AI Products

AI Product Lifecycle

BRICQS treats AI products as first-class citizens — not just deployments. Every product has a tracked lifecycle, versioned releases, a planning history, and a live infrastructure topology.

The 9 lifecycle stages

Each AI product automatically moves through these stages as you build and ship it. The current stage is derived from the product's build status, deployment status, and evaluation status — you don't manage it manually.

plan

Initial state. The product exists in the registry but has no runtime configuration yet. Use this stage to let the AI Planner generate your architecture blueprint.

build

A build is in progress. The AI model or runtime component is being assembled and linked to the project vault, storage, and database.

configure

Build complete. The product is ready for environment variable and secrets configuration before its first deployment.

test

Deployed to a preview environment. Real traffic can be sent to the preview endpoint. Iterate here before promoting.

evaluate

Formal evaluation in progress — automated benchmarks, manual review, or both. A product cannot be released until evaluation passes.

release

Evaluation approved. A versioned release has been created with an immutable snapshot of the model, config, and capabilities. Release comparison diff is computed automatically.

deploy

The release is deployed to production. The real Azure Container App is live and serving traffic.

monitor

The deployment is live and operational. Azure Monitor platform metrics (CPU, memory, requests, latency) are streaming in real time.

improve

A new iteration has started. The current production release is still live while the next version moves through the plan → build cycle.

AI Planning Engine

When you create a new AI product and describe what it does, BRICQS's AI Planning Engine generates a complete architecture blueprint. The blueprint includes:

Architecture pattern

The high-level design — RAG pipeline, agent loop, fine-tuned GPU model, etc.

Execution graph

Node-by-node flow: trigger → model → retriever → response, with edge types.

Cloud cost estimate

Projected test and production costs with cost drivers and scaling notes.

Latency projection

p50 and p95 latency estimates, identified bottleneck, and optimization tips.

GPU requirement

Whether a GPU deployment is needed, and why.

Security review

Data sensitivity rating, identified risks, mitigations, and compliance notes.

The blueprint is stored on the project record. You can regenerate it at any time by updating your product description. All fields are optional — the AI fills in what it can from the context you provide.

Infrastructure Graph

Every AI product has an Infrastructure tab that renders a live topology diagram. The topology is derived from actual provisioned resources associated with the project, not from the blueprint — what you see is what exists.

What shows up

·Runtime node — always shown; links to the active deployment endpoint

·API Endpoint — the real Azure Container Apps FQDN

·Secrets Vault — if the project has any secrets configured

·Storage Bucket — if cloud storage is provisioned for this project

·Database — if a managed Postgres instance is linked

·Edge Network — if WAF / CDN routing is active for the project

·GPU node — shown only for GPU catalog deployments

Live status dots indicate whether each resource is actively healthy. The topology updates each time you navigate to the Infrastructure tab — it reflects the real state of your project at that moment.

API reference

POST/projectsCreate an AI product. Triggers blueprint generation if description is provided.

GET/projects/{id}Fetch a product with its full blueprint, releases, and environments.

GET/registry/productsList all products in your org with lifecycle stage derivation.

GET/registry/products/{id}Full product detail with versions, releases, and environments.