LinkedIn Tech News Auto-Poster

Automated pipeline that fetches the latest TechCrunch headline, generates a professional LinkedIn post with AI, and publishes it to LinkedIn using a saved browser session.

What This Project Does

1. Pulls the latest article from TechCrunch RSS (last 24 hours only)
2. Generates a LinkedIn-ready post using Gemini or Ollama
3. Publishes the post to LinkedIn through Playwright
4. Runs daily via GitHub Actions
5. Uses session-based authentication to reduce repeated 2FA prompts

Architecture

• main.py: Single entry point for the entire workflow (fetch, generate, publish, logging)
• save_linkedin_session.py: One-time helper to create/update LinkedIn session state
• .github/workflows/daily-linkedin-post.yml: Daily scheduler and CI runner
• .env.example: Environment variable template

Requirements

• Python 3.11+
• Playwright Chromium browser
• Gemini API key (optional if using Ollama-only)

Install dependencies:

pip install -r requirements.txt
playwright install chromium

Local Setup

1. Copy .env.example to .env
2. Fill required values
3. Create a LinkedIn session file
4. Run main.py

Environment Variables

Core:

• FEED_URL=https://techcrunch.com/feed/
• MODEL_PROVIDER=gemini
• GEMINI_API_KEY=your_key
• GEMINI_MODEL=gemini-2.5-flash-lite
• LINKEDIN_STATE_FILE=linkedin_state.json
• HEADLESS=true
• LOG_LEVEL=INFO

Resilience flags:

• ENABLE_OLLAMA_FALLBACK=true
• SKIP_ON_GEMINI_QUOTA=true
• SKIP_ON_GEMINI_MODEL_NOT_FOUND=true
• SKIP_ON_LINKEDIN_TIMEOUT=true

Fallback login (optional, recommended):

• LINKEDIN_EMAIL=you@example.com
• LINKEDIN_PASSWORD=your_password

Ollama fallback (optional):

• OLLAMA_MODEL=llama3.1
• OLLAMA_URL=http://localhost:11434/api/generate

Session-Based LinkedIn Authentication

This project posts with a saved Playwright auth state file instead of logging in from scratch on each run.

Generate session state:

python save_linkedin_session.py

The script opens a real browser. Complete LinkedIn login and any 2FA, then press Enter in terminal to save session state.

If session expires later, rerun save_linkedin_session.py and update the GitHub secret.

Run Locally

python main.py

GitHub Actions Automation

Workflow file:

• .github/workflows/daily-linkedin-post.yml

Current behavior:

• Runs daily at 09:00 UTC
• Supports manual trigger with workflow_dispatch
• Installs dependencies and Playwright
• Decodes LinkedIn session from secret
• Executes main.py

Required GitHub Secrets

• GEMINI_API_KEY
• LINKEDIN_SESSION_B64
• LINKEDIN_EMAIL
• LINKEDIN_PASSWORD

To create LINKEDIN_SESSION_B64 locally (PowerShell):

[Convert]::ToBase64String([IO.File]::ReadAllBytes("linkedin_state.json"))

Paste the output into the LINKEDIN_SESSION_B64 repository secret.

Reliability and Failure Handling

Built-in skip-safe behavior:

• No news in 24h: skips run cleanly
• Gemini quota exhausted: optional skip instead of fail
• Gemini model mismatch: optional skip instead of fail
• LinkedIn timeout/UI drift: optional skip instead of fail

This keeps CI healthy while still logging actionable causes.

Common Troubleshooting

1) Gemini 429 RESOURCE_EXHAUSTED

Cause: quota/rate limit reached.

Fixes:

• Wait for quota reset
• Ensure correct project/key pairing
• Keep Ollama fallback enabled

2) Gemini 404 model not found

Cause: model name not supported for your API/version/project.

Fixes:

• Use GEMINI_MODEL=gemini-2.5-flash-lite
• Keep model fallback logic enabled

3) LinkedIn start post timeout

Cause: expired session, verification checkpoint, or UI variant.

Fixes:

• Refresh session with save_linkedin_session.py
• Update LINKEDIN_SESSION_B64 secret
• Keep SKIP_ON_LINKEDIN_TIMEOUT=true for non-blocking CI

Security Notes

• Do not commit .env
• Do not commit linkedin_state.json
• Keep all sensitive data in GitHub Secrets

Portfolio Notes

This repository demonstrates:

• Practical workflow orchestration in Python
• API fallback and resilience design
• Browser automation with session persistence
• CI/CD automation with scheduled jobs
• Production-style logging and graceful degradation