🚀 zero-cloud-dev-stack

Aider + Ollama + Open WebUI + CrewAI — for Windows

Ollama Aider Open WebUI Python License Windows

"Everything Claude Code does, but running on your machine. Free, private, and yours."

🇪🇸 Versión en Español · 📋 Troubleshooting · ⚡ Quick Start

📖 About

Hardware tested on

Credits

⚠️ Debunked Myths

🧰 Prerequisites

node --version # Node.js
choco --version # Chocolatey 
ollama --version # Ollama → irm https://ollama.com/install.ps1 | iex
py -0 # Must show Python 3.11 available

⚡ Quick Start

# 1. Download models (one-time)
ollama pull qwen2.5-coder:32b # 19GB — takes ~2 hours
ollama pull qwen2.5-coder:7b # 4.7GB — takes ~30 min
# 2. Coding agent (≈ Claude Code):
cd C:\your-project
C:\aider-env\Scripts\aider --model ollama/qwen2.5-coder:32b --no-auto-commits
# 3. Web interface:
C:\webui-env\Scripts\open-webui serve
# Open: http://localhost:8080
# 4. Multi-agent:
C:\crew-env\Scripts\python.exe C:\crew_local.py
# 🔴 KILL SWITCH — free all RAM:
ollama stop qwen2.5-coder:32b
ollama stop qwen2.5-coder:7b

📦 Installation

Step 1 — Aider (Claude Code alternative)

# Always use py -3.11, never just "python"
py -3.11 -m venv C:\aider-env --clear
# Verify it's 3.11:
C:\aider-env\Scripts\python.exe --version
# Must show: Python 3.11.x
# Install using absolute path (not activate):
C:\aider-env\Scripts\python.exe -m pip install "aider-chat>=0.50"
# Verify:
C:\aider-env\Scripts\aider --version
# aider 0.86.2

Step 2 — Open WebUI

py -3.11 -m venv C:\webui-env
C:\webui-env\Scripts\python.exe -m pip install open-webui
# Start:
C:\webui-env\Scripts\open-webui serve
# Wait for: "Uvicorn running on http://0.0.0.0:8080"
# Then open: http://localhost:8080

Step 3 — CrewAI

py -3.11 -m venv C:\crew-env
C:\crew-env\Scripts\python.exe -m pip install crewai crewai-tools

🤖 CrewAI — Coder + Auditor Script

from crewai import Agent, Task, Crew, LLM
llm_large = LLM(model="ollama/qwen2.5-coder:32b", base_url="http://localhost:11434")
llm_fast = LLM(model="ollama/qwen2.5-coder:7b", base_url="http://localhost:11434")
coder = Agent(
 role='Senior Python Developer',
 goal='Write efficient, clean code without inventing non-existent libraries',
 backstory='Python expert. Only deliver executable code with real imports.',
 llm=llm_large, verbose=True
)
auditor = Agent(
 role='QA and Security Auditor',
 goal='Detect bugs, invented libraries, and security issues',
 backstory='You are skeptical. If code imports something that does not exist, reject it.',
 llm=llm_fast, verbose=True
)
task_code = Task(
 description='Write an async Python function to process a large CSV with pandas.',
 expected_output='Clean Python code in a markdown block with all imports.',
 agent=coder
)
task_audit = Task(
 description='Review the generated code. Verify real imports and no obvious bugs.',
 expected_output='Approved with comments, or list of required corrections.',
 agent=auditor
)
crew = Crew(
 agents=[coder, auditor],
 tasks=[task_code, task_audit],
 process='sequential', # sequential = doesn't saturate RAM
 verbose=True
)
result = crew.kickoff()
print("\n=== FINAL RESULT ===")
print(result)

🔧 Troubleshooting

❌ Cannot import 'setuptools.build_meta'

❌ No compatible version of aider-chat>=0.50 found

❌ activate still uses Python 3.14 after creating env with 3.11

❌ Only one usage of each socket address when running ollama serve

❌ Models don't appear in Open WebUI

❌ ERR_CONNECTION_REFUSED on localhost:8080

❌ Model download interrupted

❌ PC very slow with everything running

⚡ RAM Optimization

# Check what's loaded in RAM:
ollama ps
# Free specific model:
ollama stop qwen2.5-coder:32b
# Limit parallel requests:
$env:OLLAMA_NUM_PARALLEL = "1"

📊 Honest Limitations

When the model doesn't know something

🌐 Optional: Two-Machine Network Setup

# On the server machine:
$env:OLLAMA_HOST = "0.0.0.0:11434"
ollama serve
# On the client machine, change all URLs from localhost to the server IP:
C:\aider-env\Scripts\aider --model ollama/qwen2.5-coder:32b --openai-api-base http://192.168.1.100:11434

🗑️ Uninstall

ollama rm qwen2.5-coder:32b
ollama rm qwen2.5-coder:7b
Remove-Item -Recurse -Force C:\aider-env
Remove-Item -Recurse -Force C:\webui-env
Remove-Item -Recurse -Force C:\crew-env
# Optional: free ~25GB of model cache
Remove-Item -Recurse -Force "$env:USERPROFILE\.ollama"
Remove-Item -Recurse -Force "$env:USERPROFILE\.cache\huggingface"

📋 Stack Summary

Guide generated in June 2026 — Installation completed at 4:35am successfully 🌙

Hardware: Acer Aspire AI · Intel Core Ultra 9 · Arc GPU · 32GB RAM · Windows 11

⭐ If this saved you time, consider starring the repo!