PLATFORM GUIDE

How Every Agent Works

A complete walkthrough of all 9 modules — what they do, how to use them, and what to expect.

MODULE A

Supplier Risk Analysis

Real-time AI risk scoring for any supplier worldwide.

⚠️

Note: All scores are AI-generated from public data. Not financial or legal advice. Always verify critical decisions with primary sources. Takes 15–45 seconds.

INPUT

Enter supplier details

Provide the company name, industry, and country. The more specific, the better — use full legal names (e.g. "Evergrande Group" not "Evergrande").

STEP 1

News Agent scans intelligence

Queries AI knowledge for bankruptcies, sanctions, disasters, scandals, and leadership changes. Qwen 2.5 72B

STEP 2

Risk Agent scores the supplier

Applies structured scoring logic to produce a 0–100 risk score with specific risk factors and recommended actions. DeepSeek-R1

STEP 3

Price Agent checks market

Analyzes commodity price trends for the supplier's industry — giving market context to the risk picture.

OUTPUT

Full risk report

Risk score + level + summary + factors + actions + confidence + news summary + price data.

0–25

✅ Low Risk

Supplier appears stable. Safe to proceed with standard due diligence.

26–50

🟡 Moderate Risk

Some concerns. Monitor closely and consider requesting financial statements.

51–75

🟠 High Risk

Significant issues. Proceed with caution. Seek alternative suppliers.

76–100

🔴 Critical Risk

Severe risk. Consider suspending this supplier relationship immediately.

📄
MODULE A

Document Analysis

Upload any supplier PDF — AI reads and extracts financial risk signals.

⚠️

Note: Works best with text-based PDFs (not scanned images). Supports annual reports, contracts, financial statements. Max ~30 pages analyzed. Takes 30–60 seconds.

INPUT

Upload a PDF file

Annual report, contract, financial statement, or any supplier document. Enter the supplier name for context.

STEP 1

Text extraction

System extracts text from PDF using pypdf / pdftotext. Up to 12,000 characters (roughly 30 pages) are analyzed.

STEP 2

AI financial analysis

Reads the extracted text and identifies revenue trends, debt levels, profit/loss, red flags, and positive indicators. Qwen 2.5 72B

OUTPUT

Structured risk verdict

Financial health summary + red flags + positive indicators + key risks + risk level + recommendation (SAFE / MONITOR / CAUTION / AVOID).

Best documents to upload: Annual reports (10-K, 10-Q), audited financial statements, supplier contracts with financial terms.
Scanned PDFs won't work well. The document must contain selectable text. If you can copy text from the PDF, it will work.
Combine with Supplier Risk. Run the supplier name check first, then upload their annual report for deeper financial analysis.
📸
MODULE A

Visual Check

Upload a factory or warehouse photo — AI assesses physical condition and risk.

⚠️

Note: Supports JPG, PNG, WebP images. Analysis is based on visible elements only — AI cannot detect hidden structural issues. Takes 20–45 seconds.

INPUT

Upload a factory photo

Any photo of the supplier's factory, warehouse, or facility. Clear, well-lit photos give better results.

STEP 1

Vision AI analyzes image

A multimodal vision model examines the image for physical condition, safety issues, equipment state, and capacity signals. Qwen 2.5 VL (fallback chain)

OUTPUT

Physical risk assessment

Condition rating (Poor/Fair/Good/Excellent) + Safety level + Red flags + Positive signs + Capacity status + Visual risk score (0–100).

What the AI looks for: Flood or fire damage, abandoned equipment, structural issues, overcrowding, safety hazards, modern vs outdated machinery.
Best photo types: Wide shots of the factory floor, exterior building shots, warehouse storage areas, production lines.
If vision model is unavailable: The system will return a "manual inspection required" notice — not an error. This is the text-only fallback.
🚀
MODULE B

DealFlow AI

Find global buyers for your product — AI scouts, qualifies, and writes outreach emails.

