factcircuit
FactCircuit: an auditable claim-and-evidence verification loop with immutable provenance, staged checks, bounded retrieval, and reproducible evaluation.
Instantly run SQL queries on your CSVs — or ask in plain English. No install, no sign-in, right in your browser.
Translation pending
DataHTML · JavaScriptGitHub @spelilloOnlineOwner unconfirmed
Have you tried it? One tap is enough. No sign-in needed.
No updates detected yet. GitHub releases will appear here.
Owners of public repositories under a personal GitHub account can sign in to verify ownership and manage the introduction themselves. No administrator approval is needed.
Sign in, generate a verification file, and commit it to this repository’s default branch within 30 minutes. Files on other branches or forks do not count. Then check the file below to manage the project without administrator approval. Keep the file to retain your access.
If needed, you will go to sign-in. After signing in, return here and press your chosen verification button again.
Run /nomorevibe verify in your project folder to verify ownership and manage the introduction yourself.
Verify in one minuteAsk about other verification options
Query any CSV or JSON file with real SQL, or plain English, entirely in your browser. No database to install, no server to host.
Taken from the repository README
Loading the full README. The saved excerpt is shown below.
Query any CSV or JSON file with real SQL, or in plain English, entirely in your browser. There's no database to install and no server to host. When you upload a file, it becomes a SQLite database running client-side in your browser tab.Read the full README ↗Query any CSV or JSON file with real SQL, or in plain English, entirely in your browser. There's no database to install and no server to host. When you upload a file, it becomes a SQLite database running client-side in your browser tab.
Live app: synth-sql.com
JOINs, GROUP BY, window functions, and subqueries all work against your file.json_extract()..xlsx: the results as an Excel Table, and a native Excel chart linked to those cells, so editing the data updates the chart. The workbook is generated by Python (XlsxWriter) running in the browser through Pyodide, so results never leave your machine.| Route | File | What it is |
|---|---|---|
/ | synth.html | The app: home view (upload and saved workspaces), then the query workspace |
/learn | synth.html | Learn SQL: three interactive courses (/learn?lesson=b01 opens a lesson). See LEARN.md |
/welcome | welcome.html | About page: Lite vs. Normal Mode comparison, how it works, and background on the project |
/about | none | Permanent redirect to /welcome |
/guides | guides.html | Index of the SQL how-to guides (/query-csv-with-sql, /sql-cheat-sheet, and others) |
/terms, /privacy | terms.html, privacy.html | Terms of Service and Privacy Policy |
/?auth=signin or /?auth=signup opens the account modal on load. The /welcome page's Sign in link uses it.
Routes and redirects are defined in vercel.json. The static pages share marketing.css.
The app is plain HTML, CSS, and JavaScript with no bundler or framework. synth.html holds the markup and styles and loads these scripts in order, sharing globals declared in state.js:
| File | Handles |
|---|---|
state.js | Shared global state (the database, loaded tables, current user) |
csv-parser.js, json-parser.js | Parsing uploaded files into rows |
auth.js | Supabase auth and the account modal |
checkout.js | Account tiers and feature access |
grid.js | The virtualized results grid |
charts.js | Charts, the full-screen chart view, SVG download, and Excel export |
chat.js | The AI assistant panel |
app.js | Home view, uploads, workspaces, and everything else |
learn-datasets.js, learn-curriculum.js, learn-grader.js, learn-progress.js, learn.js | The Learn page: lesson tables, the three courses, answer checking, progress, and the page itself (styles in learn.css). See LEARN.md |
Other pieces:
excel_chart.py under Pyodide, which loads only the first time someone exports. The XlsxWriter wheel is vendored in vendor/ so the export doesn't depend on PyPI at runtime.api/chat.js) that proxies to Groq, so the API key never reaches the browser.main on GitHub deploys to Vercel automatically.You'll need Node.js and the Vercel CLI (npm i -g vercel), since the routes under api/ are Vercel serverless functions.
git clone https://github.com/spelillo/synth.git
cd synth
npm install
vercel dev
vercel dev serves synth.html and the api/ functions together, matching production. To work on the front end alone, any static server works (for example python3 -m http.server 8123, then open /synth.html, or /synth.html?view=learn for the Learn page), but the AI assistant, checkout, and the vercel.json routes won't be available.
npm test # Learn: answer grader, all 60 lessons against their data, datasets, progress
npm run test:e2e # Learn: browser test of the whole page (needs Playwright + Chromium)
npm test also runs on every pull request (.github/workflows/test.yml).
The API routes read these from your environment (create a .env.local, which is gitignored):
| Variable | Used for |
|---|---|
SUPABASE_URL | Server-side Supabase client (admin operations) |
SUPABASE_SERVICE_ROLE_KEY | Bypasses RLS for admin routes; never expose this to the client |
GROQ_API_KEY | Powers the AI assistant (api/chat.js) |
CRON_SECRET | Authenticates scheduled Vercel Cron jobs |
The client-side Supabase URL and anon key are hardcoded at the top of auth.js, since there's no bundler to inject them at build time. The anon key is meant to be public. Point both at your own Supabase project for local development.
SQLite (via WebAssembly) · Pyodide + XlsxWriter · Supabase · Groq · Vercel Serverless Functions · vanilla JS/HTML/CSS (no framework, no bundler)
There's no LICENSE file, so standard copyright applies: the code isn't licensed for reuse, modification, or redistribution.