4emet ADK (foremet)

4emet is a CLI-first Agent Development Kit that turns any SaaS (via OpenAPI) into an AI agent. Run foremet init in your repo, answer a few questions, and get a LangGraph agent that can call your API.

Install

pip install -e .
# or from PyPI (when published): pip install foremet

Quick start

1. Init (in your repo): answer prompts to create adk.config.json and generate the agent.

   foremet init
   

   You will be asked for: OpenAPI spec URL or path, product name, API base URL, auth env var name, LLM provider (openai, anthropic, gemini, or ollama), model name, and optional output directory (default ./agent).

2. Set env: copy .env.example to .env and set your SaaS API key and the LLM key for your chosen provider.

   cp .env.example .env
   # Edit .env: set SAAS_API_KEY and one of OPENAI_API_KEY, ANTHROPIC_API_KEY, GOOGLE_API_KEY (or run Ollama locally; no key needed)
   

3. Run the agent (interactive chat):

   foremet run
   

Commands

• foremet init – Questionnaire; writes adk.config.json, .env.example, and generates agent/ (tools, prompt, run script).
• foremet run – Loads config and runs the agent in an interactive loop. Requires .env with your SaaS auth env var and the LLM key for your provider (e.g. OPENAI_API_KEY, GOOGLE_API_KEY for Gemini; Ollama uses local server, optional OLLAMA_BASE_URL).
• foremet refresh – Re-fetches OpenAPI and regenerates agent/ from current config.

Config

adk.config.json (created by init) includes: openapi_url or openapi_path, product_name, product_description, base_url, auth_type, auth_env_var, output_dir (default ./agent), llm_provider (openai, anthropic, gemini, ollama), model.

Secrets are never stored in config; only the name of the env var (e.g. SAAS_API_KEY) is saved. Set the actual key in .env or your environment.

Generated agent

The agent/ directory (or your custom output_dir) contains:

• tools_generated.py – LangChain tools for each OpenAPI operation.
• prompt_generated.py – System prompt for the agent.
• run.py – Standalone script: python agent/run.py "your message" (run from repo root).

foremet run uses the same runner and is the recommended way to chat.

Testing

pip install -e ".[dev]"
pytest tests/ -v

See the plan (Section 10) for the full test plan before release (unit, integration, E2E, manual QA, security, packaging).

LLM providers

• openai – Set OPENAI_API_KEY in .env.
• anthropic – Set ANTHROPIC_API_KEY in .env.
• gemini – Set GOOGLE_API_KEY in .env.
• ollama – Run Ollama locally; optional OLLAMA_BASE_URL (default http://localhost:11434).

Requirements

• Python 3.11+
• API key for your chosen LLM provider (or local Ollama). Set the matching env var in .env.