Introduction
Welcome to the official documentation for Tillu Smart Search. Tillu is a premium, high-performance, AI-driven search ecosystem designed to connect standard web engines with multi-modal AI agents and web apps.
Agent-centric response design: Tillu extracts relevant content, strips clean HTML, performs vector context synthesis, and normalises responses into predictable structures.
System Structure
The Tillu Search pipeline is split into independent sub-components:
- SearXNG Aggregation: Fetches organic listings, images, news, or videos.
- Cleaner Scraper: Strips markup/scripts to retain readability text blocks.
- Synthesis LLM: Generates cohesive responses using a model chain starting with
openai/gpt-oss-120b.
Getting Started
Integrating Tillu into your agentic loops is simple. Run queries via HTTP requests using environment parameters defined in your deployment.
Environment Variables
SEARXNG_URL=http://localhost:8080
GROQ_API_KEY=gsk_your_key_hereAPI: Unified Pipeline
GET /api/unified
Runs the full sequence: Search -> Extract -> Summarise -> Structured response output.
Query Options
| Parameter | Type | Default | Function |
|---|---|---|---|
q | string | - | Search query to execute. |
mode | string | fast | fast (search+LLM), full (search+scrape+LLM), search (raw search). |
category | string | general | general, science, news, videos, social. |
API: Search proxy
GET /api/search
Requests raw search query arrays aggregated across dozens of public search engines.
API: Smart Scraper
GET /api/scraper
Downloads target webpage URLs and returns sanitised read-only text segments.
API: Context Synthesiser
GET /api/synthesiser
Accepts manual context lists and synthesises answers via LLM models.
Pipelines & Models
Tillu uses a redundant fallback model pool to guarantee search reliability:
openai/gpt-oss-120b(Primary)qwen/qwen3-32b(Secondary fallback)openai/gpt-oss-20b(Tertiary fallback)
Deployment
Tillu is optimized for zero-config deployment on Vercel or local Node.js containers.
npm run dev # Starts local tester dev server
npm run deploy # Direct deployment to production Vercel