:

data: 🎉 Awesome! Marked 'Gym' as complete!
data: [DONE]

---

🏗️ Project Structure

AI-Assistant-Task-Manager-Backend/
├── src/
│   ├── agent.py          # LangChain agent + tools
│   ├── api.py            # FastAPI routes
│   └── api/
│       └── health.py     # Health check endpoint
├── tests/
│   └── test_agent.py     # Unit tests
├── .env                  # Environment variables (create this)
├── pyproject.toml        # Project dependencies
├── uv.lock              # Locked dependencies
└── README.md             # This file

---

📚 Tech Stack

• FastAPI - Modern, fast web framework
• LangChain - AI agent orchestration framework
• LangGraph - Agent workflow management
• Groq - Ultra-fast LLM inference (model: openai/gpt-oss-120b)
• Pydantic - Data validation
• Server-Sent Events (SSE) - Real-time streaming
• uvicorn - ASGI server
• uv - Fast Python package manager

---

🔧 Configuration

CORS Settings

Edit src/api.py to add allowed origins:

app.add_middleware(
    CORSMiddleware,
    allow_origins=[
        "http://localhost:5173",           # Vite dev server
        "https://tasky22.vercel.app",      # Production frontend
        "https://your-domain.com"          # Your domain
    ],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

Change LLM Model

Edit src/agent.py:

# Current model
llm = ChatGroq(model="openai/gpt-oss-120b")

# Other Groq models
llm = ChatGroq(model="llama-3.1-70b-versatile")
llm = ChatGroq(model="mixtral-8x7b-32768")

---

📖 API Documentation

Once the server is running, visit:

• Swagger UI: http://localhost:8000/docs
• ReDoc: http://localhost:8000/redoc

---

🔗 Related Repositories

• Frontend Repository: AI-Assistant-Task-Manager-Frontend

---

⭐ If you like this project, give it a star on GitHub!