LinkedIn Trend Posts Generator

A local-first pipeline that turns tech/AI trends into LinkedIn-ready posts. Includes a CLI review/publish workflow and a modern web UI backed by a FastAPI service.

Interfaces

• CLI: review, approve, and publish posts from the terminal.
• Web UI: run research queries, generate posts, and edit them in a minimal UI.

Features

• 🤖 Local LLM integration (Ollama; OpenAI fallback optional)
• 🧭 Trend aggregation from Hacker News, RSS feeds, Reddit, and GitHub Trending
• ✍️ Post generation pipeline with validation + source citations
• 🎨 Automatic AI image generation for posts using Stable Diffusion
• ✅ Manual review + approval workflow (CLI + Web UI)
• 🔗 LinkedIn OAuth 2.0 posting flow with token refresh
• 💾 Local storage (SQLite + file-based artifacts)
• 🌗 Light/Dark theme in the Web UI

Project Structure

.
├── src/                     # Core application
│   ├── api/                 # FastAPI backend for the web UI
│   ├── sources/             # Trend sources (HN, RSS, Reddit, GitHub)
│   ├── llm.py               # LLM integration
│   ├── trends.py            # Trend fetch/aggregate
│   ├── post_generator.py    # Post generation pipeline
│   ├── review_cli.py        # Review/approve CLI
│   ├── publish_cli.py       # Publish CLI
│   ├── linkedin_api.py      # LinkedIn API client + OAuth
│   └── database.py          # SQLite storage + migrations
├── scripts/                 # Utilities (OAuth flow, testing, generation)
├── config/                  # Settings + prompts
├── data/                    # SQLite DB + generated posts
├── web/                     # Astro web UI
├── .env.example             # Environment template
├── requirements.txt         # Python dependencies
└── README.md

Setup

1) Python environment

python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

2) Configure environment

cp .env.example .env

Update .env with your settings. The database is created automatically on first run.

3) Local LLM (Ollama recommended)

# Install Ollama: https://ollama.ai
ollama pull llama2

4) Image Generation (Hugging Face)

Get a free API token for automatic image generation:

# 1. Visit https://huggingface.co/settings/tokens
# 2. Create a new token
# 3. Add to .env:
HUGGINGFACE_API_TOKEN=hf_your_token_here

See IMAGE_GENERATION_SETUP.md for detailed setup instructions.

5) Optional: LinkedIn OAuth (publishing)

python scripts/linkedin_oauth.py

This stores tokens in .env. See LINKEDIN_SETUP.md for the full walkthrough.

6) Optional: Reddit source setup

See REDDIT_SETUP.md to configure Reddit API credentials.

Running the Web UI

Start the API (FastAPI)

uvicorn src.api.app:app --reload --port 8000

Environment notes:

• WEB_UI_ORIGINS controls CORS for the API (defaults to local dev hosts).
• The web UI reads PUBLIC_API_BASE to reach the API (default: http://localhost:8000).

Start the Web UI (Astro)

cd web
npm install
npm run dev

Visit http://localhost:4321.

CLI Workflow

Generate a post from current trends

python scripts/generate_post_from_trend.py --save-to-db

Review and approve posts

python src/review_cli.py

Publish approved posts

python src/publish_cli.py

Notes

• .env is ignored by git. Do not commit secrets.
• Publishing requires LinkedIn app credentials and OAuth approval.
• Web UI currently focuses on research + post creation; CLI handles review/publish.

Roadmap (high level)

• Automation & scheduling for end-to-end runs
• Notifications (Slack/email)
• Expanded trend source controls + analytics