Skip to content

RELEASE_NOTES_V4.md

Latest commit

 

History

History
72 lines (56 loc) · 2.97 KB

RELEASE_NOTES_V4.md

File metadata and controls

72 lines (56 loc) · 2.97 KB

Claude V4.0.0 — Whisper Service + Cursor Workflow

Release Date: February 17, 2026 Upgrade from v3.3.0: Drop-in replacement. New whisper-service/ directory added.

What's New

Whisper Shared Service (whisper-service/)

Speech-to-text is now a Docker container instead of running inside the Telegram bot process.

  • Flask API on port 9000 — POST /transcribe accepts any audio file, returns text + segments + timing
  • Supports: English, Hebrew (he), auto-detect, translation to English
  • Pre-downloads model at build time — container starts in seconds
  • Resource-limited: 2GB RAM, 4 CPUs max — won't starve other services
  • Any app can use it: Telegram bot, orchestrator agents, cron scripts, future web UIs

The Telegram bot was updated to call http://127.0.0.1:9000/transcribe instead of loading Whisper in-process, saving ~700MB RAM.

Client-Side Setup (Cursor + Claude Code)

The infrastructure setup prompt now includes a complete guide for the client-side setup:

  • Cursor IDE ($20/month) — VS Code fork with AI. Connect via Remote SSH, browse files on the mini PC, make inline AI-assisted edits with Ctrl+K
  • Claude Code CLI — runs in Cursor's built-in terminal for big builds
  • Three interfaces documented: Telegram (mobile), Claude Code (heavy lifting), Cursor (visual)

Architecture After V4

Your Phone                Your Laptop
    │                         │
    │ Telegram                │ SSH (Cursor + Claude Code CLI)
    │                         │
    └──────────┬──────────────┘
               │
        Mini PC (Ubuntu, always-on)
               │
    ┌──────────┴──────────────────┐
    │  Orchestrator (port 8000)   │
    │  Telegram Bot (port 8787)   │
    │  Whisper Service (port 9000)│
    │  Docker containers          │
    │  PostgreSQL, Redis, Gitea   │
    └─────────────────────────────┘

Files Changed from V3.3.0

Added:
  whisper-service/              — NEW: Shared speech-to-text Docker service
  whisper-service/app.py        — Flask API server
  whisper-service/Dockerfile    — Container with pre-downloaded Whisper model
  whisper-service/docker-compose.yml
  whisper-service/requirements.txt
  whisper-service/README.md
  RELEASE_NOTES_V4.md         — This file

Modified:
  VERSION                       — 4.0.0
  CHANGELOG.md                  — V4 entry
  START_HERE.md                 — Added whisper-service to structure, version bump
  telegram-bot/bot.py           — Uses Whisper API service instead of in-process model
  telegram-bot/requirements.txt — Removed openai-whisper dependency
  telegram-bot/.env.example     — WHISPER_API_URL replaces WHISPER_MODEL
  MINI_PC_SETUP_PROMPT.md       — Phase 4.5 Whisper deploy step, Client-Side Setup section