A fast, local AI code assistant for scanning, refactoring, and planning changes across your codebase.
- ✅ Prompt-first UI: type your request immediately on launch
- ✅ Modes: Ask / Edit / Plan (Tab to switch)
- ✅ Local LLM via Ollama
- ✅ Multi-language and all files scanning
- ✅ HTML reports and CLI automation
- Install Ollama
curl -fsSL https://ollama.ai/install.sh | sh - Pull a model
ollama pull qwen2.5-coder:14b-instruct-q4
# Build
cd /path/to/sidekick
make build
# or
./build.sh
# or
go build -o sidekick
# Install system-wide (optional)
sudo ./sidekick installsidekick- Prompt line: type your request immediately
- Mode: press Tab to switch Ask/Edit/Plan
- Menu: use ↑/↓ to select, Enter to open
- Ask: answer questions about the code
- Edit: return diffs for code changes
- Plan: provide a step-by-step plan
# Scan a directory
sidekick scan /path/to/project
# Use a specific model
sidekick scan --model qwen2.5-coder:14b-instruct-q4
# HTML report
sidekick scan --format html --output report.htmlSettings are stored at ~/.sidekick/config.json.
Common settings:
- default model
- Ollama URL
- output format
Sidekick scans all files (excluding hidden directories and sensitive files such as .env, private keys, etc.).
- Ollama not running:
ollama serve - Model missing:
ollama pull <model>
MIT