⚡ Poseidev

The AI-Powered CLI Code Editor — Multi-model orchestration, file operations, agent collaboration, and auto-build, all from your terminal.

---

✨ Features

Feature	Description
🤖 Multi-Model AI	4 models: Kimi K2, Qwen 3.5 397B, GLM 5.1, DeepSeek V4 Pro
📁 File Operations	AI creates, edits, deletes, and renames files directly on disk
🤝 Model Collab	All 3 NVIDIA models confer in rounds (Architect → Engineer → Reviewer)
🔧 Auto-Build	6-phase AI pipeline: architecture → schema → backend → frontend → integration → polish
🔍 Code Analysis	Health reports, security scans, complexity analysis, dependency checks
🧪 Test Generation	AI-powered test writing and coverage analysis
🔒 Security Scanning	Static analysis + AI-driven vulnerability detection
💬 REPL Chat	Persistent conversation with context, streaming, and markdown rendering
🎨 Beautiful TUI	Gradients, panels, spinners, syntax highlighting, and themed output

---

🛠️ Install from GitHub (for developers)

npm install poseidev

OR

# 1. Clone the repo
git clone https://github.com/Imdevsup/poseidev.git
cd poseidev/cli

# 2. Install dependencies
npm install

# 3. Build
npm run build

# 4. Link globally (makes 'poseidev' command available everywhere)
npm link

# 5. Set your API key (required — no keys are bundled)
poseidev config set-key nvidia-kimi YOUR_API_KEY

# 6. Start using
poseidev

⚠️ Important: Poseidev ships with no API keys. Every user must provide their own. Get a free key at build.nvidia.com.

---

💻 Usage

Interactive REPL (main mode)

poseidev

Opens an interactive session. Type naturally — the AI creates files, edits code, and answers questions.

  ❯ create a REST API with Express and MongoDB

  ◆ Writing files...

  ✓ src/index.ts       42 lines · 1.2 KB · /project/src/index.ts
  ✓ src/routes/api.ts   28 lines · 890 B  · /project/src/routes/api.ts
  ✓ package.json        25 lines · 856 B  · /project/package.json

  ✓ All 3 file(s) written and verified

Slash Commands

Command	What it does
/model	List and switch between AI models
/key	View and set API keys
/collab <task>	All 3 models collaborate on a task
/build	Start the 6-phase auto-build engine
/read <file>	Read and display a file
/ls [path]	List files in a directory
/create <file>	Create a new empty file
/delete <file>	Delete a file
/rename <old> <new>	Rename or move a file
/search <query>	Search project files
/help	Show all 25+ commands

One-Shot Commands

# Chat without entering the REPL
poseidev chat "explain this error: Cannot read property of undefined"

# Analyze code health
poseidev analyze health

# Security scan
poseidev security scan

# Generate tests
poseidev test generate

# Auto-build a full project
poseidev build

---

🤝 Multi-Model Collaboration

The /collab command makes all 3 NVIDIA models work together in rounds:

  ━━━ Multi-Model Collaboration ━━━
  Models conferring in rounds, building on each other's input

  🧠 Architect     Kimi K2 Instruct
  ⚡ Engineer       Qwen 3.5 397B
  🔍 Reviewer       GLM 5.1

  ───── Round 1 — 🧠 Architect ─────
  [Proposes design and file structure]

  ───── Round 2 — ⚡ Engineer ─────
  [Builds on Architect's plan, adds implementation]

  ───── Round 3 — 🔍 Reviewer ─────
  [Critiques both, finds issues]

  ───── Round 4 — 🧠 Architect ─────
  [Incorporates all feedback into final plan]

  ───── Final — Implementation ─────
  [Engineer writes all files based on agreed plan]

---

⚙️ Configuration

API Keys

Poseidev supports multiple AI providers. Set keys from the terminal or inside the REPL:

# From terminal
poseidev config set-key nvidia-kimi YOUR_KEY
poseidev config set-key nvidia-glm5 YOUR_KEY
poseidev config set-key nvidia-qwen YOUR_KEY
poseidev config set-key deepseek YOUR_KEY

# From inside the REPL
/key nvidia-kimi YOUR_KEY
/key deepseek YOUR_KEY

Models

Model	Provider	Key Required
Kimi K2 Instruct	NVIDIA NIM	nvidia-kimi
Qwen 3.5 397B	NVIDIA NIM	nvidia-qwen
GLM 5.1	NVIDIA NIM	nvidia-glm5
DeepSeek V4 Pro	DeepSeek	deepseek

Switch models anytime:

/model qwen/qwen3.5-397b-a17b

Settings

poseidev config show          # View all settings
poseidev config reset         # Reset to defaults

Or from the REPL:

/tokens max 4096     # Set max tokens
/tokens temp 0.7     # Set temperature
/level expert        # Set experience level
/stream              # Toggle streaming

---

📁 File Operations

The AI can manipulate files directly. All operations are verified — Poseidev checks that files were written correctly.

  ✓ src/index.ts      42 lines · 1.2 KB · C:\project\src\index.ts     ← verified
  ✓ package.json      25 lines · 856 B · C:\project\package.json      ← verified
  ⚠ config.ts         written but size mismatch                        ← flagged

Verification checks:

• ✅ File exists after write
• ✅ Content size matches expected
• ✅ Deletion confirmed (file gone)
• ✅ Rename verified (old gone, new exists)

---

🏗️ Project Structure

poseidev/
├── cli/
│   ├── src/
│   │   ├── index.ts              # Entry point + CLI commands
│   │   ├── repl.ts               # Interactive REPL
│   │   ├── config/
│   │   │   └── configManager.ts  # Persistent config storage
│   │   ├── models/
│   │   │   └── modelManager.ts   # Multi-model + fallback chain
│   │   ├── files/
│   │   │   └── fileOperations.ts # File ops engine + verification
│   │   ├── collab/
│   │   │   └── collabEngine.ts   # Multi-model collaboration
│   │   ├── build/
│   │   │   ├── autoBuildEngine.ts # 6-phase auto-build
│   │   │   └── codeExtractor.ts  # Code extraction from AI
│   │   ├── chat/
│   │   │   ├── conversationStore.ts
│   │   │   └── markdownRenderer.ts
│   │   ├── ui/
│   │   │   └── display.ts        # TUI components
│   │   ├── commands/             # 14 CLI command modules
│   │   ├── prompts/
│   │   │   └── promptEnhancer.ts
│   │   ├── search/
│   │   │   └── searchEngine.ts
│   │   └── onboarding/
│   │       └── wizard.ts         # First-run setup wizard
│   ├── package.json
│   └── tsconfig.json
├── .gitignore
└── README.md

---

🧩 Requirements

• Node.js ≥ 18.0.0
• npm ≥ 8
• At least one API key (NVIDIA NIM keys are free at build.nvidia.com)

---

📝 License

MIT © Poseidev

---

Built with ⚡ by the Poseidev team
_{Multi-model AI orchestration for the terminal}