An n8n automation that researches real sources with SerpAPI, uses Groq for structured synthesis, accounts for Google Calendar holidays, persists plans in Data Tables and delivers missions through Telegram. Deterministic rules control idempotency, grounding, dates, states and feedback.
- n8n
- JavaScript
- Groq
- SerpAPI
- Google Calendar
- Telegram
- Docker
Context and problem #
Daily learning requires choosing useful topics, finding trustworthy sources, fitting activities into a schedule and preserving history. The project automates that cycle without delegating critical rules to the model.
My contribution #
The repository combines data modelling, deterministic generation of four n8n workflows, reusable JavaScript logic, OAuth and HTTP integrations, tests, operational documentation and secret scanning.
Architecture #
The weekly planner researches and validates sources before materializing topics and missions. A daily workflow delivers the eligible mission; another resolves callbacks and updates only the addressed item. Five Data Tables store categories, topics, plans, missions and destinations.
- ✓ 4 workflows / 149 nodes
- ✓ 5 Data Tables / 48 fields
- ✓ Grounding by exact source URL
- ✓ Idempotent state transitions
Evaluation and quality #
The suite records 25 functional tests and 1,021 structural and security checks. Groq, SerpAPI, Calendar and Telegram's /start command were validated against real services.
Limitations #
The corrected Planner still needs a real end-to-end retest, followed by the Dispatcher and a mission callback. Production schedules therefore remain inactive.
Next steps #
Reimport the corrected exports, prove idempotency across two runs, validate real delivery and feedback, and only then enable schedules.
Test the product and inspect the code
The live demo runs the core flow with safe local data. The public repository preserves the complete backend, tests, infrastructure and technical documentation.