wispy-cli 0.1.0

package/LICENSE

@@ -0,0 +1,17 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   Copyright 2026 Minseo & Poropo
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

package/README.md

@@ -0,0 +1,162 @@
+# 🌿 Wispy CLI
+
+AI workspace assistant with multi-agent orchestration. Like Claude Code or Codex, but with workstream isolation, budget management, and a team of specialized sub-agents.
+
+## Quick Start
+
+```bash
+npm install -g wispy-cli
+export GOOGLE_AI_KEY=your-key   # or ANTHROPIC_API_KEY or OPENAI_API_KEY
+wispy
+```
+
+That's it. No server setup needed — it starts automatically.
+
+## Features
+
+### Chat & Tools
+```bash
+wispy                        # Interactive REPL
+wispy "read my package.json" # One-shot with tool use
+wispy --continue "and fix it"# Continue previous session
+wispy --new "fresh start"    # New session
+```
+
+Wispy can **read/write files, run commands, search the web** — not just chat.
+
+### Workstream Isolation
+```bash
+wispy -w backend "fix the API"     # Backend workstream
+wispy -w frontend "add dark mode"  # Frontend workstream  
+wispy overview                      # See all workstreams at a glance
+wispy search "auth"                 # Search across all workstreams
+```
+
+Each workstream has its own conversation history, context, and budget.
+
+### Multi-Agent Orchestration
+```bash
+# AI automatically delegates to specialized sub-agents:
+# 🔍 explorer — codebase search
+# 📋 planner  — strategy design  
+# 🔨 worker   — implementation
+# 🔎 reviewer  — code review & QA
+```
+
+#### Execution Modes
+- **Pipeline** — Sequential chaining: `explore → plan → implement → review`
+- **Ralph mode** — Persistence: retry until reviewer confirms done (max 5 rounds)
+- **Async agents** — Fire-and-forget background tasks
+
+### 7 Providers (Auto-detected)
+
+| Provider | Env Variable | Free Tier |
+|----------|-------------|-----------|
+| Google AI (Gemini) | `GOOGLE_AI_KEY` | ✅ |
+| Anthropic (Claude) | `ANTHROPIC_API_KEY` | ❌ |
+| OpenAI (GPT-4o) | `OPENAI_API_KEY` | ❌ |
+| OpenRouter | `OPENROUTER_API_KEY` | Some |
+| Groq | `GROQ_API_KEY` | ✅ |
+| DeepSeek | `DEEPSEEK_API_KEY` | ✅ |
+| Ollama (local) | `OLLAMA_HOST` | ✅ |
+
+Also auto-detects macOS Keychain credentials.
+
+### Cost Optimization
+- **Smart model routing** — simple tasks → cheap model, complex → mid, critical → expensive
+- **Per-workstream budgets** — `/budget set 5.00` to cap spending
+- **Context window optimization** — auto-trims old messages to save tokens
+- **Token tracking** — see cost after every response
+
+### Security
+- Sandboxed command execution (macOS Seatbelt)
+- Process hardening (core dump disable, ptrace block)
+- Audit logging (JSON Lines)
+- Rate limiting + CORS restriction
+- API key isolation per workstream
+
+## In-Session Commands
+
+| Command | Description |
+|---------|-------------|
+| `/help` | Show all commands |
+| `/compact` | Compress conversation history |
+| `/memory <type> <text>` | Save to persistent memory |
+| `/budget` | Show spending per workstream |
+| `/budget set <usd>` | Set budget limit |
+| `/overview` | Director view — all workstreams |
+| `/search <keyword>` | Search across workstreams |
+| `/cost` | Show session token usage |
+| `/workstreams` | List all workstreams |
+| `/provider` | Show current provider |
+| `/model [name]` | Show/change model |
+| `/clear` | Reset conversation |
+| `/quit` | Exit |
+
+## Project Context (WISPY.md)
+
+Create a `WISPY.md` in your project root or `~/.wispy/WISPY.md`. It gets loaded into every conversation turn (up to 40K chars), just like Claude Code's `CLAUDE.md`.
+
+```markdown
+# WISPY.md
+## About this project
+- Next.js 15 app with TypeScript
+- Use Tailwind CSS
+- Korean comments preferred
+```
+
+## Workstream Context (work.md)
+
+Each workstream can have its own `work.md` — loaded every turn, like WISPY.md but scoped to that workstream.
+
+```bash
+/work init                    # Create work.md for current workstream
+/work set "deadline: Friday"  # Append context
+/work                         # View current work.md
+```
+
+File locations (searched in order):
+1. `~/.wispy/conversations/<workstream>.work.md`
+2. `.wispy/<workstream>.work.md`
+3. `work.md` (project root fallback)
+
+```markdown
+# backend
+## Goals
+- Migrate to PostgreSQL
+- Add rate limiting
+
+## Context
+- Using Express.js + TypeORM
+- Staging server: staging.example.com
+```
+
+## Persistent Memory
+
+Wispy remembers across sessions via `~/.wispy/memory/`:
+
+```bash
+/memory user "prefers dark mode"
+/memory project "deadline is Friday"
+/memory feedback "don't use var, use const"
+/memory references "API docs: https://..."
+```
+
+## Architecture
+
+```
+wispy (CLI REPL)
+  ↕ Gemini/Claude/OpenAI API (tool calling)
+  ↕ AWOS Server (auto-started, sandboxed tools)
+  ↕ Local Node (macOS file/browser/app actions)
+```
+
+## Requirements
+
+- Node.js 20+
+- One API key (Google AI recommended for free tier)
+- macOS or Linux
+
+## License
+
+Apache-2.0 — Minseo & Poropo 🌿

package/bin/wispy.mjs

@@ -0,0 +1,11 @@
+#!/usr/bin/env node
+
+// Wispy CLI entry point — delegates to the main REPL
+import { fileURLToPath } from "node:url";
+import path from "node:path";
+
+const __dirname = path.dirname(fileURLToPath(import.meta.url));
+const mainScript = path.join(__dirname, "..", "lib", "wispy-repl.mjs");
+
+// Dynamic import of the main module
+await import(mainScript);