CodingDrop / Drops / silentbug
Out nowsilentbug
Finds the bugs that pass your tests.
silentbug reads Python and JavaScript/TypeScript code and looks for the bugs tests and linters miss: errors that still return success, routes with no login check, and LLM calls that quietly run up the bill. With a model set up, an AI judge confirms or rejects each one before it's reported as a bug.
Secure payment by Razorpay (UPI or card). Digital product: sales are final once you have access to the repo. Refund policy.
Shortened from a real run on the demo app.
A look inside
What silentbug does
A fixed list of checks, every time
It looks for silent errors, missing auth, SQL injection, hardcoded secrets, open CORS, blocking calls in async code, missing timeouts and AI/LLM cost traps, on every run, without being asked.
An AI judge before the report
With a model set up, each suspect is confirmed or rejected by reading the code around it, so your report shows what the model confirmed, not a long list of guesses.
Runs on a local model
Point it at a local Ollama model and your code never leaves your machine. Or plug in a cloud model instead.
An LLM cost map
Every LLM SDK call in your code (OpenAI, Anthropic, Gemini, Ollama, LangChain and more) with its model, token limit and whether it runs in a loop, plus the calls likely to burn the most tokens. Estimates from the code, not measured usage.
Reports you can use
Terminal output, an HTML report, Markdown, JSON, SARIF and a local dashboard.
Reports only
It never changes your code. Each finding comes with a plain-language hint on how to fix it.
Who it's for
- Developers and small teams shipping Python or JavaScript/TypeScript backends and apps
- Teams building AI or LLM features who want to keep an eye on cost
- Anyone who wants their code checked on their own machine, with a local model
What you need
- Python 3.11 or newer
- No Node.js needed, even for JavaScript/TypeScript projects
- Docker works if you'd rather not install Python
- Checks Python and JavaScript/TypeScript projects: FastAPI, Flask, Django, Express, Next.js, NestJS, React
- Optional: a local Ollama model, or a cloud model API key, for the AI judge
Questions
Does it fix the bugs for me?
No. It reports them, each with a plain-language hint on how to fix it. It never changes your code.
Does my code leave my machine?
Not with a local model. With a cloud model, the code around each suspect is sent to the provider you chose, after secrets are masked.
Will it catch every bug?
No tool does. It looks for the kinds listed above, and with a model set up it confirms each one before reporting it.
Which languages does it check?
Python and JavaScript/TypeScript.
Will you help me set it up?
Yes. Every purchase includes a free setup call of about 30 minutes to get it running on your machine. Reply to your purchase email, or email goyalmanish3333@gmail.com with your GitHub username, and I'll find a time with you. Custom changes are a separate, paid job.
How do I get it after paying?
A GitHub invite to the private repo goes to the username you entered at checkout, usually within about a minute. Accept it and follow the README.
silentbug, yours to run and rebrand
- The full source code in a private GitHub repo
- A GitHub invite to the repo in about a minute after you pay
- A free setup call of about 30 minutes to get it running on your machine: reply to your purchase email to book it
- A README that takes you from clone to first run, and a guide to rebrand it
- A commercial licence: use it, change it and put your own name on it. One licence per developer; the source code can't be shared or resold.
Diwali price · ends 8 Nov Full source code · no subscription
Buy silentbugSecure payment by Razorpay (UPI or card). Digital product: sales are final once you have access to the repo. Refund policy.