Changelog

What we've shipped

A record of real feature releases, infrastructure milestones, and platform improvements — in reverse chronological order.

July 2026PLATFORM

Real Azure Monitor metrics for all deployments

CPU%, memory%, GPU utilization%, request count, and response time are now pulled directly from Azure Monitor platform metrics for every Container App deployment. No instrumentation required — metrics appear automatically for any running deployment.

  • CpuPercentage, MemoryPercentage, GpuUtilizationPercentage from Azure Monitor
  • ResponseTime and Requests metrics shown in deployment detail drawer
  • Requests/sec added to the deployments summary strip
  • High-latency recommendation added to Deployment Intelligence (threshold: 3 000 ms)
June 2026DEPLOYMENTS

Three-environment deployments: development, preview, production

Each GPU model deployment can now target one of three isolated environments. Development, preview, and production each provision a separate Azure Container App with a distinct FQDN. One-click promotion from preview to production replaces the production deployment with zero downtime.

  • Development / preview / production environment selector in deploy modal
  • Live deploy pipeline panel showing real provisioning stages
  • Promote-to-production action retires the old production deployment after the new one reaches running state
  • Source selector scaffolded (Model Catalog functional; GitHub coming soon)
May 2026INFRASTRUCTURE

Deployments cloud-OS redesign

The deployments page was rebuilt as an operational cloud-OS interface. Deployment cards show live status, replica counts, GPU type, and real Azure Monitor metrics. A summary strip shows fleet-level aggregates. An activity feed shows real audit events.

  • SummaryStrip: active deployments, GPU hours, active regions, failure rate, runtime count, requests/sec
  • DeploymentCard: real CPU/GPU/memory/latency from Azure Monitor
  • ActivityFeed: polls /deployments/activity every 15 s — real audit_log rows
  • Detail drawer: Overview / Logs / Revisions / Scale / Metrics / Env Vars / History tabs
April 2026DATABASES

Managed Postgres — Table Editor, SQL Editor, GitHub migrations

The Databases section was restructured into four dedicated panels. Table Editor provides a spreadsheet-style row interface. SQL Editor supports saved/shared/private query management. The Database panel shows real server stats. GitHub integration applies migrations automatically on push.

  • Table Editor: inline cell editing, new row, column sort, pagination
  • SQL Editor: saved queries with Private/Shared/Favorites scoping
  • Real Azure Postgres SKUs in the compute tier picker
  • GitHub OAuth integration with migration auto-apply on push to main
March 2026EDGE

Edge network: WAF, bot scoring, Geo-IP, Redis CDN

The BRICQS edge layer reached production. Every domain served through BRICQS routes through a real edge router with WAF rule evaluation, multi-signal bot scoring, RIR-delegated geo-IP resolution, and Redis CDN with zero stored credentials.

  • 9 OWASP WAF rule categories active on all traffic
  • 6-signal bot scoring with composite score threshold
  • 259,824 IPv4 ranges from ARIN/RIPE/APNIC/LACNIC/AFRINIC
  • Redis CDN authenticated via managed identity — key-based auth permanently disabled
February 2026RUNTIMES

Five AI runtimes now in production

All five BRICQS AI execution models reached production status: LLM API gateway, GPU container runtime, agent reasoning loop, RAG retrieval pipeline, and workflow orchestration engine.

  • LLM API: OpenAI-compatible endpoint, any provider, secrets-vault-backed
  • GPU Runtime: LLaMA-3, Mistral-7B, Phi-3, Whisper, Qwen2.5 on dedicated T4 containers
  • Agent: tool-using reasoning loop with calculator and HTTP tools, full trace output
  • RAG: PDF/DOCX/TXT ingestion, embedding, vector search, grounded answers
  • Workflow: multi-step pipelines with cross-runtime calls and template variable passing