Skip to main content

AI Smart Reports 📝

Vartovii provides deeper insights through its AI Smart Report engine. Unlike standard dashboards that just show numbers, AI Smart Reports generate comprehensive, human-readable research briefs (PDFs) that summarize available risk, market, and operational signals.

Overview​

The AI Smart Report system uses Gemini 3.1 Pro preview (with automatic Gemini 2.5 Pro fallback) to synthesize data from multiple sources (Risk Engine, Market Data, Team Intelligence, Security Audits) into a professional 10-page document.

Key Capabilities​

  • Executive Summary: A concise "Buy/Sell/Hold" thesis generated by AI.
  • Risk Analysis: Deep dive into the 6 pillars of the Trust Score.
  • Competitor Comparison: Automatically compares the target project against top rivals in its category.
  • Team Profiling: Detailed background checks on founders and core developers.
  • Visual Data: Includes generated charts for "Trust Score Composition", "Tokenomics", and "Development Activity".

How It Works​

  1. Data Aggregation: The backend ReportDataAggregator collects live data from:
    • Database (Trust Scores, Team Data)
    • BigQuery (Historical trends)
    • External APIs (DefiLlama, Certik, TweetScout)
  2. AI Analysis: The gathered context is sent to Gemini 3.1 Pro preview with a specialized system prompt designed for structured research output.
  3. Deep Research Workflow: A PDF-only workflow layer then assembles the output into a staged research artifact with:
    • Entity Brief
    • Agent Workflow
    • Forensic + Corporate Fusion
    • Outlook Confidence
    • 30/60/90 Day Action Plan
  4. PDF Generation: The AI now returns a schema-backed JSON response that is validated before the Python PDF renderer consumes the canonical report sections, which:
    • Renders Markdown text to formatted PDF.
    • Generates dynamic charts using Matplotlib.
    • Applies Vartovii branding and styling.
  5. Delivery: The user receives a downloadable PDF file.

Trust Score Consistency​

Corporate AI reports now reuse the same canonical Trust Score path as the overview/dashboard surface.

If cached report text contains stale embedded Trust Score values after fresher review data arrives, the report is refreshed before delivery so the report card and the supporting narrative stay aligned to the same available data snapshot.

The downloadable PDF can include deeper workflow-oriented narrative blocks without changing the dashboard JSON contract used by on-screen report views.

Report Sections​

  1. Cover Page: Project Logo, Current Trust Score, Date.
  2. Executive Thesis: High-level verdict.
  3. Fundamental Analysis: Financials, Treasury, Tokenomics.
  4. Technical Analysis: Development activity, Code quality, Security.
  5. Social Analysis: Community strength, Sentiment.
  6. Team & Backers: Founder implementation capabilities and Investor quality.
  7. Disclaimer: Signal-based summary, not legal, investment, or employment advice.

Technical Implementation​

The system is built on a Python backend pipeline:

# Pseudo-code workflow
aggregator = ReportDataAggregator(project_slug)
context_data = await aggregator.collect_all_data()

analyst = AIAnalyst(model="gemini-3.1-pro-preview")
report_content = await analyst.generate_investment_memo(context_data)

pdf_gen = PDFReportGenerator()
pdf_file = pdf_gen.create_pdf(report_content, context_data)
# Returns binary stream to frontend

Prompt Integrity Contract (Updated 2026-03-03)​

To reduce hallucinations in generated memos, report prompt policy now enforces:

  • Use direct evidence from provided review input only.
  • Do not invent or fabricate quotes.
  • Do not present unsupported claims as sourced facts.

Usage​

Users can generate reports directly from the Crypto Dashboard:

  1. Select a Project.
  2. Click the "AI Report" button in the header.
  3. Wait for the "Thinking..." process (approx. 15-30 seconds).
  4. Download the generated PDF.