gpt-friend — Telegram diary bot + LLM chat

A private Telegram bot for journaling (text/photos), quick summaries, and a “chat” mode. Stores data locally in SQLite.

Features

• Diary: text entries and photo entries (photos are saved to disk).
• Summaries: daily / weekly / monthly.
• Tags: #tags inside entry text.
• Privacy: allowlist by Telegram user_id (or open access if the list is empty).
• Web search (optional): enable/disable via WEB_SEARCH_ENABLED.
• UI language: Russian/English via BOT_LANG.

Requirements

• Python 3.10+ (recommended 3.11+)
• Telegram Bot API token
• OpenAI API key

Quick start

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

cp .env.example .env
# fill in .env

python3 bot.py

Configuration (.env)

The bot reads settings from environment variables and an optional .env file (do not commit secrets).

Required:

• TELEGRAM_TOKEN — Telegram bot token
• OPENAI_API_KEY — OpenAI API key
• ALLOWED_USER_IDS — Telegram user_id allowlist (comma/space-separated, e.g. 123,456). Empty => open access.

Optional:

• BOT_LANG — ru or en (default: ru)
• OPENAI_MODEL — default model (example: gpt-5.2)
• OPENAI_VISION_MODEL — model for photo understanding (default: OPENAI_MODEL)
• OPENAI_WEB_SEARCH_MODEL — model for web search (default: OPENAI_MODEL)
• WEB_SEARCH_ENABLED — 1/0
• BOT_DB_PATH — SQLite path (default: diary.db)
• PHOTO_DIR — photo directory (default: diary_photos)
• DIARY_PAGE_SIZE — diary list page size
• LOG_LEVEL — INFO, DEBUG, ...

To find your Telegram user_id, message @userinfobot (or similar).

Data

• Database: BOT_DB_PATH (default diary.db)
• Photos: PHOTO_DIR (default diary_photos/)

Set up regular backups for these files/folders.

Localization

All user-facing strings are centralized in texts.py. Switch UI language via BOT_LANG.

ETH: 0x5F911FA3045dfE3d8E7c23a916FE8f2Aa45f8bd7

Roadmap

Ideas and next steps live in ROADMAP.md.