AI Product Assistant: smart consulting bot for e-commerce

AI consultant in Telegram and on the website: selection and recommendations of products.

Goal:
Automate the selection and recommendation of products for the online store saleway.org. The client wants managers and users to quickly receive professional consultation in a chat format (#Telegram-bot + #widget on the website).

Tasks:
Create a knowledge base for all products in the store (1000+ items).
Use a vector storage for searching by descriptions and characteristics.
Develop an #AI-consultant based on LLM (#Gemini / #OpenAI).

Set up the response algorithm:
- clear and professional recommendations;
- always provide a link to the website + product model;
- if there are doubts — clarify with the user (“I can suggest X or Y, please specify for whom exactly?”).

Functionality:
- Search and recommendation of products by key query (for example, “vitamins for joint pain”).
- Detailed product description with a link to the website.
- User survey for more accurate selection (by age, gender, symptoms).
- Automatic database update when new products are added.

Implementation:
#Telegram-bot for consultants and clients.
#Chat-widget on the website.

Technology stack:
#N8N — process orchestration.
#Supabase Vector Store — storage and search of vectorized descriptions.
#Gemini / #OpenAI (4o-mini) — #LLM for vectorization and consultations.
#Telegram API — integration with the messenger.

Result:
A fully automated #AI-assistant has been obtained, significantly reducing the workload on managers, improving the quality of consultations, and increasing conversion to orders.