Mission 02 / Retrieval

Standout4Growth RAG Chatbot

Workflow and public UI evidence

A dual-context assistant that combines personal-brand guidance with factual company retrieval and conversation memory.

DemonstratedLive on standout4growth.com: the assistant answers a company question ('who is Alice?') from a Supabase vector store while holding its brand-coach persona.

The problem

A single assistant needed to coach users conversationally while keeping company answers grounded in a controlled knowledge source.

My responsibility

Designed the n8n agent workflow, Supabase retrieval path, model connection, and memory behavior documented in the project portfolio.

Evidence

Screenshots from the live system and the workflow itself — captions describe only what is visible.

System route

The architecture below is a readable abstraction of the documented workflow, not a simulated live execution.

  1. Receive and classify a chat request
  2. Route factual questions through vector retrieval
  3. Combine retrieved context with the assistant role
  4. Preserve conversation continuity with memory

Reliability notes

  • Grounded retrieval separates factual company context from coaching behavior.
  • Evaluation results and production usage remain pending evidence review.

Evidence currently available

  • Agent architecture screenshot
  • Public chatbot interface

Technology used

  • n8n
  • Supabase
  • Gemini
  • LangChain
  • OpenRouter

Evidence still being recovered

Client-approved metrics, execution recordings, exact delivery dates, and sanitized implementation artifacts will be added only after verification. This dossier deliberately avoids unsupported impact claims.

← Return to mission archive