MVP → Phase 1 → Phase 3 Technical Blueprint
(Apple Notes edition — clean, readable, emoji-enhanced)
⸻
🎯 Purpose of This Document
This Architecture Guide explains how TalentQuote works internally and how each subsystem interacts.
It covers: • Frontend stack • Backend automations • Database model • Integrations • AI layer • Security • Scaling strategy • Migration path (Phase 3)
This guide does not define screens (UX) or user stories — those live in separate documents.
⸻
🟦 1. Core Architectural Philosophy
TalentQuote v3.0 is built around three pillars:
1️⃣ No-code/low-code orchestration for speed
n8n acts as the “brain” of the system. It allows fast iteration, automated workflows, and reduced dependency on developers.
2️⃣ Supabase as the single source of truth for all talent & client data
PostgreSQL + Auth + Storage + RLS give us enterprise-grade reliability with SaaS speed.
3️⃣ Modular, replaceable components
Every part of the stack can evolve independently: • Webflow → replaced by Next.js later • Airtable → replaced by Supabase Recruiter Console • n8n workflows → replaced by microservices where needed • Workers/Functions → introduced only for scale-critical tasks
This ensures TalentQuote avoids unnecessary complexity while remaining future-proof.
⸻
🟧 2. Frontend Architecture (MVP → Phase 1)
🌐 Webflow (Marketing + Public Pages)
Purpose: • Landing pages • Pricing • Talent showcase (blurred profiles) • Client education • Lead capture forms
Why: • Designers iterate quickly • Consistent branding • SEO and speed built-in
Forms → Trigger n8n for lead processing.
⸻
⚙️ Wized (Authenticated App Layer)
Purpose: • Client dashboard • Recruiter dashboard (temporary) • Talent mini-dashboard (Phase 1) • MSA unlock state • Stripe/DocuSign embeds • Instant Match browsing
Why: • Works natively with Webflow • Auth, roles, CRUD, and data binding built-in • Direct connection to Supabase
This enables a full SaaS experience without building a frontend from scratch.
⸻
🛡️ Cloudflare
Purpose: • DNS • SSL • CDN • Firewall • Bot protection • Caching for global clients
This ensures the platform loads fast worldwide and protects talent data from scraping or attacks.
⸻
🟩 3. Backend & Automation Architecture
🤖 n8n — The Automation Brain
Everything flows through n8n: • Stripe webhooks • DocuSign callbacks • AI interview processing • Psychometric scoring • Slack alerts • Availability decay engine • Candidate skill-gap prompts • Recruiter notifications • Shortlist creation • ETL to Metabase • Airtable → Supabase sync
n8n acts as a visual backend, orchestrating all data flow with minimal code.
⸻
🗄️ Supabase — The System Heart
Core responsibilities: • Primary database (PostgreSQL) • Auth: clients, candidates, recruiters • Storage: PDFs, videos, thumbnails • RLS: row-level security for all views • APIs: automatically generated • Realtime streams for dashboards
Key Tables (conceptual): • candidates • candidate_skills • candidate_videos • psychometrics • availability_status • clients • roles • role_shortlists • msas, sows • stripe_events • recruiter_actions • ai_processing_logs
Supabase remains the true source of all operational data.
⸻
🗃️ Airtable — Recruiter ATS (Temporary Phase 1)
Purpose: • Candidate evaluation pipeline • Video generation tracking • Recruiter approvals • Edge-case notes • Exceptions handling • Quality control
Why temporarily: • Faster onboarding for recruiter team • Extends operational flexibility • Ideal for hybrid AI + human review in MVP phase
Migration plan: Replace Airtable with a native Recruiter Console in Supabase once volume exceeds ~3 recruiters or ~800 candidates.
⸻
💳 4. Integrations Layer
Stripe — Billing & Subscription Engine
Handles: • Checkout • Subscriptions • Buyout fees • Invoice generation • Payment retries • Webhooks → n8n → Supabase
Output feeds into: • Client Unlock State • Onboarding Dashboard • MRR/ARR calculations • Financial reporting
⸻
DocuSign — Legal Automation
Handles: • MSA signing • SOW signing • Signature verification • Audit trails • Document storage URLs
DocuSign → webhook → n8n → Supabase → dashboard unlock sequence.
⸻
🎥 HeyGen — AI Video Generation
Purpose: Generate short, high-quality AI video intros for each approved candidate, used to humanize profiles and reinforce trust.
