AI News Blog Workflow

A scheduled Python workflow that fetches AI news from RSS feeds, summarizes it with a local Ollama model, and publishes the result as a post to your Google Blogger blog.

Requirements

• Python 3.10+
• Ollama installed and running locally (with at least one model pulled, e.g. ollama pull llama3.2)
• A Blogger blog and Google Cloud project with the Blogger API enabled

Setup

1. Clone or create the project and install dependencies

cd ai-news-blog-workflow
python3 -m venv .venv
source .venv/bin/activate   # or .venv\Scripts\activate on Windows
pip install -r requirements.txt

If you don’t have a config.yaml yet, copy the example and edit it:

cp config.yaml.example config.yaml

2. Configure RSS feeds and options

Edit config.yaml:

• feeds: List of RSS/Atom feed URLs. You can use AI/tech-focused feeds (e.g. TechCrunch AI, VentureBeat, The Verge AI) or any other sources—e.g. popular HN blogs OPML (use the xmlUrl values as feed URLs).
• ollama: Set model to an Ollama model you have pulled (e.g. llama3.2, gemma3:latest). Optionally set host if Ollama runs elsewhere. Use timeout_seconds (e.g. 300) to avoid hanging on slow responses. Use prompt_persona and prompt_requirements to tune the summarizer’s tone and output format.
• blogger: Set blog_id (see below). Paths to credentials.json and token.json default to the project directory.

3. Google Cloud and Blogger API

1. Go to Google Cloud Console and create or select a project.
2. Enable the Blogger API: APIs & Services → Enable APIs and Services → search for “Blogger API” → Enable.
3. Create OAuth 2.0 credentials:
   • APIs & Services → Credentials → Create Credentials → OAuth client ID.
   • If prompted, configure the OAuth consent screen (e.g. External, add your email as a test user).
   • Application type: Desktop app.
   • Download the JSON and save it as credentials.json in this project directory. Do not commit this file.

4. Get your Blogger Blog ID

• In Blogger, open your blog and go to Settings → Other. Under “Blog tools” you may see the Blog ID.
• Or: open your blog’s URL in the browser and view the page source; search for blogID or similar.
• Or: use the Blogger API “list by user” to list your blogs and their IDs.

Put this value in config.yaml under blogger.blog_id (or set the BLOGGER_BLOG_ID environment variable).

5. One-time OAuth (create token.json)

Run the auth helper once. It will open a browser for you to sign in and grant access; then it saves a refresh token to token.json:

python auth_blogger.py

Keep token.json and credentials.json out of version control (they are listed in .gitignore).

6. Run the workflow

python run.py

This will:

1. Fetch recent items from the configured RSS feeds (last 48 hours, deduplicated).
2. Call Ollama to generate a single blog post (title + HTML body) summarizing those items.
3. Create a new post on your Blogger blog and publish it (or save as draft if blogger.publish_as_draft is true in config.yaml).

Progress is logged to stdout (timestamp, step, and key info). When run from cron, output is appended to logs/ai-news-blog.log.

Scheduling with cron

To run the workflow daily (e.g. at 9:00 AM), use cron. Ensure the cron job runs in an environment where Ollama is available (same machine or set ollama.host / OLLAMA_HOST if Ollama is on another host).

A ready-to-paste cron line for this project is in crontab.example (paths already set for this repo). Output is written to logs/ai-news-blog.log.

crontab -e

Paste the line from crontab.example, then save and exit. Confirm with crontab -l. On macOS, cron runs only when the machine is awake; ensure Ollama is running at the scheduled time.

Configuration reference

Setting	Description
feeds	List of RSS feed URLs.
ollama.model	Ollama model name (e.g. llama3.2).
ollama.host	Optional. Ollama server URL (default: http://localhost:11434).
ollama.max_tokens	Max tokens for the summary (default: 4096).
ollama.timeout_seconds	Optional. Request timeout in seconds (e.g. 300).
ollama.prompt_persona	Optional. Multiline string: AI persona/tone injected at the start of the prompt.
ollama.prompt_requirements	Optional. Multiline string: format and output requirements (title, HTML, TITLE line).
blogger.blog_id	Your Blogger blog ID.
blogger.credentials_path	Path to OAuth client JSON (default: credentials.json).
blogger.token_path	Path to stored refresh token (default: token.json).
blogger.publish_as_draft	If true, posts are created as drafts.

Environment variables:

• BLOGGER_BLOG_ID overrides blogger.blog_id.
• CONFIG_PATH overrides the config file path (default: config.yaml).

Project layout

File / folder	Purpose
run.py	Entrypoint: load config, fetch feeds, summarize, publish.
feeds.py	RSS/Atom fetch, parse, dedupe by URL.
summarize.py	Build prompt, call Ollama, return title + HTML body.
blogger_client.py	OAuth and publish post to Blogger.
auth_blogger.py	One-time OAuth flow to create token.json.
config.yaml	Feeds, Ollama and Blogger settings (copy from config.yaml.example).
crontab.example	Example cron line for scheduled runs.
logs/	Cron output goes to logs/ai-news-blog.log.

Troubleshooting

• “No valid token”: Run python auth_blogger.py again and complete the browser flow.
• “Credentials file not found”: Ensure credentials.json from Google Cloud (Desktop app) is in the project directory or at the path set in config.
• Ollama connection errors: Ensure Ollama is running (ollama serve or the Ollama app) and that the model in config is pulled (ollama pull <model>).
• Ollama timeout or very long runs: Increase ollama.timeout_seconds (e.g. 600) or lower ollama.max_tokens (e.g. 2048). Slow or large models can exceed the default 300s timeout.
• Blog ID: Double-check blog_id in config or BLOGGER_BLOG_ID; wrong ID will cause API errors.
• Feed fetch warnings: Failed feeds are logged; the run continues with the rest. Remove or replace broken feed URLs if needed.

License

Use and modify as you like.