Ask your codebase anything, in plain English

Why this exists

You join a new project. The README is six months stale. The Slack thread that explained the auth flow is buried. The person who wrote the billing module left.

So you grep. You open file after file. You trace imports manually. You build a mental model of the codebase one bread crumb at a time.

This takes hours. Sometimes days.

The irony is that all the answers are already in the code. The function names, the import chains, the route definitions, the commit messages — it's all there. You just can't query it in the way your brain actually thinks.

You think: "where does auth start?"

Your tools want: rg -n "auth" --type py | head -20 grep -r "login" src/ | grep -v test | grep -v node_modules git log --all --oneline -- '**/auth*'

Dino closes that gap. You ask the messy English question. It indexes your repo into SQLite with FTS5, parses your intent, runs typed operators, and returns evidence-grounded answers with file paths, line numbers, and confidence scores.

No model. No API key. No cloud. Your code never leaves your machine.

How it works

Dino is a Python CLI. It indexes your codebase into a local SQLite database with FTS5 full-text search, then answers natural language questions using rule-based NLU — no neural networks, no embeddings, no API calls.

The pipeline: normalize → spell correct → extract entities → score intents → build operator plan → execute → rank evidence.