Pikit

Token-efficient coding harness for pi-mono. Multi-model workflows, RTK compression, progressive skill loading, and model-tier routing.

Quick Start

./scripts/install.sh                # Install bun, pi-mono, RTK, deps
./scripts/setup.sh /path/to/project # Link pikit to your project
cd /path/to/project && pi           # Start coding

Model Profiles

/config apply anthropic    # smart=Opus, general=Sonnet, fast=Haiku
/config apply codex        # smart=GPT-5.4, general=GPT-5.4-mini, fast=GPT-5.4-nano
/config apply gemini       # smart=Gemini-3.1-Pro, general=Flash, fast=Flash-Lite
/config apply deepseek     # smart=DeepSeek-R1, general=V3, fast=V3
/config apply qwen         # smart=Qwen3.5-235B, general=Qwen3.5-32B, fast=Qwen3.5-8B
/config apply local        # Ollama: qwen3.6-36b + gemma4-26b
/config apply lmstudio     # LM Studio: same models, OpenAI-compatible API
/config apply vllm         # vLLM: high-throughput serving
/config apply llamacpp     # llama-swap: qwen3.6-36b (reasoning) + gemma4-26b (general)
/config aliases            # Show current mappings

Workflows

/workflow feature "add user authentication with JWT"
/workflow bugfix "TypeError in handleAuth when session is null"
/workflow refactor "simplify the payment processing module"
/workflow research "compare state management approaches for React"
/workflow tri-review "review the latest PR"
/workflow tri-dispatch "compare caching strategies"
/workflow tri-debug "TypeError in handleAuth after upgrading passport"
/workflow tri-security "audit src/api/ for vulnerabilities"
/workflow autoloop "Goal: increase coverage to 90% Metric: npx vitest --coverage Guard: npm run build"
/workflow predict "security and reliability gaps in src/api/"
/workflow reason "Should we use event sourcing for order management? Domain: software"
/workflow self-improve "--target src/ --metric 'npm test' --objective 'fix failing tests'"
/workflow self-test "npm test"
/workflow self-lint "npx eslint src/"
/workflow self-perf "node bench/run.js --target 200ms"

Workflow	Steps	Models
feature	Brainstorm → Plan → Implement (loop) → Test → Lint → Review (parallel) → Report	all tiers
bugfix	Reproduce → Diagnose → Fix → Regression test → Verify → Summary	fast + smart
refactor	Analyze → Plan → Refactor (loop) → Verify → Before/after comparison	fast + smart
research	Clarify → Search → Analyze → Synthesize	smart + general
tri-review	Gather → 3-tier review (parallel) → Consolidate	all tiers
tri-dispatch	3 models in parallel → Compare	all tiers
tri-debug	3 models diagnose in parallel → Compare theories	all tiers
tri-security	3 parallel audits (injection, auth, config) → Report	all tiers
predict	Recon → 3 analysts (parallel) → Devil's advocate debate → Rank → Report	all tiers
reason	Generate A → Critique → Generate B → Synthesize → 3 blind judges (parallel) → Converge (loop) → Report	all tiers
autoloop	Baseline → Audit (git history) → Fix → Commit → Measure → Guard → Keep/Revert (loop) → Report	fast + smart
self-improve	Baseline → Improve (loop) → Report	fast + smart
self-test	Run tests → Fix failures (loop) → Summary	fast + smart
self-lint	Run linter → Fix violations (loop) → Summary	fast + smart
self-perf	Profile → Optimize (loop) → Report improvement	fast + smart

Write your own workflows with the creating-workflows skill — see /workflow list for available workflows.

Token Efficiency

1. RTK — Compresses shell output 60-90% before entering context
2. Model routing — Cheap models for simple steps, expensive for complex
3. Token budgets — Set a per-workflow ceiling; auto-downgrade to fast models when approaching limit
4. Parallel steps — Run independent steps concurrently (tri-review, tri-dispatch)
5. Module pruning — /module hide search removes unused tools from the prompt
6. Progressive skills — Only name+description loaded; full content on demand
7. Dry run — /workflow --dry-run feature "..." previews steps without executing
8. Per-step stats — /rtk --steps shows compression breakdown by workflow step

Skills

Built-in skills for code quality and workflow authoring:

Skill	Purpose
brainstorming	Collaborative ideation techniques
planning	Implementation plan authoring
executing	Step-by-step plan execution
clean-code	Naming, functions, stepdown rule
dry	Don't Repeat Yourself + Rule of Three
yagni	Build only what's needed now
writing-workflows	Workflow YAML authoring guide
writing-tests	Test writing best practices
skill-authoring	How to write new skills

Tools

Tool	What it does
memory_set/get/list/delete	Per-project key-value stores
todo_add/complete/list	Task tracking within sessions
web_search	Search via DuckDuckGo
web_read	Fetch URL as readable markdown
ask_user	Pause and ask the user a question

Commands

Command	Usage
/workflow <name> <input>	Run a multi-step workflow
/config apply <profile>	Switch model profile
/module show/hide <name>	Toggle tool groups
/todo add/done/list/clear	Manage session todos
/health	Check local model endpoint availability
/rtk	Show RTK compression stats
/test-gen <target>	Generate tests for a file or directory
/changelog [since]	Generate changelog from git history