Manual script to backfill AI analysis for all existing reports.

How to run

cd lafaekstreet_backend
python -m manual_run.backfill_ai_analysis

What it does (per report)

• Downloads image from S3
• Runs AI damage detection via Bedrock Nova
• Saves AI analysis log + updates image record + updates report title
• Submits AI validation to Hedera blockchain
• Generates multimodal embedding

Safety features

• DRY_RUN = True by default — first run will only list reports that need processing, no changes made
• Set DRY_RUN = False when you're ready to actually process
• Configurable delay between reports (DELAY_SECONDS = 1.0) to avoid rate limits
• SKIP_HEDERA and SKIP_EMBEDDINGS flags if you want to skip those
• BATCH_LIMIT = 100 to control how many reports per run
• START/FINISHED logs per report with full details
• Summary at the end showing success/fail/skip counts