Daily News Digest

_{Verified on AgentGraph — trust infrastructure for AI agents. View profile}

Daily News Digest

A Python pipeline that fetches news from 50+ RSS feeds, uses Claude AI to create a personalized summary, generates a two-host AI podcast, and delivers everything via email with an optional Audiobookshelf podcast link.

Features

50+ RSS sources across tech, AI, robotics, finance, fintech, crypto, legal/regulatory, science, health, climate, automotive, global news, US news, LA local, and more
Claude AI summarization — filters hundreds of articles down to the best 20-30, organized by topic
Two-host AI podcast — generates a natural conversation between "Alex" and "Sam" from the digest
Multi-voice TTS — ElevenLabs (premium) with automatic Edge-TTS fallback
Audiobookshelf integration — auto-uploads podcast episodes for streaming
Duplicate detection — 7-day article history prevents repeating content
Error notifications — emails you if the script fails, API credits run out, or no new articles are found
Graceful degradation — podcast failures don't block email delivery; TTS failures fall back silently

Architecture

RSS Feeds (50+)
      │
      ▼
 news_digest.py ──── Claude AI ──── HTML Email
      │
      ▼
 podcast_generator.py ──── Local LLM (Ollama)
      │
      ▼
 audio_generator.py ──── ElevenLabs / Edge-TTS
      │
      ▼
 audiobookshelf_client.py ──── Audiobookshelf library scan

File	Purpose
`news_digest.py`	Main entry point — fetches RSS, calls Claude, sends email, orchestrates podcast pipeline
`podcast_generator.py`	Generates a two-host podcast script via local LLM (Ollama/LM Studio)
`audio_generator.py`	Converts script segments to multi-voice MP3 with pydub assembly
`audiobookshelf_client.py`	Triggers Audiobookshelf library scan and provides podcast URL for email

Quick Start

1. Install dependencies

pip install -r requirements.txt

System requirement: ffmpeg must be on your PATH (required for MP3 encoding).

2. Configure environment variables

cp .env.example .env

Edit .env with your credentials. At minimum you need:

Variable	Description
`ANTHROPIC_API_KEY`	Get from https://console.anthropic.com/
`GMAIL_ADDRESS`	Your Gmail address (sender)
`GMAIL_APP_PASSWORD`	16-char app password (Google Account > Security > 2-Step Verification > App passwords)
`RECIPIENT_EMAIL`	Comma-separated recipient list

3. Run

python news_digest.py

This will fetch news, generate the digest email, and (if podcast is configured) generate and upload a podcast episode.

Podcast Setup (Optional)

The podcast pipeline requires a local LLM and a TTS engine. If AUDIO_OUTPUT_DIR is not set, the podcast step is skipped entirely.

Local LLM (Ollama)

Install Ollama and pull a model:

ollama pull qwen2.5:14b

Audiobookshelf (Optional)

Run Audiobookshelf via Docker to host podcast episodes:

docker run -d --name audiobookshelf \
  -p 13378:80 \
  -v /path/to/audio:/podcasts \
  -v /path/to/config:/config \
  -v /path/to/metadata:/metadata \
  --restart unless-stopped \
  ghcr.io/advplyr/audiobookshelf

Podcast environment variables

Variable	Default	Description
`AUDIO_OUTPUT_DIR`	(none)	Directory for generated MP3s — setting this enables the podcast pipeline
`LOCAL_LLM_URL`	`http://localhost:11434`	Ollama API endpoint
`LOCAL_LLM_MODEL`	`qwen3.5:9b`	Model for script generation
`AUDIOBOOKSHELF_URL`	`http://localhost:13378`	Audiobookshelf instance
`AUDIOBOOKSHELF_API_KEY`	(none)	API token from Audiobookshelf settings
`AUDIOBOOKSHELF_LIBRARY_ID`	(none)	Library UUID to scan
`ELEVENLABS_API_KEY`	(none)	Premium TTS (falls back to Edge-TTS if unset)
`ELEVENLABS_VOICE_ALEX`	(auto-rotate)	Pin a specific ElevenLabs voice for Alex
`ELEVENLABS_VOICE_SAM`	(auto-rotate)	Pin a specific ElevenLabs voice for Sam
`ELEVENLABS_MODEL`	`eleven_multilingual_v2`	ElevenLabs model
`INTRO_MUSIC_PATH`	(none)	MP3 file for intro music with fade-in
`OUTRO_MUSIC_PATH`	(none)	MP3 file for outro music with fade-out
`PODCAST_TEST_MODE`	`false`	Generate a ~2-minute sample instead of full episode

Audio details

Output format: 128kbps CBR mono 44.1kHz MP3
Filename pattern: digest-YYYY-MM-DD.mp3
300ms silence between speaker changes
Daily voice rotation: same date always produces the same voice pairing
Old audio files auto-cleaned after 10 days

Scheduling

macOS/Linux (cron)

crontab -e

Add:

0 8 * * * cd /path/to/news-digest && /usr/bin/python3 news_digest.py >> digest.log 2>&1

Optional log rotation (Linux/macOS):

1. Create `news-digest-logrotate.conf`:

```conf
/path/to/news-digest/digest.log {
  daily
  rotate 30
  compress
  missingok
  notifempty
  copytruncate
}

Run it daily:

0 7 * * * /usr/sbin/logrotate /path/to/news-digest/news-digest-logrotate.conf

The script also deletes rotated digest.log.* files older than LOG_RETENTION_DAYS (default 30).


Verify:
```bash
crontab -l

Note: On macOS, cron may require Full Disk Access in System Preferences > Privacy.

