Building AI Financial Intelligence MVP Built with n8n and Supabase

Built with n8n
, ,
1

Hello everyone :waving_hand:t5:

I’m stanley Efemdi, founder of Elixacore technology, an AI financial-intelligence platform being developed for the African market.

The current MVP uses n8n, Supabase, Telegram and AI. It can receive transaction data through an API, validate transactions, detect duplicates, save records to Supabase, categorise spending and answer natural-language questions about expenses.

I joined the n8n community to improve the reliability, security and scalability of the product, and to connect with experienced builders working in fintech, automation and AI.

I would especially appreciate constructive feedback on:

  • secure multi-user workflow design
  • error handling and retry logic
  • scaling webhook-based transaction ingestion
  • safely connecting Telegram to a central transaction API
  • preparing n8n workflows for controlled pilot testing

The project is already working at MVP level, and I am now focused on strengthening the architecture before expanding further.

1 « J'aime »
2

Welcome @Efemdi-Elixacore! The use case is solid - transaction deduplication + natural-language queries on expenses is a real gap in African fintech tooling. A few concrete points on what you’re asking:

  • For scaling webhook ingestion: switch to n8n queue mode (Redis + workers) so your webhook collector and transaction processor run on separate processes. This prevents a heavy AI query from blocking incoming webhooks.
  • For error handling + retry: use a global Error Workflow with a Supabase-backed retry queue. On failure, write the payload to a table with status=pending, then a scheduled workflow retries those rows until they succeed or hit a max retry count.
  • For Telegram-to-API security: store the bot token in n8n credentials only, and add an IF node that validates the incoming chat_id against an allowlist in Supabase before processing any command.
1 « J'aime »