Inbox Point is a self-hosted service that brings customer conversations from Telegram and VK into a private Telegram workspace for operators. Customers stay in their messenger; your team answers in Telegram.
- A customer writes to your Telegram bot or VK community.
- Inbox Point answers configured questions about schedules, prices, addresses, FAQs, or custom sections.
- A question needing a person opens an operator request in the customer's Telegram Topic, creating the topic on first contact.
- An operator replies in that topic.
- Inbox Point sends the answer back to the customer's original conversation.
Operators close finished requests. Later conversations open a new request and reuse the customer's topic when available.
- Text conversations from Telegram bots and VK communities.
- Configurable schedules, prices, addresses, FAQs, and custom menu sections.
- A protected emergency web inbox when Telegram is unavailable.
- Saved outbound messages with retries and controls for delivery incidents.
- Channel setup, content editing, preview, and revision history.
- Service monitoring, verified snapshots, and external backups.
Attachments, voice messages, AI-generated replies, CRM entities, operator assignment, multi-tenant SaaS, and billing are outside the current release. Unsupported attachments prompt the customer to resend their question as text.
- Durable delivery. Replies are saved in SQLite before sending and retried within set limits; pending work survives a restart.
- Uncertain outcomes. A confirmed failure differs from a lost confirmation: when Telegram may have accepted an action, the service asks for resolution instead of blindly repeating it.
- Duplicate protection. Stored event IDs and operation keys prevent duplicate requests and side effects. Repeatedly failing VK events are isolated for review.
- Emergency fallback. An active request can move to the protected web inbox. A request uses one workspace at a time.
- Verified recovery. Snapshots are checked before restore. Restored admin sessions are invalidated; devices must sign in again.
- Useful monitoring. Readiness checks channel and delivery-worker activity, queue delays, and unresolved inbound or operator-relay incidents.
Actual Russian-language UI with synthetic data: fictional North Side Dance content above; a failed VK reply, retry action, and service state below.
Reproduce the screenshots without real accounts.
TypeScript · Node.js · Fastify · SQLite · Vue 3 · Playwright
Requires Node.js 24.20.x and npm 11.
npm ci
npm run devAt http://127.0.0.1:3000, use /setup for channels, /manage for content,
and /ops for monitoring and emergency requests. Local loopback access needs no
password. Set ADMIN_PASSWORD for /login and optional remembered sessions;
production requires it. dev and start load an untracked .env;
see .env.example.
Run npm run check for formatting, lint, types, tests, and builds. For browser
checks, run npx playwright install chromium, then npm run check:e2e.
- Deployment — one isolated instance per organization.
- Administration and channels — authentication, connections, and content behavior.
- Monitoring and backups / restore.
- Acceptance and failure drills.
Core flows and reliability mechanisms are implemented and covered by automated checks. Production acceptance still requires verification on the target deployment: real channels, restart and failure drills, restore, rollback, and the operator workflow.