LookML Quality Assistant 🚀

An AI-powered agentic pipeline built with the Google ADK to refine, audit, and improve LookML files. This project uses a deterministic "Project Picker" to lock in context before engaging a Gemini-powered assistant.

---

📂 Project Structure

lookml_llm_quality/
├── agent.py               # The Orchestrator (Pipeline definition)
├── custom_agents/
│   ├── __init__.py        # Makes folder a Python package
│   └── picker.py          # Custom Project Selection logic
├── tools.yaml             # MCP configuration (Templates env vars)
├── .env                   # Local secrets (GIT IGNORED)
├── .env.example           # Template for secrets
├── .gitignore             # Git exclusion rules
├── requirements.txt       # Python dependencies
└── toolbox                # MCP binary (Ensure chmod +x)

---

##🛠 Setup & Installation###1. Environment & DependenciesEnsure you are using Python 3.13+.

# Create and activate virtual environment
python3 -m venv .venv
source .venv/bin/activate

# Install requirements
pip install -r requirements.txt

###2. Configure SecretsCopy the example file and fill in your credentials.

cp .env.example .env

Open .env and add your Looker Client ID and Secret.

###3. Prepare the ToolboxThe MCP binary must be executable.

chmod +x toolbox

---

##🚀 Running the AgentLaunch the terminal-based assistant:

python agent.py

###How it works:1. Selection: The script connects to Looker and lists all available projects. 2. Deterministic Input: You select a project by number in the terminal. 3. AI Handoff: The selected project is passed into the Gemini session state. 4. Chat: The AI wakes up already knowing exactly which project you are working on.

---

##🔒 Security & Git* .env: Never committed. Contains raw credentials.

• tools.yaml: Committed. Uses ${VAR} syntax to pull secrets from the environment safely.
• toolbox: Generally ignored (binary), but kept in the .gitignore to prevent bloat.

---

##🤖 Model ConfigurationThis project is optimized for Gemini 3 Pro, which features advanced reasoning for LookML and BigQuery tool orchestration.

---