AI Assistant with RAG
Configurable AI assistant that answers from each organization’s own knowledge using a RAG architecture.
LLMs · RAG · Pinecone · n8n · JavaScript · APIs
Overview
I built an AI assistant that can be embedded on websites and configured per client. It holds a conversation with visitors and, when the question needs company-specific knowledge, retrieves context from a vector store (RAG) before answering. The same base is reused: knowledge and behavior are configured for each implementation.
The challenge
Generic LLMs do not know a company’s products, docs, or processes. Answers sound fluent but miss the actual business context.
The approach
I implemented a reusable RAG pipeline: conversation, retrieval from a vector store, and grounded responses — configurable per client and embeddable on the web.
Key capabilities
- Chat embedded on the website
- Conversations with an LLM
- Company-knowledge retrieval
- RAG architecture and vector store
- API automations and integrations
- Per-client configuration
- Context management
How it works
The product
Technology
- LLMs
- RAG
- Pinecone
- n8n
- JavaScript
- APIs
Highlights
- Answers from the organization’s own knowledge, not only the model’s general training.
- The same core is reused by configuring knowledge and behavior per client.
More selected work
Need something in this direction?
I take a limited number of freelance engagements on architecture, SaaS, integrations, automation, and applied AI. Tell me the problem — I will tell you if I am a fit.