API Reference

⚠️ Disclaimer: AI-generated responses may contain inaccuracies. Always verify important information from the original sources provided.

Getting Started

No API keys required! Simply make requests to the endpoints below.

Base URL

https://anveshasearch.vercel.app

🔍 Quick Search

POST /api/search

Search 7 engines in parallel, extract page content, get AI-synthesized answer.

Request Body:

{
    "query": "best python frameworks 2025",
    "num_results": 5
}

Response:

{
    "success": true,
    "query": "best python frameworks 2025",
    "ai_answer": "Based on the sources, the top Python frameworks...",
    "results": [
        {
            "title": "Top Python Frameworks",
            "url": "https://example.com/...",
            "content": "Django, Flask, and FastAPI lead...",
            "engine": "DuckDuckGo"
        }
    ],
    "total_results": 25,
    "engine_stats": {
        "duckduckgo": "✓ 5 results",
        "startpage": "✓ 5 results",
        "bing": "✓ 5 results",
        "brave": "✓ 5 results"
    }
}

🧠 Deep Research (Streaming)

GET /api/deep-research-stream

Real-time streaming research with live AI response generation.

Research Depth Modes:

Mode Questions Est. Time
surface 3 ~45 seconds
moderate 6 ~2 minutes
deep 10 ~4 minutes

Query Parameters:

Parameter Type Required Description
topic string Yes The research topic
depth string No surface, moderate, or deep
num_questions int No Override question count (1-10)

Example:

GET /api/deep-research-stream?topic=quantum+computing&depth=moderate

SSE Event Types:

status         → { "type": "status", "message": "Generating questions..." }
questions      → { "type": "questions", "questions": ["Q1", "Q2", ...] }
question_start → { "type": "question_start", "num": 1, "total": 6 }
engine_result  → { "type": "engine_result", "engine": "DuckDuckGo", "count": 5 }
question_done  → { "type": "question_done", "num": 1 }
report_start   → { "type": "report_start", "sources": [...], "total_sources": 25 }
report_token   → { "type": "report_token", "token": "The" }
report_done    → { "type": "report_done", "report": "# Full Report..." }
done           → { "type": "done" }

🤖 AI Fallback System

The API automatically handles AI model failures:

  1. Primary: Llama 3.3 70B (DeepInfra)
  2. Cache: Felo.ai cached answer
  3. Fallback: Qwen 72B (DeepInfra)
  4. Error: Friendly error message

📊 Other Endpoints

POST /api/felo

Search Felo.ai specifically for AI-powered results.

GET /api/engines

Get list of available search engines and their status.

GET /api/health

Health check endpoint.

⚡ How It Works

  1. Parallel Search — Queries 7 engines simultaneously
  2. Deduplication — Removes duplicate URLs
  3. Content Extraction — Fetches actual page content
  4. AI Synthesis — Streams response token-by-token

❌ Error Handling

{
    "success": false,
    "error": "Query is required"
}

All endpoints return success: false with an error message on failure.