INPUT

Describe your product

Enter your product (e.g. "handmade leather bags"), industry (e.g. "fashion"), target region (EU, North America, Asia…), and your company name.

STEP 1

Scout Agent finds buyers

AI searches its knowledge for matching importers, retailers, distributors, and manufacturers in your target region. Returns up to 5 qualified leads. Qwen 2.5 72B

STEP 2

Qualify Agent scores each buyer

Each buyer gets a fit score (0–100) based on company size, type, and alignment with your product and industry.

STEP 3

Outreach Agent writes the email

Click "Generate Email" on any buyer card — AI writes a personalized cold outreach email with subject line and follow-up message. Qwen 2.5 72B

OUTPUT

Ready-to-send leads

Buyer list with company, country, type, fit reason, contact format, and personalized outreach email per buyer.

Be specific with your product. "Premium handmade leather handbags" gives better matches than "bags".
Try different regions. Run separate searches for EU, North America, and Asia Pacific to build a full global pipeline.
Contacts are estimated. Email formats shown are generic (info@company.com). Always verify real contacts via LinkedIn or company websites.
📊
MODULE C

Analytics Dashboard

Live overview of all your suppliers, leads, and an AI daily briefing.

AUTO

Data populates automatically

Every supplier you analyze and every buyer you find is saved in memory during your session. The dashboard aggregates all of it.

SECTION 1

Stats overview

Total suppliers analyzed, how many are high risk (score > 50), and total leads found in this session.

SECTION 2

Supplier risk heatmap

All analyzed suppliers listed with their risk score and level — color coded green / yellow / orange / red.

SECTION 3

Top leads

Your top 5 found buyers with company name, country, and fit reason.

SECTION 4

AI Daily Briefing

AI generates a 3-priority business intelligence report based on your session data — what to do today, what to watch, and what opportunity to act on. Qwen 2.5 72B

Analyze at least 2–3 suppliers first before opening Dashboard — the briefing needs data to generate meaningful insights.
Session-based only. Data resets when the server restarts. The Nightly Audit Trail provides persistent tracking.
🔮
MODULE D

MarketPulse AI

Three-layer commodity market prediction — 24 hours to 10 years.

INPUT

Enter commodity and industry

Specify the raw material (e.g. "cotton"), your industry (e.g. "textile"), region, and country. All three prediction layers run simultaneously.

LAYER 1

Tomorrow Agent — 24–48 hours

Uses Tree-of-Thought reasoning to evaluate weather patterns, supply chain signals, and currency movements for short-term price prediction. Returns direction (UP/DOWN/STABLE), probability %, and price change estimate. DeepSeek-R1

LAYER 2

Trend Agent — 1–5 years

Analyzes industry demand growth, ESG regulations, technology disruption, and trade policy changes for medium-term outlook. Returns CAGR estimate, top opportunities, threats, and SME action items. Qwen 2.5 72B

LAYER 3

Mega Trend Agent — 5–10 years

Forecasts climate change impact, AI/automation disruption, geopolitical realignment, and demographic shifts specific to your country. Returns mega trends with probability scores and survival strategies. Mistral Large

BONUS

Micro-Econ Agent

See Micro-Econ tab for details on the HHI monopoly analysis and Utility Maximization purchasing optimization.

Layer 1 is most actionable. Use it before making large purchase decisions — e.g. whether to stock up on cotton this week or wait.
Full run takes 30–90 seconds. All three agents run in parallel — you can also run each layer individually for speed.
Predictions are probabilistic, not certain. The confidence score tells you how reliable the prediction is based on available signals.
🔐
MODULE A — NEW

Cyber Risk Agent

Free OSINT scan of any supplier's IT security posture — no API keys needed.

⚠️

Note: This scan uses only publicly available information (OSINT). No hacking or unauthorized access is performed. Results depend on what is publicly exposed. Takes 30–60 seconds.