Windows (Task Scheduler)

Option A: GUI

Win + R > taskschd.msc
Create Basic Task > Name: Daily News Digest
Trigger: Daily at 8:00 AM
Action: Start a program
- Program: python
- Arguments: news_digest.py
- Start in: C:\path\to\news-digest
In Settings, enable "Run task as soon as possible after a scheduled start is missed"

Option B: PowerShell

$action = New-ScheduledTaskAction -Execute "python" -Argument "news_digest.py" -WorkingDirectory "C:\path\to\news-digest"
$trigger = New-ScheduledTaskTrigger -Daily -At 8:00AM
$settings = New-ScheduledTaskSettingsSet -StartWhenAvailable -RestartCount 3 -RestartInterval (New-TimeSpan -Hours 1)
Register-ScheduledTask -TaskName "DailyNewsDigest" -Action $action -Trigger $trigger -Settings $settings -Description "Fetches news and emails daily digest"

RSS Sources

Category	Sources
Tech & AI	Hacker News, TechCrunch, TechCrunch AI, Ars Technica, The Verge, MIT Tech Review, VentureBeat AI, The Information
Robotics	IEEE Spectrum Robotics, The Robot Report
Automotive & EVs	Electrek, InsideEVs, The Drive
Social	Platformer
Web3	The Block, Decrypt, CoinDesk
Fintech & Crypto Industry	Finextra, CoinTelegraph, TechCrunch Fintech, Crunchbase News
Legal & Regulatory	Reuters Legal, The Register, Rest of World
Health & Longevity	STAT News, Longevity Technology
Climate	Canary Media, CleanTechnica
Major News	BBC News, Reuters, NPR, AP News, Al Jazeera
US News	Politico, The Hill, USA Today
Los Angeles Local	LA Times, LAist
Finance	Bloomberg, Financial Times, MarketWatch
Science & Space	Science Daily, Phys.org, Nature, NASA, Space.com
Entertainment	The Hollywood Reporter

Error Notifications

The script automatically emails you when something goes wrong:

Error Type	What It Means
API Credits Exhausted	Your Anthropic account has run out of credits
API Rate Limit Exceeded	Too many requests; wait or upgrade your plan
API Authentication Failed	Your API key is invalid or revoked
Email Sending Failed	Gmail credentials are wrong or app password expired
No New Articles Found	All articles were already sent previously (slow news day or feed issue)
Unexpected Error	Something else went wrong (full traceback included)

Customization

Modify your interests

Edit the INTERESTS variable in news_digest.py to change what topics Claude prioritizes. Interests are organized by priority tier (Priority > High > Moderate) with strict exclusion filters.

Add/remove news sources

Edit the RSS_FEEDS dictionary in news_digest.py.

Change article limits

Set MAX_ARTICLES_PER_SOURCE in .env (default: 20).

Use a different Claude model

Set DIGEST_MODEL in .env (default: claude-sonnet-4-20250514).

Troubleshooting

Gmail "Less secure app" errors

Make sure 2-Step Verification is enabled on your Google account
Use an App Password, not your regular password
App passwords are 16 characters with no spaces

Cron job not running (macOS/Linux)

Check the log file: tail -f digest.log
Ensure Python path is correct: which python3
macOS may require Full Disk Access for cron in System Preferences > Privacy

Task Scheduler not running (Windows)

Open Task Scheduler and check the task's History tab for errors
Ensure "Run whether user is logged on or not" is set if needed
Verify the Python path: where python in Command Prompt

RSS feed errors

Some feeds may be rate-limited or require different parsing
Check console output for specific feed errors

Podcast not generating

Verify Ollama is running: curl http://localhost:11434/v1/models
Ensure model is installed: ollama pull qwen2.5:14b
Check that ffmpeg is on your PATH: ffmpeg -version
Verify AUDIO_OUTPUT_DIR exists and is writable

API credit issues

Check your usage at https://console.anthropic.com/
The script will email you when credits run low or are exhausted

Name		Name	Last commit message	Last commit date
Latest commit History 37 Commits
.env.example		.env.example
.gitignore		.gitignore
README.md		README.md
audio_generator.py		audio_generator.py
audiobookshelf_client.py		audiobookshelf_client.py
news_digest.py		news_digest.py
podcast_generator.py		podcast_generator.py
requirements.txt		requirements.txt

Folders and files

Latest commit

History

Repository files navigation

Daily News Digest

Features

Architecture

Quick Start

1. Install dependencies

2. Configure environment variables

3. Run

Podcast Setup (Optional)

Local LLM (Ollama)

Audiobookshelf (Optional)

Podcast environment variables

Audio details

Scheduling

macOS/Linux (cron)

Windows (Task Scheduler)

RSS Sources

Error Notifications

Customization

Modify your interests

Add/remove news sources

Change article limits

Use a different Claude model

Troubleshooting

Gmail "Less secure app" errors

Cron job not running (macOS/Linux)

Task Scheduler not running (Windows)

RSS feed errors

Podcast not generating

API credit issues

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages