aigen-cli

Zero-API, browser-based AI dataset generation.

Turn any web AI (Gemini, ChatGPT, Claude, Perplexity) into a batch generation engine. No API keys. No per-call costs. Just your browser and a config file.

$20/mo subscription → unlimited generations
vs
$0.001–$0.024 per API call → ~$5 for 200 items

---

Install

Via NPM (no Python setup needed)

npx aigen-cli auth gemini
npx aigen-cli run aigen.yaml

# Or install globally:
npm install -g aigen-cli
aigen auth gemini

The NPM wrapper auto-installs the Python package on first run.

Via pip

pip install aigen-cli

# With semantic dedup + quality scoring:
pip install aigen-cli[ml]

# With Playwright backend:
pip install aigen-cli[playwright]

# Everything:
pip install aigen-cli[all]

---

Quick Start

1. Authenticate once

aigen auth gemini

A Chrome window opens — sign in, then close it. Your session is saved encrypted.

2. Create a config

aigen init

Interactive wizard generates aigen.yaml.

3. Generate

aigen run aigen.yaml

Headless browsers start, data flows into output/dataset.json.

---

Commands

Command	Description
aigen auth <platform>	One-time login. Saves encrypted browser session. Platforms: gemini, chatgpt, claude, perplexity
aigen run <config.yaml>	Primary command. Auth-first headless generation from config
aigen generate	Legacy generation via an already-running Chrome debug session
aigen init	Interactive wizard to create a generation config
aigen doctor	System health check (Chrome, Python, keyring, optional deps)
aigen status	Show generation stats for an output directory
aigen packs	List built-in domain config packs (medical, legal, ecommerce, code)
aigen push-hf	Push a completed dataset to HuggingFace Hub
aigen sessions	List, export, or import browser sessions
aigen schedule	Add/remove/run cron-scheduled generation jobs
aigen mcp	Run as an MCP server for AI agent integration

Run aigen --help or aigen <command> --help for full options.

---

Config File

project: "My Dataset"
description: "Short description"
target: 100              # Total items to generate
batch_size: 5            # Items per browser request
agents: 2                # Parallel browser tabs
platforms:
  - gemini
  - chatgpt

schema:
  fields: [question_text, expected_answer, difficulty, marks]
  required: [question_text, expected_answer, difficulty]

topic_pool:
  - chapter: "Chapter 1"
    topic: "Algebra"
    question_type: short_answer
    difficulty: easy
    marks: 3

output:
  format: json           # json | csv | jsonl
  path: output
  filename: dataset.json

# Optional: push to HuggingFace after generation
huggingface:
  repo_id: "your-username/dataset-name"
  private: false

# Optional: quality + compliance
quality:
  enabled: true          # requires pip install aigen-cli[ml]
  mode: heuristic
  threshold: 6.0

compliance:
  pii_detection: true
  auto_redact: true

---

Domain Packs

Ready-made configs for common use cases:

aigen packs                          # list available packs
aigen generate --pack medical_mcq    # run a built-in pack

Built-in packs: medical_mcq, legal_qa, ecommerce, code_review

---

Architecture

┌─────────────────────────────────────────────────────────┐
│                      aigen-cli                           │
│                                                          │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐  ┌──────────┐ │
│  │ Gemini   │  │ ChatGPT  │  │  Claude  │  │Perplexity│ │
│  │  Tab     │  │   Tab    │  │   Tab    │  │   Tab    │ │
│  └────┬─────┘  └────┬─────┘  └────┬─────┘  └────┬─────┘ │
│       └──────────────┴─────────────┴──────────────┘       │
│                          │                                │
│              ┌───────────▼───────────┐                    │
│              │    BrowserPool        │                    │
│              │  (auth-first agents)  │                    │
│              └───────────┬───────────┘                    │
│                          │                                │
│              ┌───────────▼───────────┐                    │
│              │  GenerationEngine     │                    │
│              └───────────┬───────────┘                    │
│                          │                                │
│    ┌─────────────────────┼─────────────────────┐         │
│    │                     │                     │         │
│ ┌──▼──┐  ┌───────────┐ ┌▼────────┐  ┌────────▼──┐      │
│ │Parse│  │  Validate  │ │ Dedup   │  │  Output   │      │
│ │JSON │  │  (schema)  │ │ Engine  │  │ Writers   │      │
│ └─────┘  └───────────┘ └─────────┘  └───────────┘      │
└─────────────────────────────────────────────────────────┘

---

Why This Exists

API costs scale linearly with usage. Generating 1000 items via API can cost $20–50. The same models via web interface are included in a flat $20/month subscription.

aigen-cli automates the web interface so you get:

• Unlimited generations — rate-limited by platform anti-bot, not your billing
• Zero API keys — just log in once like a normal user
• Full transparency — every raw response is saved for audit
• No vendor lock-in — swap platforms instantly via config

---

License

MIT