NexusFlow: AI Automation & Data Monitoring Platform
Mojtaba Karimiabdar
How it works: data from four kinds of sources is validated, cleaned, stored, checked for changes, explained by AI and turned into alerts.
The challenge
Teams lose hours checking competitor prices, partner feeds and spreadsheets for changes, and the important ones still slip through.
The solution
NexusFlow is a platform I built to do that work automatically. It collects business data from websites, REST APIs, signed webhooks and CSV or Excel uploads, keeps every version, spots the changes that matter, explains them with AI and alerts the right people.
The cover shows the real web console with a month of demo data: changes per day with unusual days marked, open alerts and the latest changes.
How it works
Data moves through six steps, and every step is authorized, isolated per organization, safe to retry and written to an audit trail:
Validate: data is checked against a typed schema, and bad data is refused.
Clean: values are normalized, de-duplicated and enriched.
Store: every record keeps its full version history.
Detect: real changes and unusual days are found automatically.
Explain: an AI summary says what changed and why it matters. It is opt-in, and personal data is redacted first.
Alert and report: rules notify the right people, and reports export as JSON, CSV, Excel or PDF.
n8n orchestrates the work through five workflows, Python executes it in FastAPI and RabbitMQ workers, and people decide in a web console with passkey, two-factor or single sign-on.
A real dataset page: every record keeps its full version history, sensitive fields are masked and exports take one click.
Every record, every version
Nothing is ever silently overwritten:
Current records at a glance: price, stock and link for each product, with a version number.
Full history: every change is stored as a new version, so you can see what changed and when.
Sensitive fields protected: encrypted at rest and masked unless your role may read them.
One-click export: CSV and JSONL exports, plus scheduled Excel and PDF reports.
Security built in, from database-level isolation and passkeys to a safe AI design, checked by 2,561 automated tests on every change.
Security and quality
Tenant isolation in the database: PostgreSQL row-level security on every table.
Modern sign-in: passkeys, two-factor codes and single sign-on (OIDC, SCIM), with five roles and 38 permissions.
Encrypted everywhere: sensitive values encrypted at rest (AES-256-GCM) and TLS on every connection.
Hostile content contained: web pages and uploads are parsed in a sandbox with no database access and no secrets.
Safe AI by design: offline by default; external AI only when opted in, with personal data redacted and output validated.
Tamper-evident audit: an append-only, hash-chained audit trail per organization, verified every day.
2,561 automated tests at 90% coverage run on every change, together with an end-to-end run of the whole stack behind TLS, a backup-and-restore drill, an OWASP ZAP scan of every API operation, and CodeQL, Semgrep, Bandit, Trivy, pip-audit and Gitleaks.
Tech stack: Python, FastAPI, PostgreSQL, n8n, RabbitMQ, Redis and Docker.
Want your own data sources monitored like this? Send me a message with the websites, feeds or spreadsheets you check today.
Like this project
Posted Oct 12, 2026
Collects business data from websites, APIs and spreadsheets, detects important changes, explains them with AI and alerts your team. 2,561 tests, 90% coverage.