How it works: • Candidate completes the async interview flow (text + recorded answers). • n8n composes a video script + metadata (name alias, role, stack, language). • n8n calls the HeyGen API to create or update a talking-head style intro video. • HeyGen returns a video URL or file reference. • n8n stores the reference in Supabase (candidate_videos table) and optionally pushes the file to Cloudflare Stream for secure delivery. • The video is: • shown as a blurred thumbnail pre-MSA, and • played in full quality post-MSA unlock for clients.
Error handling: • Any HeyGen failure is logged in ai_processing_logs / video_jobs. • Recruiters are notified in Slack if a candidate video fails or needs manual review.
⸻
Cal.com — Interview Scheduling
Features: • Inline scheduling UI • Calendar sync • Reminder system • Recruiter assignment
⸻
OpenAI — AI Layer
Used for: • Extracting skills from CV • Generating psychometric labels • Summarizing candidate videos • Detecting tone, clarity, reasoning • Creating “TalentScore (basic)” metrics
Runs entirely through n8n jobs, storing results in Supabase.
⸻
🧠 5. AI & Psychometrics Architecture
🔹 Phase 1: Lightweight AI • GPT-4o for psychometrics • 3–5 signal categories • Stored as metadata • No ML training required • Real-time scoring per candidate video • Video assets generated via HeyGen and linked to the candidate’s record in Supabase
This is supportive, not authoritative.
⸻
🔹 Phase 2: Behavioral + Performance Correlation
Once TalentQuote has: • 1,000+ videos • 300+ hires • 50+ feedback loops
We add: • Success correlation modeling • Role-cluster psychometrics • Redeployment predictions • Automated shortlist ranking (client-specific)
⸻
🔹 Phase 3: Predictive Engine • Neural patterns across talent history • Skill drift detection • Hiring fit probability • Culture match algorithms
This evolve TalentQuote into a hiring intelligence platform, not just a marketplace.
⸻
🔐 6. Security Architecture • Cloudflare WAF • Supabase RLS (strict role access) • Encrypted storage • n8n credential vault • Video URLs signed + expiring • Non-downloadable PDF embedding • Anti-scraping measures • Rate limiting on all profile views • Client roles → visibility levels (blurred → full)
Everything is designed to protect talent identity and enforce trust.
⸻
📈 7. Analytics & Observability
Metabase • MRR / ARR • Talent availability trends • AI accuracy patterns • Recruiter performance • Contract duration patterns • Conversion funnel (lead → MSA → SOW → hire)
Cloudflare Logs / Logtail • Traffic anomalies • Latency • DDOS detection
Future: Sentry + Doppler
For error tracking and secrets management.
⸻
🧱 8. Migration Path — Phase 3 (2026–2027)
🎯 Goal: scalable, investor-grade architecture
🔧 Frontend • Move dashboards to Next.js + Plasmic • Keep Webflow for marketing • Shared components synced to Figma
⚙️ Backend Micro-Layers
Instead of heavy microservices, we use Cloudflare Workers / Vercel Functions: • billing-worker • contracts-worker • ai-worker • video-worker • availability-worker
Benefits: • Near-zero cold start • Lower cost than AWS • Easy to version and deploy • Compatible with multi-venture platform (TalentQuote, DeskTify, Yobo)
🤖 n8n remains orchestrator
We do NOT remove n8n. We elevate it to: • master workflow system • error handling hub • cross-service orchestrator
🔄 Recruiter ATS Migration
Airtable → Supabase Recruiter Console with: • bulk actions • filtering • approvals • AI suggestions • recruiter KPIs
⸻
🧠 9. Scalability Notes (Critical Insights) • Supabase can handle millions of rows easily • n8n may need a dedicated cluster after ~200 workflows running concurrently • Stripe scales automatically • Cloudflare handles global delivery at enterprise scale • Wized may require partial migration as UI complexity grows • Multiple ventures can share micro-layers (AI, contracts, billing)
⸻
🎉 10. Architecture Summary (at a glance) • Frontend: Webflow + Wized • Backend: n8n + Supabase • Recruiters: Airtable (temporary) • AI: GPT-4o + psychometric tagging • Payments: Stripe • Legal: DocuSign • Security: Cloudflare + RLS • Dashboards: Metabase • Future: Next.js + Workers + shared micro-layers
TalentQuote’s architecture is intentionally: • simple to build • fast to iterate • powerful to scale • attractive to investors • aligned with a multi-product ecosystem