EngineeringNewsBot

Telegram bot built with Scala 3 and ZIO.

It polls engineering RSS and Atom feeds, stores state in local JSON files, publishes new posts to Telegram, and can forward published posts into NotebookLM through Telegram commands.

Project Layout

• scala-bot/ - main Scala application
• scala-bot/src/main/scala/bot/ - feed fetching, state management, Telegram publishing
• scala-bot/src/main/scala/bot/notebooklm/ - NotebookLM command handling and CLI integration
• flake.nix - Nix development shell and helper commands

Required Environment

• TELEGRAM_BOT_TOKEN

Optional:

• TELEGRAM_CHAT_ID

If TELEGRAM_CHAT_ID is not set, send the bot any private message and it will register that chat automatically.

NotebookLM Integration

NotebookLM support uses the external notebooklm CLI from the unofficial notebooklm-py project.

Recommended setup with Nix:

nix develop
setup-notebooklm
export NOTEBOOKLM_COMMAND="$PWD/scala-bot/.venv-notebooklm/bin/notebooklm"
export PATH="$PWD/scala-bot/.venv-notebooklm/bin:$PATH"
export PLAYWRIGHT_BROWSERS_PATH="$PWD/scala-bot/.playwright-browsers"
notebooklm login

This keeps the Python package and Playwright browser assets inside the project directory while system dependencies come from the Nix shell.

Common Telegram commands:

/nb create <title>
/nb notebooks
/nb switch <id|title>
/nb remember <id> [title]
/nb add https://example.com/post
/nb add
/nb ask <question>
/nb summary
/nb history
/nb sources
/nb report [--lang en] [--format study-guide] [--length ...] [--style ...] [instructions]
/nb artifacts
/nb artifact <id>
/nb forget
/clearnotes

Main reply flow:

1. Let the bot publish a post.
2. Reply to that Telegram message with /nb add.
3. The bot sends the original article URL to your active NotebookLM notebook.

/clearnotes removes tracked NotebookLM commands and replies from the private chat without deleting published blog posts.

Example flow:

Published engineering post in Telegram:

Reply-based NotebookLM workflow in Telegram:

Generated NotebookLM study guide:

Report flags:

• --lang or --language sets the NotebookLM output language
• --format briefing-doc|study-guide|blog-post|custom selects the report format
• --length short|medium|long appends a length hint
• --style executive|technical|academic|brief appends a style hint unless it maps to a real NotebookLM report format

Examples:

/nb report --lang en --format study-guide
/nb report --lang en --length long --style executive Focus on risks and trade-offs

Run With Nix

nix develop
cd scala-bot
export TELEGRAM_BOT_TOKEN=...
export NOTEBOOKLM_COMMAND="$PWD/.venv-notebooklm/bin/notebooklm"
sbt run

Run With Script

cd scala-bot
export TELEGRAM_BOT_TOKEN=...
./run.sh

If scala-bot/.env exists, run.sh loads it automatically.

Optional Variables

export STATE_FILE=./data/state.json
export NOTEBOOK_STATE_FILE=./data/notebook_state.json
export FETCH_INTERVAL_MINUTES=30
export PUBLISH_INTERVAL_SECONDS=30
export POSTS_PER_PUBLISH_CYCLE=1
export MAX_POSTS_PER_SOURCE=1
export POST_DESCRIPTION_LENGTH=1200
export TELEGRAM_UPDATES_TIMEOUT_SECONDS=25
export PUBLISHED_TELEGRAM_MESSAGE_LIMIT=500
export NOTEBOOKLM_COMMAND=notebooklm

State Files

The bot stores:

• known post GUIDs
• unpublished post queue
• published post counter
• registered private chat id
• Telegram getUpdates offset
• recent Telegram message_id -> blog post link mappings
• per-user NotebookLM state in NOTEBOOK_STATE_FILE

Default state path: scala-bot/data/state.json.

Delivery Defaults

Default behavior is intentionally conservative:

• MAX_POSTS_PER_SOURCE=1
• POSTS_PER_PUBLISH_CYCLE=1
• PUBLISH_INTERVAL_SECONDS=30

Telegram limits still apply:

• text message limit: 4096 characters
• photo caption limit: 1024 characters