DEV Community

massiron
massiron

Posted on • Originally published at massiron.com

X-Ray Your Python Codebase in 0.2 Seconds — Without an LLM

#atlas

You inherit a 50k-line Python monolith. Senior dev who wrote it left last month. No tests, no docs, no architecture diagram. Where do you even start?

Grep for imports? Parse the AST yourself? Spend two days drawing boxes and arrows in Mermaid?

There's a faster way. Let me show you.

What is code-atlas-py?

code-atlas-py (CLI: atlas) is a deterministic code intelligence tool. It builds a full call graph, dependency graph, and symbol index of any Python project in ~0.2 seconds. It scores every function for risk — complexity, coupling, instability — and surfaces the most dangerous spots before you touch them.

No LLM. No cloud. No tokens. Same result every run.

Step 1: Install 

pip install code-atlas-py
Enter fullscreen mode Exit fullscreen mode

That's it. No config file, no server, no database. It works on any Python project out of the box.

Step 2: Scan your project 

cd /path/to/your/project
atlas scan
Enter fullscreen mode Exit fullscreen mode

You'll see output like:

✓ Scanned 1,247 symbols in 0.19s
  Functions: 892 | Classes: 203 | Imports: 152
  Top risk: payment/processor.py:process_payment (risk: 8.7)
              - complexity: 14 | coupling: 23 | instability: 0.87
Enter fullscreen mode Exit fullscreen mode

The scan found your riskiest function before you opened a single file.

Step 3: Ask questions in natural language 

atlas ask "which functions call process_payment?"
Enter fullscreen mode Exit fullscreen mode 
process_payment
  ├─ checkout.create_order (checkout/orders.py:142)
  ├─ billing.sync_subscriptions (billing/sync.py:67)
  └─ webhooks.stripe_handler (webhooks/stripe.py:33)

3 callers, 4 callees (including 2 external API calls)
Risk impact if changed: high (3 downstream modules affected)
Enter fullscreen mode Exit fullscreen mode

Step 4: Generate an interactive HTML report 

atlas html --open
Enter fullscreen mode Exit fullscreen mode

This opens a clickable, filterable call graph in your browser. Each node is colored by risk score. Click any function to see its callers, callees, and complexity breakdown. Filter to show only functions with risk > 7 and instantly see the 5% of your codebase you should review before making any changes.

Step 5: Use it in CI/CD 

atlas scan --fail-on-risk 8 --json > atlas-report.json
Enter fullscreen mode Exit fullscreen mode

Fail a PR if it introduces code with risk score ≥ 8. No more merges that silently add coupling. Ship an HTML artifact alongside your test results.

Who is this for?

  • Engineers inheriting legacy code: Find the hot spots in 0.2s instead of 2 days of reading.
  • Tech leads doing code review: See risk before you read line by line — focus your attention where it matters.
  • DevOps teams: Add codebase health to your CI/CD pipeline.
  • Anyone using AI coding agents: code-atlas-py integrates with deepstrain so the AI understands structure, not just text. It makes AI edits safer by showing risk impact before the tool makes a change.

How it compares

Tool What it does Deterministic? Local? Free?
pylint/flake8 Style + lint + simple complexity Yes Yes Yes
sourcegraph Code search with indexing server Yes No Subscription
LLM code review Ask "is this risky?" No Varies Freemium
code-atlas-py Call graph + risk scoring + symbol search Yes Yes Free tier

Limitations (honest ones)

  • Python only (for now).
  • Static analysis — it can't know runtime call paths (e.g., getattr-based dispatch).
  • Risk scores are heuristics — useful for triage, not a replacement for reading the code.
  • The free tier covers the core engine. Advanced modules (7 for Pro, 12 for Enterprise) are paid.

Try it 

pip install code-atlas-py
cd your-project
docker run -v $(pwd):/code --rm code-atlas-py atlas scan
# or just:
atlas scan
Enter fullscreen mode Exit fullscreen mode

Repo: github.com/mete-dotcom/code-atlas
Site: massiron.com/atlas

What's the riskiest function in your codebase right now? Run atlas scan and find out.

Top comments (0)