📚 Step 1: the context (1 hour)
Don’t read the code yet. Read these first to understand why QWED exists.- The story: Read how QWED evolved from a simple prototype to an enterprise architecture.
- The system: Understand the deterministic verification logic and how QWED safeguards the future of AI.
- The technical core: Explore how visionary concepts map to the 8-engine specialized architecture.
- Security framework: Scan the latest enterprise security framework.
🛠️ Step 2: the setup (30 mins)
- Clone the Repo:
- Install Dependencies:
- Environment Variables:
- Create a
.envfile (copy.env.example). - Note: We now use PostgreSQL via Docker. Ensure you have Docker running (
docker-compose up -d) to run the full suite, but for this task, standard unit tests will suffice.
🎯 Step 3: the task (Phase 0)
The goal: building a “SQL firewall” for AI agents
Enterprises want to use “Text-to-SQL” agents (e.g., “Show me top 10 users”). The Risk: LLMs hallucinate. If an LLM generatesDROP TABLE users or SELECT * FROM passwords, the company is destroyed.
QWED’s role: QWED parses the SQL using AST analysis and blocks dangerous queries before they reach the database.
Real-world scenarios (your task)
You are building the Safety Test Suite for our SQL Engine. Your Workspace:tests/test_sql_safety.py (Create this file).
Task: Write 5-10 test cases that simulate these “Bad AI” behaviors:
Output:
Write the Tests first (TDD). Create scenarios for both Safe Queries (e.g.,
SELECT name FROM users) and **Unsafe Queries. Assert that our engine raises a SecurityViolation` error for the unsafe ones.
🚀 Step 4: submission
- Create a branch:
git checkout -b feature/sql-safety-tests - Push your changes.
- Open a Pull Request (PR).
Questions?
Ping me anytime. I value:- Curiosity: Ask “Why did you design it this way?”
- Clarity: Write simple, readable code.
- Speed: Ship small, verified changes.