# Datasheet: Quantum XR (Research Engine)

## Product Overview
* **Tagline:** Autonomous Research & Math-Verified Intelligence
* **Recommended for:** Analytical Teams, Data Science Teams, and Corporate Intelligence departments requiring high-accuracy quantitative reporting.

---

## Core Differentiators
* **Sandboxed Code Execution:** Standard RAG search engines easily hallucinate numbers. Quantum XR spins up coordinated agent networks that write and run local Python code to verify all calculations and data tables.
* **Multi-Source Synthesis:** Synthesizes intelligence from web databases, academic papers, local files, and APIs into structured PDF/JSON reports.
* **Deterministic Validation:** LLM creative drafting is strictly funneled through quantitative checks to ensure zero factual errors.

---

## System Requirements

### Hardware Specs
| Component | Minimum | Recommended (Production) |
|-----------|---------|-------------------------|
| **CPU** | 2 cores | 4+ cores (for concurrent agent processing) |
| **RAM** | 2 GB | 4 GB |
| **Disk** | 500 MB | 2 GB (for local database and research assets) |
| **GPU** | Not required | 8+ GB VRAM (Only if running local Ollama models) |

### Software & Environment
* **Supported OS:** Windows 10/11, Linux (Ubuntu/Debian), macOS 11+.
* **Runtimes:** Python >= 3.9.
* **LLM Backends:** OpenAI API, Anthropic Claude API, Google Gemini API, or local Ollama runtime.

---

## Technical Architecture & Appendix

### Verification Sandboxing
Every quantitative research query triggers a sandboxed environment where a Specialist Agent writes verification scripts:
```python
def verify_financial_claim():
    # Automatically generated code cell
    revenue = [12.5, 14.8, 18.2]
    growth = [(revenue[i] - revenue[i-1])/revenue[i-1] for i in range(1, len(revenue))]
    assert abs(sum(growth)/2 - 0.207) < 0.01, "Growth rate verification failed"
```
If the code returns an assertion error, the agent corrects its narrative draft automatically.

### Network Requirements
* HTTPS outbound access to LLM provider endpoints (Gemini, Claude, OpenAI).
* Local network access to port `11434` if using Ollama.