🔬 Compare personal grocery inflation to US CPI with Gemini and Telegram
⚡ 0 views · 🔬 Document Extraction & Analysis
Description
Quick overview
This workflow accepts grocery receipt photos via an n8n form or Telegram, uses Google Gemini to extract and normalize line items, calculates a personal month-to-month inflation rate, pulls the matching official CPI series from the US BLS API, and returns a report (optionally with an ElevenLabs voice note).
How it works
- Receives receipt photos or commands from an n8n Form trigger or a Telegram bot trigger.
- For uploaded images, uses Google Gemini to extract the store, purchase date, and purchasable line items into structured data.
- Checks previously saved receipt lines in an n8n Data Table to reuse known product mappings, and uses a Google Gemini agent to assign consistent product keys to any new line-item texts.
- Calculates unit prices, saves/upserts the normalized receipt lines to the Data Table, and fetches the official CPI time series from the US BLS Public Data API for the configured series.
- Computes a spending-weighted personal inflation rate between the two most recent months with receipts and compares it to the official index change for the same months.
- Uses Google Gemini to write an “economist’s note” from the precomputed results and optionally generates an audio version via the ElevenLabs text-to-speech API.
- Builds a formatted HTML report for form submissions or a Telegram message (plus optional voice note) and sends the response back to the originating channel.
Setup
- Add a Google Gemini (Google AI Studio) credential for the Gemini chat model used to read receipts and write the report.
- (Optional) Add an ElevenLabs HTTP Header Auth credential (xi-api-key) and set the voice/model settings if you want audio notes.
- (Optional) Add Telegram bot credentials (BotFather token) and configure the Telegram trigger/send nodes if you want Telegram intake and replies.
- Review and adjust the settings values (BLS series ID, label, currency symbol, fallback official rate, narrator prompt, and ElevenLabs voice/model IDs) to match your use case.
- Run the workflow once to let it create the receipt_economist_lines n8n Data Table automatically, then test with the sample household or upload receipts from at least two different months.
Requirements
- Gemini API key; n8n with Data Tables (the table is created on the first run). ElevenLabs and Telegram are optional.
Customization
- In Settings, change the BLS series, currency or narrator voice. For another country, point the official-index HTTP Request node at its statistics office and adjust the Calculate node.
🔗 Nodes Used
HTTP Request, Telegram, Telegram Trigger, AI Agent, Basic LLM Chain, Structured Output Parser
📥 Import
Download workflow.json and import into n8n:
Workflow menu → Import from File