Evidence Gallery: the engine, run in public.
Every report below was generated by OpenThunder’s keyless deterministic engine against the public source of a well-known open-source repository: no AI model, no account, no API key. Each card shows the exact commit analyzed and the date it was generated, and each report is regenerated on a weekly schedule from the latest engine source, so this page is continuous, reproducible proof that the analysis works, not a one-time demo.
Each card shows what the engine found: modules mapped, critical flows traced, cycles and fragile zones detected. Architecture health is one structural signal among them, not a grade: large, mature systems naturally score lower, and that is not a defect.
10 repositories · last generated 2026-07-21 (UTC)
Want accuracy, not just reach? The Accuracy Benchmark runs the same keyless detectors against fixtures with a KNOWN planted issue (or a known-clean control) and publishes expected vs detected, misses and false positives included.
fastify
Fast, low-overhead Node.js web framework
express
The classic Node.js web framework
hono
Small, fast web framework for the edges
axios
Promise-based HTTP client for browser and Node.js
zod
TypeScript-first schema validation
vite
Next-generation frontend build tool
nest
Progressive Node.js framework for server-side apps
prisma
Next-generation TypeScript ORM
langchainjs
Building LLM applications in JS/TS
next.js
The React framework for the web
The engine is deterministic: the same commit produces the same report. Clone the repo, run the same two commands, compare.
git clone --depth 1 https://github.com/<owner>/<repo>
cd <repo>
openthunder lens generate --no-narrate
openthunder lens export --out report.htmlHonesty notes: these are automated static-analysis reports of public source code, generated with no human curation of the findings; they are not security audits, and not an endorsement by or affiliation with any of these projects. Scores are OpenThunder’s opinionated architecture heuristics, useful for comparison and for judging the engine, not a verdict on project quality. A repository that fails to analyze is skipped, never shown broken; the skip list is recorded in manifest.json.