AI Launch Pipeline#

One-click end-to-end workflow for monitoring AI product launches. Runs four stages in sequence: RSS monitoring → product search → screenshot capture → trend analysis.

Quick Start#

bash

# Full pipeline (one click)
python scripts/run_pipeline.py

# Skip screenshot stage (no Playwright needed)
python scripts/run_pipeline.py --skip-screenshot

# Run a single stage
python scripts/run_pipeline.py --stage rss
python scripts/run_pipeline.py --stage search
python scripts/run_pipeline.py --stage screenshot
python scripts/run_pipeline.py --stage analysis

Stages#

#StageWhat it doesOutput1RSS MonitorFetches configured RSS feeds, detects new AI launch postsdata/raw_launches.json2Product SearchEnriches each launch with DuckDuckGo search resultsdata/enriched_launches.json3ScreenshotCaptures full-page screenshots of product pages (optional)screenshots/*.png, data/screenshot_results.json4Trend AnalysisIdentifies keywords, top organizations, source distributionanalysis/trends.json, analysis/launch_analysis_report.md

Configuration#

Edit config/rss_feeds.yaml to add or remove RSS feeds:

yaml

feeds:
  - name: OpenAI Blog
    url: https://openai.com/blog/rss.xml
    category: llm

Environment Variables#

VariableDefaultDescriptionPIPELINE_BASE_DIRskill rootBase directory for all outputsPIPELINE_DATA_DIR``{BASE}/dataJSON data output directoryPIPELINE_SCREENSHOT_DIR``{BASE}/screenshotsScreenshot output directoryPIPELINE_ANALYSIS_DIR``{BASE}/analysisAnalysis report directoryPIPELINE_CONFIG``config/rss_feeds.yamlRSS feed config path

Dependencies#

• Python 3.10+
• PyYAML — pip install pyyaml
• Playwright (optional, for screenshots) — pip install playwright && playwright install chromium

Output#

After a full run:

text

data/
  raw_launches.json          # Stage 1 output
  seen_ids.json              # Dedup state
  enriched_launches.json     # Stage 2 output
  screenshot_results.json    # Stage 3 output
screenshots/
  *.png                      # Captured screenshots
analysis/
  trends.json                # Structured trend data
  launch_analysis_report.md  # Human-readable report

Scheduling#

Pair with OpenClaw cron for automated daily runs:

bash

# Add via cron tool
schedule: { kind: "cron", expr: "0 8 * * *" }
payload: { kind: "agentTurn", message: "Run the AI launch pipeline: python ~/.openclaw/skills/ai-launch-pipeline/scripts/run_pipeline.py --skip-screenshot, then summarize the analysis report." }