postak 0.1.0

AI chat bot that turns your Telegram channel into a chat list

Postak

Turn a Telegram channel into a personal AI chat list.

Start a chat with /new. Each post becomes a conversation title; its comment thread is the chat. Postak replies as you type and names the post after the first reply.

[!CAUTION] Postak is under active development. APIs, commands, and storage may change.

Features

• Channel as chat list — each /new post becomes an auto-titled conversation.
• Rich streaming replies — rich Telegram messages: tables, checklists, ...
• Per-thread memory — each conversation keeps its own context.
• multi-channel support · Custom prompts · Commenter mentions · Scoped access · Thread tools · Pluggable models and storage

How it works

1. Send /new in your channel — or in its discussion group — to start a conversation.
2. Telegram forwards the post into its discussion group as a comment thread.
3. Reply in the thread — Postak answers, streaming as it types.
4. The first reply auto-titles the post, so your feed reads like a chat list.

Requirements

• Python 3.10+
• A Telegram bot (@BotFather) and a channel with a linked discussion group
• An OpenAI-compatible LLM endpoint

Start in five minutes

1. In Telegram, create a channel and enable comments by linking a discussion group. Add your bot as an admin of both.

[!IMPORTANT] Turn the bot's privacy mode off so it can see comments: in @BotFather run /setprivacy → select your bot → Disable. If the bot was already in the group, remove and re-add it after changing this.

2. Configure and run:

uv sync
cp .env.example .env    # fill in BOT_TOKEN, TARGET_CHANNEL_ID, LLM_MODEL, …
uv run postak

# or, from PyPI:
#   pip install postak
#   postak    # reads .env from the current directory

3. Send /new in your channel or discussion group, then reply in the new thread.

[!TIP] To serve another channel later, link its discussion group in Telegram, then either run /postak add in that group or just send /new there — no restart needed. See Admin commands.

Configuration

Postak reads settings from the environment (via a .env file):

Variable	Required	Default	Description
BOT_TOKEN	✔	—	Telegram bot token from @BotFather
TARGET_CHANNEL_ID		—	Channel id(s) to serve, comma-separated (e.g. -1001234567890,-1009876543210). Leave empty and a Postak admin can add channels later — see Admin commands
LLM_MODEL	✔	—	Model name, e.g. gpt-4o-mini
LLM_ENDPOINT		OpenAI API	Base URL of an OpenAI-compatible endpoint
LLM_API_KEY		not-needed	API key for the LLM endpoint
POSTAK_ADMINS		—	Comma-separated Telegram user ids who may manage Postak. Required if TARGET_CHANNEL_ID is empty, since that's the only way to add a channel
DATABASE_URL		sqlite+aiosqlite:///postak.db	Store URL: a SQLite path, or :memory:
HISTORY_WINDOW		20	Max recent messages kept as context (the system prompt is always kept)

Library usage

from aiogram import Bot, Dispatcher

from postak import OpenAIGenerator, Postak

postak = Postak(
    generator=OpenAIGenerator(model="gpt-4o-mini", api_key="..."),
    store="sqlite+aiosqlite:///postak.db",
    channels=[-1001234567890],
    admins=[123456789],
)

bot = Bot("telegram-token")
dp = Dispatcher()

postak.attach(dp)
dp.run_polling(bot)

Admin commands

Postak admins manage everything with /postak (/postak help lists it all):

Access

• admin list — show admins · admin add|remove <user_id> — grant or revoke rights
• access list — show all access rules
• access allow|revoke <user_id> global|group|thread — per-user access
• access public on|off global|group|thread — open or close a scope to everyone; a public/allowed group (or global scope) also lets non-admins run /new in an already-linked group, not just comment

Channels

• add [chat_id] — link a channel, discovering its discussion group via Telegram; defaults to the chat you're in, or pass either side's id
• remove [chat_id] — unlink it and drop that group's access rules

A Postak admin can skip add entirely: sending /new in a group Postak doesn't know yet links its channel and starts the conversation in one step.

In a thread

• digest — summarize the conversation
• compress — replace history with a compact summary
• title — regenerate the post title · settitle <text> — set it yourself
• regenerate — redo the last answer
• delete — delete the message you replied to

Model

• model get — show the current model · model set <model> — switch it

System prompt — targets the surrounding thread when run inside one, else the global default

• sysprompt — show it · sysprompt <text> — set it · sysprompt delete — reset it
• The global prompt applies to conversations opened after the change and is admin-only; a thread's prompt takes effect on its next reply and can also be managed by anyone allowed to chat in that thread

Development

uv sync
uv run python -m unittest discover -s tests
uv run ruff check .
uv run mypy postak

License

MIT