Skip to content

AI-STACK-REPORT

Archived record

This page describes the Firebase-era platform or a migration step that has completed. It is kept as history and is not a current runbook. The current platform is described from the home page.

Last Updated: February 23, 2026 Status: Active Review (Cost-first baseline)

Scope

Review current ContentForge AI stack, enforce Gemini 2.5 minimum, and recommend model allocation with cost as primary priority and performance as secondary.

Current Stack (Code Reality)

  • AI router: functions/lib/ai-model-router.js
  • Active provider switch: CONTENTFORGE_AI_PROVIDER
  • Current AI entry points:
  • Draft generation: functions/cms-functions.js (generateArticleDraft)
  • Article metadata analysis: functions/index.js
  • Architecture multimodal analysis: functions/index.js
  • Supported providers: Vertex, OpenAI, Anthropic

Gemini 2.5 Cost Signals (Vertex pricing page)

From Google Vertex AI Generative AI pricing (retrieved February 23, 2026):

  • Gemini 2.5 Flash Lite (lowest cost in Gemini 2.5 family)
  • Input (text/image/video): $0.10 / 1M tokens
  • Output text: $0.40 / 1M tokens
  • Gemini 2.5 Flash
  • Input (text/image/video): $0.30 / 1M tokens
  • Output text: $2.50 / 1M tokens
  • Gemini 2.5 Pro
  • Input: $1.25 / 1M tokens
  • Output text: $10.00 / 1M tokens

Pricing may change; verify on the Vertex pricing page before final capacity commitments.

Decision (Cost First, Performance Next)

Default Allocation

  • Draft generation → gemini-2.5-flash-lite
  • Rationale: highest volume path and most token-sensitive workload.
  • Article analysis → gemini-2.5-flash
  • Rationale: better quality/consistency than Lite with moderate cost.
  • Multimodal architecture analysis → gemini-2.5-flash
  • Rationale: image+text analysis reliability at lower cost than Pro.

Escalation Rule

Use gemini-2.5-pro only for targeted high-stakes workflows where quality gain is validated and budget-approved.

Implementation Completed

  • Updated Vertex default models in functions/lib/ai-model-router.js:
  • draft default → gemini-2.5-flash-lite
  • analysis default → gemini-2.5-flash
  • multimodal default → gemini-2.5-flash
  • generic fallback default → gemini-2.5-flash-lite

This enforces Gemini 2.5 minimum for default Gemini usage.

CONTENTFORGE_AI_PROVIDER=vertex
CONTENTFORGE_VERTEX_DRAFT_MODEL=gemini-2.5-flash-lite
CONTENTFORGE_VERTEX_ANALYSIS_MODEL=gemini-2.5-flash
CONTENTFORGE_VERTEX_MULTIMODAL_MODEL=gemini-2.5-flash

Follow-up Monitoring

  • Track per-endpoint token and cost spend by purpose (draft, analysis, multimodal).
  • Sample quality checks weekly for draft and architecture paths.
  • Promote to Pro only where measured quality delta justifies cost.

Deployment Readiness Additions

  • New admin endpoint: aiStackReadiness (Cloud Functions)
  • New script: functions/check-ai-stack-readiness.js
  • Deployment runbook: documentation/pipeline-ai-deployment.md