Private / redacted
AI-Assisted Contract Workflow
A redacted workflow where a React chatbot gathers missing contract details, creates or updates a Django editing session, and opens the generated contract in a React configurator.
- Role
- Designed and delivered the contract configurator, conversational assistant, Django integration, FastAPI service, and production flow.
- Result
- The assistant created or updated editing sessions, then returned generated contracts to the configurator with editable values populated.
- Stack
- React · TypeScript · Django · FastAPI · RAG · Stripe
System architecture
The building blocks in view
Read the system as text
- Contract userUses the conversational assistant and contract configurator.Connects to React chatbot (converses).
- React chatbotA conversational app exported through the same public-site framework as the configurator.Connects to FastAPI AI service (conversation + edit URL), React configurator (opens edit URL).
- React configuratorAn editable contract app exported through the same public-site framework.Connects to Django application (reads and writes).
- Django applicationShared contract data, editing sessions, and application integration.Connects to Editing sessions (persists).
- Editing sessionsPersisted contract state created or updated from structured assistant output.
- FastAPI AI serviceMulti-step LLM flow that receives chatbot context and creates or updates Django-backed editing sessions.Connects to Django application (creates / updates).
- StripePurchase and controlled-access events reach the Django application by webhook.Connects to Django application (webhook).
Key decisions
What shaped the implementation
Separate the assistant from the configurator
- Reason
- Allowed the conversational workflow to gather missing information before users entered the editable contract experience.
- Trade-off
- Added an integration boundary to maintain.
Keep editing sessions in Django
- Reason
- Gave both React applications one shared contract state before the generated contract opened in the configurator.
- Trade-off
- Assistant output had to be coordinated through the application backend.
Outcome
Verified delivery
Product workflow
Chatbot, backend editing session, generated contract, configurator, payment, and controlled access.