AI planning
Natural language becomes a structured automation plan with assumptions, warnings, and safety notes.
AI browser automation studio
Turn plain-English browser tasks into editable automation workflows.
Flowwright 2.0 is a portfolio orchestration surface: natural-language planning, a visual browser preview that tracks the hot step, a debugger-style timeline, step inspector, preflight quality checks, and reviewable extractions with exports—without pretending to be a database app.
- Orchestration-first
- Plans, runners, retries—not CRUD tables.
- Demo-realistic
- Deterministic mock AI + simulated worker for repeatable demos.
- Interview-ready
- Clear types, modular libs, obvious extension points.
End-to-end story
Prompt → plan → visual worker playback → reviewable data
Plain-English task
Goal + constraints
Automation plan
Typed steps
Browser worker
Simulated run
Extracted results
Review + export
Flowwright 2.0 is a demo / portfolio build — no live headless browser in this repo, but the UX mirrors a real automation debugger.
Everything reinforces automation—not admin screens
Employers should feel orchestration, observability, and extraction within seconds. No fake “analytics” charts as a substitute for substance.
Visual worker playback
A simulated browser chrome shows URL, hot targets, loading, and capture—like a lightweight debugger, not a static list.
Live run logs
Append-only audit trail with levels, timestamps, and step correlation—the kind of evidence ops teams expect.
Data extraction
Rows land in a typed results grid with confidence and review flags, then export to JSON or CSV.
Human review
Edit targets and descriptions inline before (or after) generation; humans stay in the loop.
Responsible boundaries
Explicit guardrails: permissioned sites only, no bypassing access controls, synthetic demo fixtures here.