Copy-paste workflows that survive contact with reality. Less theory. More “paste this and win”. 🏆
Chef’s note: if a recipe asks for a tiny change, keep it tiny. The turkey rule applies in the kitchen too. 🦃
curl -fsSL https://raw.githubusercontent.com/UnicoLab/smlcode/main/scripts/install-remote.sh | bash
omlx start # or configure Ollama — see Providers
mkdir -p /tmp/slm-demo && cd /tmp/slm-demo
printf 'package main\n\nfunc Hello() string { return "hi" }\n' > hello.go
slmcode init
slmcode run -v "Add a short Go doc comment to Hello(). Tiny change only."
cat hello.gocd ~/code/my-app
slmcode init
slmcode blocks apply python # apply Python pipeline + quality pack
slmcode config set permission review
slmcode run -v "Add input validation to the login handler"
slmcode apply # interactive per-file review; --all applies everything
slmcode diff
slmcode commit -m "slmcode: validate login input"slmcode run --agent explorer --dry-run "Where is authentication handled?"
# or
slmcode run --agent docs "Summarize the public HTTP API surface"Maps without mayhem. Excellent for onboarding days and panic Mondays.
slmcode config set think_passes 2
slmcode config set retries 2
slmcode config set max_context_kb 16
slmcode config set parallel 1
slmcode run -v "…"If it wanders: tighten AGENTS.md, pin atomic-coding, force explore once.
SLMCODE_FORCE_EXPLORE=1 slmcode run -v "…"Quickly switch your project's entire pipeline, agents, and quality checks to match the language:
# Switch a Go project to use Go pipeline + go-worker + go-tester + QA gate
slmcode blocks apply go
# Switch a Python project
slmcode blocks apply python
# Browse available packs
slmcode blocks list
# Validate custom blocks
slmcode blocks validateLanguage packs set QA gates, pin skills, and override worker/tester agents automatically. The PipelineEditor in Studio also has a one-click preset selector.
Local worker, sharper reviewer — budget diplomacy at its finest.
slmcode tui
# then:
/agent edit worker provider=ollama model=qwen2.5-coder:14b endpoint=http://127.0.0.1:11434
/agent edit reviewer provider=openai model=gpt-4o-mini endpoint=https://api.openai.com/v1Or edit YAML under .slmcode/agents/.
# during a run: Ctrl+C or /stop
slmcode session list
slmcode session resume run-…
# TUI: /resumeYour checkpoint remembers. Your sandwich does not. Prioritize accordingly.
export SLMCODE_PROVIDER=openrouter
export SLMCODE_ENDPOINT=https://openrouter.ai/api/v1
export SLMCODE_API_KEY=sk-or-…
export SLMCODE_MODEL=anthropic/claude-3.5-sonnet
slmcode doctor
slmcode run -v "Refactor the retry helper and add tests"# terminal A
slmcode studio
# terminal B
cd the-same-project
slmcode watchEdit CONTEXT in Studio; watch the board in the terminal. Feel powerful. Use responsibly.
# AGENTS.md
- Prefer table-driven tests
- No drive-by refactors
- Match existing naming; don't invent parallel worldsslmcode skills new house-style --agents worker,reviewer,corrector
slmcode run --skill house-style "…"Bonus tip: vague skills make vague agents. “Be careful” is not a skill. “Never rewrite unrelated files” is a skill.
☀️ Made with ♥ by UnicoLab