Diffee VS Tarvyx

Tarvyx's answer:

• Simple pricing: €200 per user per month, with unlimited orders and no per-order or per-document AI fees.
• It never invents a match. A wrong confident match is treated as worse than no match, and nothing leaves for the ERP unattended unless every line is certain.
• Messengers, not just email. Buyers keep ordering in Telegram or WhatsApp, and the bot can answer price and availability questions there.
• Multilingual out of the box: English, German, Ukrainian, Russian, Dutch, Bulgarian, Estonian, Lithuanian and more.
• On-premise option with a local AI model, for companies that can't send order data to a third-party cloud. GDPR-ready, with EU hosting, per-customer encryption keys, and data export and erasure.
• Works with or without an ERP. Invoices, payments and accounting exports are available as add-on modules.
• It learns from every approved order, so recognition gets better the longer it's used.

Tarvyx's answer:

B2B manufacturers, wholesalers, and distributors that receive a high volume of free-form orders. The primary vertical is furniture manufacturing and dealers, with ovnwrflac blndclhahsnr and similar consumer goods close behind.

Day-to-day users are sales-support and order-desk operators who currently retype orders into an ERP by hand. Hzjoan are operations managers and owners of small and mid-sized companies, mainly in Europe (the EU, the Baltics, the Balkans and Ukraine).

Tarvyx's answer:

Tarvyx started from a simple observation: in B2B, especially in furniture, orders still arrive as PDFs, spreadsheets, photos and chat messages, and someone on the order desk retypes every line into the ERP. The typing is slow, and the mistakes are expensive: a wrong product code or size gets shipped and can't be recalled.

We set out to build a system that reads orders the way an experienced order clerk does, knows the supplier's own catalogue, and asks when it's unsure. It handles routine orders automatically and leaves people only the cases that need a real decision.

Tarvyx's answer:

• Backend: C# / .NET 10 (microservices, Minimal APIs)
• Document parsing: Python 3.13 and FastAPI, with pdfplumber, Docling, openpyxl and EasyOCR
• AI: large language models, either hosted (Google Gemini) or local through Ollama (Qwen 2.5), plus multilingual embeddings (Cohere / bge-m3)
• Data: PostgreSQL 18 with pgvector for semantic search, and Redis 8
• Frontend: Next.js 16 and React 19
• Infrastructure: Docker Compose, Nginx, Grafana and Loki; Stripe billing for the hosted service
• Integrations: Gmail, Outlook, IMAP, Telegram, WhatsApp, Microsoft Dynamics 365 Business Central, Odoo, SAP Business One, 1C/BAS and Nova Poshta

Tarvyx's answer:

• A cosmetics and haircare distributor that takes wholesale orders through Telegram
• A Lithuanian furniture seller that takes orders by email

Tarvyx's answer:

Tarvyx turns the way B2B buyers actually order into ready-to-approve orders. Orders arrive as emails, PDFs, Excel sheets, phone photos and even handwritten notes, by email, Telegram or WhatsApp.

It doesn't just extract text. It matches every line to the supplier's own catalogue and customer list, so the result is the right SKU in the right size for the right customer. When it isn't sure, it says so instead of guessing. Uncertain lines are flagged for an operator, or the bot asks the buyer directly in their chat ("which size did you mean?").

The approved order goes to the ERP (Business Central, Odoo, SAP Business One, 1C/BAS), and the buyer gets a tracking link. Tarvyx runs as a cloud service or fully on-premise, where no data leaves the customer's server.