🧪 ISO-QA-Lab: Local AI Testing Platform

Project Status: ✅ Completed and Validated (Version 1.0.0) Goal: To build a fully functional, self-contained, and secure Multi-Agent AI testing lab capable of migrating manual QA processes to an automated, local AI-assisted enterprise workflow.

🚨 SECURITY & OPERATIONAL WARNING: This platform is designed to run 100% locally using Ollama-based LLMs. It has zero external dependencies on cloud services (Google Cloud, Gemini, Vertex AI, etc.). Never attempt to run this code on a public cloud endpoint; it must run within a controlled, private environment.

---

🏛️ Architecture Overview (The Local Stack)

The system operates on a unified, local architecture governed by the agents-cli tool and the core qwen3.5:9b model.

• LLM Engine: Ollama (qwen3.5:9b).
• Orchestration: The agents-cli workflow.
• Knowledge Base: Local file system for all test artifacts, logs, and metrics.

The system relies on seven localized, self-contained skills, which all point to local APIs and services.

---

🧱 Core Components (The Local Skill Suite)

The core logic is contained within the local-agents-adk-* skills. Each skill provides a fully localized, self-contained module for a specific QA task:

1. /local-agents-adk-scaffold: Project scaffolding and structural setup.
2. /local-agents-adk-code: ADK Python API reference for writing agent logic.
3. /local-agents-adk-eval: The evaluation methodology (Eval-Fix Loop) for ensuring code quality.
4. /local-agents-adk-chaos: Stress-tests the agent by simulating network loss, memory exhaustion, and other failure modes.
5. /local-agents-adk-observability: Tools for monitoring local agent traces, logs, and performance metrics.
6. /local-agents-adk-deploy: Workflow for containerizing and deploying the agent to a local service endpoint.
7. /local-agents-adk-publish: Registers the agent with the local "Local Enterprise" directory, making it discoverable by other local agents.

⚙️ The Operational Workflow (How to Run It)

This process must be run sequentially for a new agent to achieve final maturity.

1. Scaffold: Create the project structure. agents-cli scaffold create [project-name]
2. Code: Develop and implement tools and core logic (/local-agents-adk-code).
3. Test (Success): Run basic evaluation to ensure happy path functionality. agents-cli eval run
4. Fix: Iterate the code based on evaluation results (Eval-Fix Loop).
5. Test (Failure): Run chaos tests to prove resilience. agents-cli chaos run
6. Deploy: Make the agent accessible via a local endpoint. agents-cli deploy
7. Publish: Register the working agent in the local directory. agents-cli publish local-enterprise

---

🚀 Project Success Criteria

• Efficiency: The automated flow demonstrably reduces manual testing time by over 50%.
• Coverage: The system can systematically test for known gaps using the Chaos skill.
• Quality: All artifacts are validated through the local Eval-Fix loop.
• Security: The workflow mandates local authentication before accessing test metrics.

---

This README serves as the official Project Definition for the Local QA Lab. All runs must be executed using the local development environment.