DeepCode - REST API Agent Setup Guide

What is DeepCode?

DeepCode is an AI coding agent that uses Ollama (local AI models) to help you build REST APIs, apps, and more. It doesn't work as a traditional REST API - instead, it communicates with Ollama running on your machine.

---

Prerequisites

1. Install Ollama

• Download: https://ollama.ai
• Windows: Run installer
• Verify: ollama --version

2. Install Python Dependencies

cd "c:\Users\Liver\OneDrive\Desktop\DeepCode"
pip install requests rich watchdog

3. Start Ollama Server

ollama serve

Leave this terminal window open. You should see:

Listening on 127.0.0.1:11434

4. Pull AI Models (in another terminal)

# Fast model (for quick iterations)
ollama pull qwen2.5-coder

# Powerful model (for complex tasks)
ollama pull deepseek-coder-v2:16b

---

Running DeepCode

# Navigate to project
cd "c:\Users\Liver\OneDrive\Desktop\DeepCode"

# Run the agent
python deepcode.py

---

How It Works

┌─────────────────────────────────────────┐
│  Your Task (e.g., "Build REST API")     │
└────────────┬────────────────────────────┘
             │
             ▼
┌─────────────────────────────────────────┐
│  DeepCode Agent (this script)            │
│  - Analyzes your task                    │
│  - Selects appropriate model             │
│  - Breaks down into actions              │
└────────────┬────────────────────────────┘
             │
             ▼
┌─────────────────────────────────────────┐
│  Ollama Server (localhost:11434)         │
│  - Qwen2 model (fast)                    │
│  - DeepSeek model (powerful)             │
└────────────┬────────────────────────────┘
             │
             ▼
┌─────────────────────────────────────────┐
│  Generated Code & Changes                │
└─────────────────────────────────────────┘

---

Example Commands

Build a Simple REST API

You > build a rest api with flask for a todo app

This will:

1. Detect it's a simple task → Use Qwen2 (fast)
2. Read existing files
3. Generate code
4. Write files to disk
5. Run commands as needed

Optimize Code

You > optimize the database queries and add caching

This will:

1. Detect it's complex → Use DeepSeek (powerful)
2. Analyze your code
3. Suggest optimizations
4. Implement improvements

---

API Endpoints (Ollama Internal)

These are what DeepCode uses internally:

Endpoint	Purpose
GET /api/tags	List available models
POST /api/generate	Generate text (simple)
POST /api/chat	Chat interface (advanced)

Note: These are NOT exposed publicly - only accessible locally at http://localhost:11434

---

Fixing 404 Errors

Error: "404 - Ollama API endpoint not found"

Cause: Ollama server not running or not accessible

Fix:

1. Open new terminal
2. Run: ollama serve
3. Wait for message: Listening on 127.0.0.1:11434
4. Go back to DeepCode terminal and try again

Error: "Model not found on Ollama"

Cause: Model not installed

Fix:

ollama pull qwen2.5-coder
ollama pull deepseek-coder-v2:16b

---

Smart Model Selection

DeepCode automatically chooses the best model:

Uses Qwen2 (⚡ Fast) for:

• "Create a simple REST API"
• "Add error handling"
• "Fix typo in main.py"
• "Update the homepage"

Uses DeepSeek (🚀 Powerful) for:

• "Optimize database performance"
• "Refactor architecture"
• "Add machine learning model"
• "Implement caching system"

Force a specific model:

You > use qwen
You > use deepseek

---

Features

Multi-Agent Collaboration

Enable to have 5 agents work together:

You > multiagent on

Agents:

• Architect: Plans structure and design
• Coder: Writes implementation
• Reviewer: Reviews for bugs
• Tester: Writes and runs tests
• Optimizer: Optimizes performance

Code Evolution Tracking

See how your code changes:

You > evolution

Smart Context Analysis

You > context

Shows:

• Tech stack detected
• Dependencies found
• File structure
• Relevant files

---

Troubleshooting

My code works but getting 404 errors

This usually means:

1. Ollama server crashed or stopped
2. Connection lost between DeepCode and Ollama
3. Model was unloaded from memory

Quick fix:

# Terminal 1: Restart Ollama
Stop-Process -Name ollama -Force
ollama serve

# Terminal 2: Run DeepCode again
python deepcode.py

DeepCode hangs or times out

Solutions:

1. Check if Ollama is responding: curl http://localhost:11434/api/tags
2. Restart Ollama
3. Check available RAM (models need 4GB+)
4. Close other applications

Model keeps failing

Solutions:

# Remove and reinstall model
ollama rm qwen2.5-coder
ollama pull qwen2.5-coder

# Or restart Ollama
ollama serve

---

System Requirements

Resource	Minimum	Recommended
RAM	4 GB	8+ GB
Disk	5 GB	10+ GB
CPU	Any	Multi-core
GPU	Not required	NVIDIA/AMD

---

Next Steps

1. Start Ollama: ollama serve
2. Run DeepCode: python deepcode.py
3. Try a task: build a simple rest api
4. Check results: Look at generated files
5. View logs: Use troubleshooting guide if issues occur

---

Getting Help

• See TROUBLESHOOTING.md for detailed fixes
• Type help in DeepCode for commands
• Check Ollama docs: https://ollama.ai