OSINTai v3.4

OSINTai v3.4

Highlights

• Default Ollama analysis model switched to osint-tuned-v3:latest.
• Default embedding model switched to bge-m3:latest.
• Added bounded async Ollama analysis and embedding calls.
• Reworked crawl scheduling with O(1) queue operations and duplicate suppression for queued, active, and visited URLs.
• Fixed resume handling for multi-seed crawls.
• Activated proxy request support with per-proxy success/failure tracking.
• Replaced XOR pseudo-Simhash with weighted 64-bit Simhash near-duplicate detection.
• Expanded indicator extraction for domains, IPv4s, URLs, emails, phones, BTC/ETH addresses, and social handles.
• Added IP nodes and relationships to graph exports.
• Hardened .gitignore for local OSINT outputs, credentials, proxy lists, and investigation artifacts.
• Updated README with current v3.4 commands and conda setup.

Validation

• conda run -n osintai python -m compileall -q src run_osintai.py
• conda run -n osintai python run_osintai.py --help
• git diff --check