Loading image 1...
WhatsAppDailySummary
n8n Automation That Summarises WhatsApp Group Conversations Nightly, Released Open Source
WhatsApp Daily Summary
WhatsApp Daily Summary
n8n Automation That Summarises WhatsApp Group Conversations Nightly, Released Open Source
n8n Automation That Summarises WhatsApp Group Conversations Nightly, Released Open Source
An in-house Javapixa tool that was subsequently released as open source. The n8n automation collects a WhatsApp group's messages through the day via the go-whatsapp-web-multidevice gateway, stores them in PostgreSQL, then each night has Google Gemini produce a structured summary of topics discussed, action items and statistics, and posts it back to a nominated WhatsApp number. One workflow handles many groups at once.
The Problem
Summarising a conversation with a language model is easy to run and hard to trust. The riskiest part is the numbers: message counts and who was active are exactly what a model is most likely to invent, and one wrong figure discredits the whole summary. The second problem is resilience, since one failing group must not take the other groups' summaries down with it.
Statistics are computed from the database rather than asked of the model, so message counts and active members are always exact. Gemini is constrained by a JSON schema to return `topics[]` and `action_items[]`, and the message is rendered deterministically instead of parsed out of free text. Ingestion and the nightly summary are decoupled through PostgreSQL tables, with `message_id` unique so webhook retries never double-count. Each group is processed in isolation, and a dedicated error workflow reports failures over WhatsApp.
Visual Journey
Project Showcase
Loading image 1...
Screenshot 1
Project visualization and interface design
TechStack
n8n
Workflow Orchestration
Google Gemini
Language Model
PostgreSQL
Database
go-whatsapp-web-multidevice
WhatsApp Gateway
KeyFeatures
Exact Statistics from the Database
Message counts and active members are computed by query rather than asked of the model, so the figures can be trusted.
Structured Output from Gemini
Gemini is forced to return topics[] and action_items[] through a JSON schema, and the message is rendered locally, so the format is identical every night.
Many Groups at Once
One workflow summarises any number of groups, and adding a group is done through a form rather than by editing the workflow.
Deduplication
message_id is unique, so a repeated webhook delivery never counts the same message twice.
Fault Isolation per Group
A failure on one group does not halt the rest, and an Error Trigger workflow sends an alert over WhatsApp.
No-Code Installation
The whole thing is exported n8n workflow JSON imported into your own instance, so there is no application code and no build step.
OurProcess
A detailed look at how we brought this project from concept to reality through our proven methodology.
Designing the Data Flow
Separated real-time ingestion from the nightly summary, then designed the PostgreSQL tables that connect the two.
Summarisation and Model Constraints
Built the summary workflow with a JSON schema for Gemini and query-computed statistics, so output stays consistent and the numbers stay exact.
Resilience and Documentation
Added per-group fault isolation, the alert workflow and the Quick Setup form, then wrote the documentation and released it as open source.
05 Outcomes delivered
- 01Statistics are computed from PostgreSQL rather than generated, so message counts and active members are exact.
- 02Gemini is constrained by a JSON schema, so topics and action items render deterministically.
- 03Ingestion and summarisation are decoupled through PostgreSQL, with message_id unique against double-counting.
- 04One group failing does not stop the others, and an error workflow reports it over WhatsApp.
- 05Ships as exported n8n workflow JSON, so there is no application code to build or deploy.
Related Services
Services Featured in This Project
These are the Javapixa services behind this project.