INPUT

Enter company name

Optionally add the domain (e.g. evergrande.com). If no domain is given, the system guesses it from the company name.

CHECK 1

SSL Certificate scan

Checks if the supplier's website has a valid SSL certificate and how many days until expiry. Expired or missing SSL is a red flag.

CHECK 2

Open port scan

Scans 12 common ports (FTP, SSH, RDP, MySQL, MongoDB, Redis…). Exposed database or remote access ports are serious vulnerabilities.

CHECK 3

Data breach check

Queries HaveIBeenPwned free domain API to check if the supplier's domain has appeared in known data breaches.

CHECK 4

Malware scan

Queries urlscan.io free API to check if the domain has been flagged for malware, phishing, or suspicious activity.

OUTPUT

Cyber risk score + AI analysis

Score 0–100 + risk level + key vulnerabilities + red flags + attack vectors + recommendation. Qwen 2.5 72B

A supplier with exposed RDP or database ports is a supply chain attack vector — attackers can compromise their systems and reach you.
Run this alongside financial risk. A financially stable supplier with poor IT security is still a major liability.
📊
MODULE D — NEW

Micro-Econ Agent

HHI monopoly analysis + Utility Maximization for smarter purchasing decisions.

INPUT

Add 2+ suppliers

For each supplier enter: name, market share %, price per unit, quality score (0–10), and units needed. Set your total budget.

STEP 1

HHI Index calculation

Herfindahl-Hirschman Index = sum of squared market shares. Measures supplier market concentration.

HHI < 1,500 = Competitive market — you have bargaining power
1,500–2,500 = Moderately concentrated — negotiate carefully
> 2,500 = Monopoly risk — supplier has pricing power over you

STEP 2

Utility Maximization

Calculates utility per dollar (quality ÷ price) for each supplier and finds the optimal purchase allocation within your budget — maximizing value, not just minimizing cost.

OUTPUT

Purchase strategy

Market structure verdict + buyer bargaining power + recommended action (BUY NOW / WAIT / NEGOTIATE / DIVERSIFY) + optimal lock-in % + ranked supplier table. DeepSeek-R1

Use before large orders. If one supplier has > 60% market share, you have low bargaining power — the HHI will flag this and recommend negotiation tactics.
Quality score is subjective. Rate each supplier based on your own experience — 1 = very poor, 10 = exceptional.
🌙
AUTONOMOUS SYSTEM

Nightly Audit Trail

Fully autonomous supplier monitoring — runs every night at midnight with zero human input.

HOW

Vultr Cron Scheduler

A single cron job runs every night at 00:00 UTC:
0 0 * * * python nightly_audit_trail.py

No human trigger. No login required. Fully autonomous.

STEP 1

Loads all tracked suppliers

Every supplier you analyze is automatically added to the tracking list (data/tracked_suppliers.json). The audit re-checks all of them.

STEP 2

Risk re-scan

Calls the same risk analysis pipeline on each supplier — updating their score if anything has changed overnight.

STEP 3

Cyber re-scan

Runs OSINT cyber check on each supplier — detects newly exposed ports, expired SSL certificates, or new breach records.

STEP 4

Alerts generated

Any supplier scoring > 75 (Critical) or > 50 (High Risk) generates an alert. Cyber score > 70 generates a cyber alert.

OUTPUT

Summary saved to dashboard

Results saved to logs/last_audit_summary.json. View anytime in the Audit Trail tab — timestamp, duration, alerts, per-supplier scores.

Click "Run Audit Now" in the Audit Trail tab to trigger a manual run for demo purposes — results appear in ~65 seconds.
On Vultr VM: Set up the cron job once and it runs every night automatically — no maintenance needed.
Suppliers are auto-tracked. Every time you run a supplier risk analysis, that supplier is automatically added to the nightly audit list.

Ready to get started?

Launch the platform and run your first analysis in under 60 seconds.

Launch TradeNexus AI →