Dynamic Form Builder

A containerized, YAML-driven form builder for prototyping enterprise forms and dispatching them to GitHub Actions or Ansible Tower — with a built-in local AI assistant.

Why this exists

Enterprise teams often hit a coordination wall between developers building automation workflows and the specialized teams that own form platforms (ServiceNow, SailPoint, etc.). Dynamic Form Builder lets developers:

• Prototype forms immediately in YAML without waiting on other teams
• Validate the dispatch payload against a real GitHub Actions or Ansible Tower workflow
• Hand off the same YAML as a structured spec to whoever owns the production form

Features

• Live YAML editor — form renders and payload updates in real time as you type
• GitHub Actions + Ansible Tower — dispatch via repository_dispatch or launch an AWX job template
• Dynamic options — populate dropdowns/checkboxes from a live API with server-side pagination
• {{env:VAR}} placeholders — tokens resolved server-side so they never reach the browser
• Conditional logic — show_if rules with chained dependencies; hidden values excluded from the payload
• AI assistant — local Ollama model proposes YAML changes reviewable in a diff modal before applying
• Folder organization — drag-and-drop reorder, rename, multi-select bulk actions
• Dispatch history — every dispatch recorded with payload + response, viewable per form
• Four themes — ServiceNow, Aurora, Midnight, Atlassian

Architecture

Browser → nginx (static + reverse proxy) → Flask API → MongoDB
                                                     → Ollama (local LLM)

Service	Purpose	Host port
frontend	nginx — SPA + /api/* reverse proxy	8090
flask-backend	REST API, dispatch, proxy, SSE chat	3000
mongodb	Forms, folders, dispatch history	27017
mongo-express	Optional MongoDB GUI	8091
ollama	Local LLM runtime (llama3.2:3b)	11434

Quick start

git clone https://github.com/MBarc/dynamic-form-builder.git
cd dynamic-form-builder
docker-compose up -d

Open http://localhost:8090. The first boot pulls the llama3.2:3b model (~2 GB) for the AI assistant — the app is fully usable while that downloads.

Note: Default MongoDB credentials are changeme/changeme. These are for local development only.

Seed example forms

After the stack is up, populate three demo forms that exercise the dynamic source: feature:

python seed_source_templates.py

Configuration

Environment variables on the flask-backend service:

Variable	Default	Description
MONGO_HOST	localhost	MongoDB host
MONGO_INITDB_ROOT_USERNAME	changeme	MongoDB username
MONGO_INITDB_ROOT_PASSWORD	changeme	MongoDB password
OLLAMA_HOST	http://localhost:11434	Ollama runtime URL
OLLAMA_MODEL	llama3.2:3b	Model used by the chat assistant
Any {{env:VAR}} name	—	Resolved into source URLs/headers

Documentation

Full YAML schema reference — field types, source: config, pagination, show_if conditional logic, and complete examples — is available inside the app under Docs in the left sidebar.

Standalone mode (no Docker)

A single-file alternative stores forms in a local forms.json instead of MongoDB. The AI assistant is not available in this mode.

pip install -r requirements-standalone.txt
python standalone.py   # → http://localhost:8080

Project layout

backend/          Flask API
frontend/         nginx-served SPA (app.js, chat.js, styles.css)
database/         MongoDB init script with sample forms
standalone.py     Single-file no-Docker alternative
seed_source_templates.py
docker-compose.yml

License

See LICENSE.