One conversation that operates the workspace.

Summary

The assistant uses Telegram as one interface. A master agent, powered by an OpenAI model and memory, analyzes the request and delegates it to a Gmail, Calendar, Sheets or Docs manager.

Status

Functional prototype in a development environment.

Type

Personal project; no client claimed.

Evidence

Visible architecture and successful n8n execution on August 20, 2026.

Public demo

No public link supplied.

Architecture

Telegram message
      ↓
Master agent ← OpenAI Chat Model + memory
      ├── Gmail Manager
      ├── Calendar Manager
      ├── Sheet Manager
      └── Docs Manager
      ↓
Telegram response

The workflow uses a multi-agent architecture: the master agent routes requests while specialized managers expose service-specific actions.

Gmail manager

• Send a message.
• Reply to a message.
• Retrieve multiple messages.
• Delete a message.
• Add a label.

Google Calendar manager

• Retrieve multiple events.
• Create an event.
• Delete an event.
• Update an event.

Google Sheets manager

• Read rows from a sheet.
• Delete rows or columns.
• Delete or create a sheet.
• Add a row.
• Add or update a row.

Google Docs manager

• Create a document.
• Retrieve a document.
• Update a document.

Evidence level

Visible

Telegram trigger and response, master agent, OpenAI Chat Model, memories and four Google Workspace managers.

Execution

One screenshot explicitly displays “Workflow executed successfully”.

Not demonstrated

Each action in a real scenario, continuous availability, production deployment and large-scale performance.

Limits and next steps

• The current README provides neither a sanitized n8n export nor a public code repository.
• No public bot or demonstration environment is available.
• Permissions, errors, confirmation before sensitive actions and escalation rules remain to be documented.
• Action-by-action test scenarios would strengthen the public evidence.