Back to FormBuddy

System design · September 2026

Forms become conversations.

Two product paths share one voice, intelligence, safety, and document pipeline.

Prepare, review, download.

01 · Complete plain-language architecture

Everything FormBuddy does, in one view

Follow the complete journey from speaking, typing, choosing a guided form, or uploading a PDF through conversation, safe document preparation, memory, review, and download. Current capabilities and future expansion are clearly separated.

Complete plain-language FormBuddy architecture showing ways to begin, the guided conversation, safe document preparation, memory boundaries, review and download, the human submission boundary, and future form tracking features.

Select the diagram to open the full-size image.

02 · Detailed technical architecture

Every component and data flow

Follow the as-built system from entry surfaces through browser orchestration, audio and interview pipelines, server functions, AI models, persistence, safety controls, operations, and the final human-reviewed PDF.

Detailed technical FormBuddy architecture showing entry surfaces, browser screens and state, voice and form pipelines, the guided and dynamic engines, server functions, AI models, persistence, trust boundaries, operations, and final PDF output.

Select the diagram to open the full-size image.

03 · Component inventory

Six clear layers

Experience

  • Guided DL-14A interview
  • Any fillable PDF
  • Live transcript
  • PDF preview and download

Orchestration

  • Voice activity detection
  • Interview state
  • Conversation context
  • Progress and corrections

Core rules

  • Deterministic DMV agent
  • Generic form engine
  • Sensitive-field filter
  • PDF read and write

AI services

  • Streaming transcription
  • Structured extraction
  • Question planning
  • Natural follow-ups and speech

Storage

  • Browser-only personal memory
  • Session resume pointer
  • Conversation sessions
  • Conversation turns

Platform

  • TanStack server functions
  • Lovable AI Gateway
  • Lovable Cloud database
  • Edge deployment and MCP

04 · Data and trust

Clear boundaries by design

Stays in the browser

Uploaded PDF bytes, the live draft, saved personal details, preferences, and the session pointer.

Reaches server services

Audio for transcription, answer text for structured extraction, and conversation turns for session continuity.

Deliberate boundary

FormBuddy prepares a draft. It does not submit forms or make legal, eligibility, or benefit decisions.

05 · Reality check

Built now versus next

The diagrams show the current system. Future capabilities are kept visibly separate from working product paths.

BuiltGuided Texas DL-14A voice interview
BuiltUpload and fill any PDF with fillable fields
BuiltStreaming speech recognition and spoken replies
BuiltBrowser memory and saved conversation context
BuiltReviewable PDF generation and download
CurrentChrome extension is a launcher to the hosted app
FutureIn-page extension automation for third-party forms
FutureNon-fillable scans, OCR, deadlines, and form tracking