Building CodeLens: A Groq-Powered AI for Automated Bug Detection and Code Rewriting

I Built an AI That Finds Your Bugs and Rewrites Your Code to Fix Them.

Developer Kamaumbugua-dev created CodeLens, a Groq-powered code review tool that identifies SQL injection, memory leaks, and O(n²) complexity. The system utilizes the llama-3.3-70b model to generate a 0–100 health score and execute full-file rewrites with inline fix comments.

Why This Matters

Manual code review is prohibitively expensive and prone to human error, often missing subtle patterns like unclosed database connections or f-strings wired directly into SQL queries. CodeLens addresses this by providing a stateless, automated alternative that mitigates the risk of O(n²) algorithms detonating in production when input scales from n=10 to n=10,000.

Key Insights

• Line Number Hallucination: To prevent LLMs from referencing non-existent lines, the system prefixes code with explicit indices (e.g., ‘1 | import sqlite3’) before processing.
• JSON Reliability: Machine-parseable output is guaranteed by using regex to strip Markdown code fences and enforcing a ‘Return ONLY valid JSON’ system prompt.
• CORS Specification: Using FastAPI’s CORSMiddleware with allow_origins=[’*’] requires allow_credentials to be False to pass browser preflight checks.
• Deployment Latency: Free-tier hosting on Render introduces 30–50 second cold starts, requiring front-end loading states to prevent user churn.
• UI Synchronization: The interface uses CSS scroll-snap-type and Math.round on scrollTop to sync vulnerability lists with detail cards without external state management.

Working Examples

Implementation to solve LLM line number hallucination by anchoring the model to a concrete reference.

def add_line_numbers(code: str) -> str:
    lines = code.splitlines()
    width = len(str(len(lines)))
    return "\n".join(
        f"{str(i+1).rjust(width)} | {line}"
        for i, line in enumerate(lines)
    )

FastAPI middleware configuration required to prevent preflight failures when using wildcard origins.

app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=False,
    allow_methods=["*"],
    allow_headers=["*"],
)

Practical Applications

• Automated Vulnerability Remediation: Chaining an analysis request with a ‘/fix’ endpoint to rewrite insecure code with parameterized queries. Pitfall: LLMs may truncate objects if token limits are hit, breaking the JSON parser.
• Stateless Code Auditing: Utilizing a React 19 and FastAPI stack for instant feedback without database overhead. Pitfall: Vite environment variables are baked at build time, causing potential mismatches if using old preview deployment URLs.

References:

• https://dev.to/kamaumbuguadev/i-built-an-ai-that-finds-your-bugs-and-rewrites-your-code-to-fix-them-4e21
• codelens-new.vercel.app
• github.com/Kamaumbugua-dev/CODELENS