A 4-engine content intelligence platform that extracts a brand's DNA, tracks live industry trends, builds SEO/AEO/GEO-scored article briefs, and writes articles that sound exactly like the brand.
Runs entirely on localhost. No subscriptions. Powered by Gemini (free tier) and Groq (free tier).
SearchOS is a self-hosted brand content platform built around a sequential 4-engine pipeline. Each engine feeds the next one, producing articles that are simultaneously optimized for:
- SEO - Traditional search engine ranking signals
- AEO - Answer Engine Optimization for featured snippets and AI Overviews
- GEO - Generative Engine Optimization for citation by ChatGPT, Perplexity, and Claude
The system manages multiple clients. Each client's full pipeline output - DNA, trends, briefs, and articles - is stored locally in a structured directory layout and a SQLite database.
Engine 01 – Brand DNA Extractor
Scrapes a brand's website across up to 60 pages using Playwright + BFS crawling, mines JSON-LD structured data, runs DuckDuckGo searches, and synthesizes everything into a structured brand profile via Gemini.
Engine 02 – Live Trend Research
Pulls real-time signals from Google News RSS and DuckDuckGo, then uses Gemini to classify each result into one of four segments: brand_news, brand_future, industry_trend, competitive. Outputs 10–14 ready-to-write article angles.
Engine 03 – Article Brief Builder
Runs gap analysis between existing brand coverage and incoming trends. Produces a structured brief with SEO, AEO, and GEO scores assigned before a word is written.
Engine 04 – Article Writer
Writes a complete article in the brand's voice using the brief and Brand DNA as context. Routes between Gemini 2.0 Flash (primary) and Groq (fallback). Filters AI clichés and computes final scores.
Backend
| Package | Purpose |
|---|---|
| FastAPI + Uvicorn | REST API server |
| Playwright | Full-render browser scraping |
| httpx | Async HTTP client for BFS crawling and fallback fetches |
spaCy (en_core_web_sm) |
NLP noun chunk extraction for keyword analysis |
| DDGS | DuckDuckGo web search - no API key |
| google-genai | Gemini 2.0 Flash - services, tone, USPs, classification, article writing |
| Groq | LLM fallback for article generation |
| aiosqlite | Async SQLite for metadata persistence |
| python-dotenv | Environment variable loading |
Frontend
| Package | Purpose |
|---|---|
| React 18 | UI framework |
| Vite 6 | Dev server and production bundler |
| Framer Motion | Animated landing page and transitions |
Storage
| Layer | What lives here |
|---|---|
data/searchos.db |
Client metadata, brief records, article metadata, SEO/AEO/GEO scores |
clients/{slug}/01_brand_dna/ |
company_dna.json |
clients/{slug}/02_trend_research/ |
trends_{timestamp}.json |
clients/{slug}/03_article_briefs/ |
brief_{timestamp}_{title}.json |
clients/{slug}/04_articles/ |
{slug}.md (final article in Markdown) |
- Python 3.11 or higher
- Node.js 18 or higher
- Chromium - installed automatically by Playwright on first run
API keys required (both free):
| Service | Free tier | Link |
|---|---|---|
| Google AI Studio (Gemini) | 1,500 requests/day on gemini-2.0-flash |
aistudio.google.com/app/apikey |
| Groq | 14,400 requests/day | console.groq.com |
No other paid services, subscriptions, or cloud infrastructure required.
git clone https://github.com/your-username/search-optimization.git
cd "Search Optimization"cd backend
pip install -r requirements.txtplaywright install chromiumpython -m spacy download en_core_web_smcd ../frontend
npm installnpm run buildThe built assets in frontend/dist/ are served directly by the FastAPI server, so you do not need to run a separate dev server in production.
Copy the example environment file and fill in your keys:
cp backend/.env.example backend/.env# backend/.env
GOOGLE_API_KEY=your_gemini_api_key_here
GROQ_API_KEY=your_groq_api_key_hereBoth keys are used only on your local machine and are never sent anywhere except the respective APIs.
# From the project root
npm run serverThis starts the FastAPI server on http://localhost:8000. The frontend is served from frontend/dist/ at the same origin - no separate frontend process needed.
Open two terminals:
# Terminal 1 - Backend
cd backend
python run.py
# Terminal 2 - Frontend (hot reload)
cd frontend
npm run devThe Vite dev server runs on http://localhost:5173 and proxies API calls to http://localhost:8000.
For architecture diagrams, full API reference, database schema, pipeline details, and development notes —
View the full technical documentation in DOCS.md →
