Tech Stack & Cost Breakdown

A core engineering goal of this digital portfolio is running, scaling, and managing the entire monorepo workspace under a $0.00/month budget.

💰 The Zero-Cost Stack Catalog

The table below lists all tools, platforms, plans, and monthly costs of the project:

Layer / Service	Tech Provider	Plan Tier	Monthly Cost	Free-Tier Limits
Frontend Web Hosting	Vercel	Personal / Hobby	`$0.00`	Bandwidth: 100GB/mo.
Serverless Routes	Vercel	Edge Functions	`$0.00`	Executions: 1M per month.
Cron Dispatcher	GitHub	Actions Run	`$0.00`	Time: 2,000 minutes/month.
Database Logs	Supabase	Free Tier	`$0.00`	Storage: 500MB, Row Limit: 50k.
Analytics Storage	Vercel Postgres (Neon)	Hobby Database	`$0.00`	Storage: 256MB, Compute: 60h/mo.
2FA Passcode Gate	Google Authenticator	TOTP Protocol	`$0.00`	Free mobile app generator.
Email Gateway	Resend	Developer Tier	`$0.00`	Quota: 3,000 emails/month (100/day limit).
AI Models (Main)	Google AI	Gemini Developer	`$0.00`	Rate Limit: 15 RPM (Requests Per Minute).
AI Models (Failover)	OpenRouter	Free Tier list	`$0.00`	Rate Limit: Varies by model (standard 10 RPM).
PDF Compilation	PDF-Lib / PDF.js	WebAssembly (client)	`$0.00`	Executed locally in the browser memory.
Audio Synthesizer	Web Audio API	Client Browser	`$0.00`	Executed locally using the Web Audio API.
URL Base64 Deflate	Browser Engine	Client CPU	`$0.00`	Executed locally in the browser thread.

🔒 Free-Tier Protection Mechanisms

To ensure active user traffic never breaches provider quotas (which can cause service suspensions or billing card charges), the workspace implements three primary guardrails:

1. Database Circuit Breakers

To prevent log spam from exhausting free storage limits, the database adapters implement row-cap circuit breakers: * Analytics Logs: Caps active visitor telemetry to 100 rows via a serverless FIFO delete query after each write, maintaining a ~20KB storage footprint. * Trace Limits: Caps active rows to 200 traces in Aileron. * Feedback Limits: Caps active rows to 50 corrections in Aileron. * Purge Mechanism: Runs a FIFO delete query after each insert to keep database storage under 1MB:

def enforce_circuit_breaker(self):
    conn = self.get_connection()
    cursor = conn.cursor()

    # 1. Cap traces to 200 rows max
    cursor.execute("SELECT COUNT(*) FROM aileron_traces")
    if cursor.fetchone()[0] > 200:
        cursor.execute("DELETE FROM aileron_traces WHERE id IN (SELECT id FROM aileron_traces ORDER BY created_at ASC LIMIT 10)")

    # 2. Cap feedback corrections to 50 rows max
    cursor.execute("SELECT COUNT(*) FROM aileron_feedback")
    if cursor.fetchone()[0] > 50:
        cursor.execute("DELETE FROM aileron_feedback WHERE id IN (SELECT id FROM aileron_feedback ORDER BY created_at ASC LIMIT 5)")

    conn.commit()
    conn.close()

Instead of storing diff configurations or YAML/JSON formatting templates in a database, CodeForge compresses state client-side using CompressionStream('deflate') and writes the output directly to the URL hash fragment. This enables link sharing with zero cloud storage costs.

3. SessionStorage Caching

Toggles and adjustments for CV customization are saved directly to the browser's sessionStorage. When the visitor navigates to /resume, the print view reads these values directly from browser memory, avoiding database lookup fees.

4. Client-Side CPU Delegation

Heavy tasks (such as Myers diff alignments, PDF parsing/merging, Web Audio oscillator calculations, and text validations) are run locally in the browser thread, keeping Vercel Edge execution times under 100ms and avoiding serverless timeouts.