Skip to main content

AI Architect chat

AI Architect is PinPole’s assistant for turning natural language (and optionally images of diagrams) into nodes and edges on the canvas. Open it from the sparkles button on the canvas toolbar.

What it does​

  1. Chat — You describe requirements, constraints, or changes. The assistant responds with explanations and structured thinking steps (requirements, service selection, topology, scale, security, cost, etc.).
  2. Streaming — Responses can stream in real time; the UI shows thinking phases with icons until the answer stabilizes.
  3. Architecture payload — When the model returns a graph, PinPole can apply nodes and edges to the current diagram, with hooks to fit view and tidy layout after generation.
  4. Diagram upload — You can attach an architecture diagram image for analysis; the assistant can summarize, suggest improvements, or help recreate topology (behavior depends on the backend prompt and model).

Provider context​

The panel is aware of the active cloud provider (AWS / Azure / GCP) where applicable, so suggestions align with the palette you are using.

Good prompts​

  • Mention ingress (e.g. API Gateway, Load Balancer), compute (Lambda, ECS, Cloud Run), data (DynamoDB, RDS, Firestore), and async pieces (SQS, Pub/Sub, Event Bus).
  • State non-functional goals: RPS, latency, multi-AZ, compliance.
  • Ask for incremental changes: “Add a dead-letter queue after the worker.”