XavvyNess Research Engine

AI-powered research agent. Give it any question or topic — get back a structured, cited research report in seconds. Powered by live web search + smart model routing (Gemini 2.5 Flash / Llama 3.3 70B / Gemini 2.0 Flash).

Demo

🎬 Video demo coming soon. Upload research-engine.mp4 to YouTube, then run python3 scripts/actor-video-gen.py --embed-readmes to embed it here automatically.

---

What it does

Most AI research tools either hallucinate without web access, or dump raw search results with no synthesis. The XavvyNess Research Engine does both:

1. Searches the live web via Tavily — real URLs, real content, not stale training data
2. Synthesizes a structured report using the best model for the depth you choose
3. Caches results for 24h — re-running the same query is instant and free
4. Returns structured output — query, summary, full report, sources array, model used, timestamp

---

Input

Field	Type	Default	Description
query	string	(required)	What to research. Be specific for better results.
depth	enum	standard	quick (3-5 paragraphs) · standard (full report) · deep (comprehensive, multi-angle)
format	enum	markdown	markdown · bullet · json
includeSource	boolean	true	Include [n] citations and source URLs in report
maxResults	integer	10	Number of web sources to pull (3–30)

Example input

{
  "query": "What are the security risks of using LLMs in production APIs?",
  "depth": "deep",
  "format": "markdown",
  "includeSource": true,
  "maxResults": 15
}

---

Example output

Real output from a live run (truncated for display):

{
  "query": "What are the best open source AI coding assistants in 2025?",
  "depth": "standard",
  "format": "markdown",
  "summary": "AI coding assistants have become essential developer tools in 2025. Windsurf, GitHub Copilot, and Cursor lead adoption. Open-source alternatives like Aider and Continue offer strong customization without vendor lock-in. Tabnine and JetBrains AI excel at refactoring and language-specific completions.",
  "report": "## Overview\n\nThe use of Artificial Intelligence (AI) in coding has become increasingly popular in recent years...\n\n## Key Findings\n\n1. **Windsurf**, **GitHub Copilot**, and **Cursor** lead in advanced code generation and IDE integration [1, 2, 3].\n2. **GitHub Copilot** integrates natively with VS Code and Neovim [3, 4].\n3. **Aider** and **Continue** are top open-source alternatives with full local model support [5, 6].\n4. **Tabnine** and **JetBrains AI Assistant** excel at refactoring and language-specific completions [7].\n5. **Gemini** is purpose-built for Android app development [1].\n\n## Sources\n[1] https://... [2] https://...",
  "sources": [
    { "title": "Windsurf", "url": "https://windsurf.com", "snippet": "AI-first code editor..." },
    { "title": "GitHub Copilot", "url": "https://github.com/features/copilot", "snippet": "AI pair programmer..." },
    { "title": "Cursor", "url": "https://cursor.com", "snippet": "The AI code editor..." },
    { "title": "Aider", "url": "https://github.com/paul-gauthier/aider", "snippet": "AI pair programming in terminal..." }
  ],
  "model": "groq/llama-3.3-70b-versatile",
  "agent": "XavvyNess Research Engine",
  "runAt": "2026-04-08T22:21:45.123Z"
}

The full report field contains the complete structured markdown report (typically 600–2000 words for standard depth).

---

Model routing (automatic)

Depth	Model	Why
quick	Llama 3.3 70B via Groq	Sub-second inference, free tier, great for fast summaries
standard	Gemini 2.0 Flash	Best balance of speed and quality
deep	Gemini 2.5 Flash	Top reasoning for comprehensive, multi-angle analysis

No configuration needed — routing is automatic based on your depth input.

---

Use cases

• Competitive research — "Who are the top 5 competitors to Linear and how do they price?"
• Technical deep-dives — "What are the differences between RAG and fine-tuning for production LLMs?"
• Market analysis — "What AI agent startups received Series A funding in Q1 2026?"
• Due diligence — "What are the known scaling issues with Supabase at 1M+ rows?"
• Content research — "What are the most cited 2025 studies on remote work and productivity?"
• Security audits — "What CVEs affect Node.js 20 LTS as of April 2026?"

---

Caching

Results are cached for 24 hours per unique query + depth + format combination. Re-running the same query within 24h returns the cached result instantly at no additional compute cost.

---

Cost comparison

Tool	Pricing	Web search	Caching	Model routing
XavvyNess Research Engine	Pay-per-result	✅ Live (Tavily)	✅ 24h	✅ Automatic
Generic GPT wrapper	High flat rate	❌ Training data only	❌	❌
Basic search scraper	Per-page	✅	❌	❌

---

Integration

Via Apify JavaScript client

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: 'YOUR_APIFY_TOKEN' });

const run = await client.actor('RBobzxRYFVgoX74uu').call({
  query: 'What is the state of open-source LLMs in 2026?',
  depth: 'standard',
});

const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(items[0].report);    // full markdown report
console.log(items[0].summary);   // 2-3 sentence summary
console.log(items[0].sources);   // array of { title, url, snippet }

Via HTTP API

curl -X POST \
  "https://api.apify.com/v2/acts/RBobzxRYFVgoX74uu/runs?token=YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "Best vector databases for production in 2026",
    "depth": "standard",
    "format": "markdown"
  }'

Via Make.com / Zapier

Use the Apify module → Run Actor action. Actor ID: RBobzxRYFVgoX74uu. Pass your query in the input JSON, then map {{report}} and {{summary}} from the output to your next step.

---

Limitations

• Web search requires the TAVILY_API_KEY environment variable. Without it, the actor falls back to AI knowledge only (clearly flagged in status messages).
• deep depth runs may take 30–90 seconds for complex topics with many sources.
• Source quality depends on what Tavily surfaces — very niche topics may return fewer authoritative results.
• Report is generated in English regardless of query language (multi-language support planned).

---

About XavvyNess

XavvyNess is an AI agent platform focused on practical, production-ready automation. This actor is part of a suite of research and development tools built for developers and operators who need real answers, not hallucinations.

Questions or feature requests → open an issue or contact us via the Apify Store.