YouTube Intelligence Platform

Mine comments, surface content opportunities, and turn viewer demand into AI-generated production briefs.

A full-stack audience research tool for YouTube creators. It ingests comments and replies at scale, then runs multi-layer analysis to answer one question: what should I make next, and why will it perform?

---

What it does

Audience Intelligence Index all comments across multiple channels. Surface pain points, questions, purchase intent signals, and content ideas — ranked by mention frequency and engagement.

Competitor Analytics Side-by-side channel comparison with AI-generated insights on content gaps, audience sentiment, and growth patterns.

Topic Intelligence Engine Clusters comments into semantic themes using TF-IDF + k-means (pure numpy, no ML deps). Tracks growth rate per theme, classifies topics as core vs adjacent, and caches results with fingerprint-based invalidation.

AI Content Strategy Kanban-style opportunity board with server-side pagination, per-card AI briefs, inline editing, notes, and archiving. Scores every opportunity across Demand, Engagement, Trend, and Relevance axes.

AI Content Recommendation Engine Generates and persists structured recommendations — title, hook, talking points, FAQs, misconceptions, and a priority score (0–100). Full workflow from draft → reviewed → approved → published.

Content Planner Converts approved recommendations into production-ready packages: video outline with timed sections, three thumbnail concepts with persuasion style, SEO keyword strategy, and a monthly content calendar.

---

Stack

Layer	Tech
Backend	FastAPI · SQLAlchemy · SQLite · async background tasks
Frontend	Next.js 14 App Router · TypeScript · Tailwind CSS
AI	Provider-agnostic — OpenAI / Anthropic / Gemini / OpenRouter / Ollama
Clustering	Pure numpy (TF-IDF + k-means) — no sklearn / torch

AI provider and model are swappable via the settings UI — no redeploy needed.

---

Quickstart

Prerequisites: Python 3.11+, Node.js 20+, a YouTube Data API v3 key.

git clone <your-repo>
cd youtube-intelligence
./start.sh

That's it. The script installs all dependencies, creates .env files, and starts both services.

	URL
App	http://localhost:3000
API docs	http://localhost:8000/api/docs

Manual setup

Backend

cd backend
python -m venv .venv
source .venv/bin/activate       # Windows: .venv\Scripts\activate
pip install -r requirements.txt
cp .env.example .env            # then add your YOUTUBE_API_KEY
uvicorn app.main:app --reload --port 8000

Frontend

cd frontend
npm install
npm run dev

---

start.sh commands

./start.sh            # install deps + start everything (default)
./start.sh start      # start without reinstalling
./start.sh stop       # stop both services
./start.sh logs       # tail live logs
./start.sh reset      # wipe DB and start fresh
./start.sh setup      # install deps only, don't start

---

Environment variables

backend/.env

YOUTUBE_API_KEY=AIzaXXX...

# AI provider — set one, leave others blank
AI_PROVIDER=none          # none | openai | anthropic | gemini | ollama | openrouter
AI_MODEL=                 # optional — defaults are set per provider
OPENAI_API_KEY=
ANTHROPIC_API_KEY=
GEMINI_API_KEY=
OPENROUTER_API_KEY=
OLLAMA_BASE_URL=http://localhost:11434

DATABASE_URL=sqlite:///./youtube_intelligence.db
CORS_ORIGINS=["http://localhost:3000"]

frontend/.env.local

NEXT_PUBLIC_API_URL=http://localhost:8000

You can also set the YouTube API key and AI provider directly in the Settings page of the app — no .env edit required after first run.

---

Getting a YouTube API key

1. Go to console.cloud.google.com
2. Create a project → APIs & Services → Enable YouTube Data API v3
3. Credentials → Create API Key
4. (Optional) Restrict to YouTube Data API v3

Daily quota: 10,000 units — importing 25 videos with comments uses ~125 units per creator.

---

Deployment

See DEPLOYMENT.md for step-by-step guides for Railway, Render, Fly.io, and VPS.

Docker

cp .env.example .env
docker compose up -d

---

Architecture

Browser (Next.js 14)
        │
        ▼
  FastAPI (Python)
        │
        ├── YouTube Data API v3
        ├── SQLite (persistent, swappable to Postgres)
        └── AI Provider (optional)
                ├── OpenAI
                ├── Anthropic
                ├── Gemini
                ├── OpenRouter
                └── Ollama (local)

---

Database

SQLite at backend/youtube_intelligence.db. For production, mount a persistent disk and point DATABASE_URL there. The SQLAlchemy layer means you can swap to PostgreSQL by changing one env var.

Scale reference: 1M comments ≈ 200MB · ~50ms queries with indexes. For 10M+ migrate to Postgres.