Hanzo AI
6562eb3345
chore: symlink AGENTS.md + CLAUDE.md → LLM.md
2026-04-06 15:06:02 -07:00
Hanzo Dev
8a2d00ce4d
Fix linting issues: specific exception types and remove f-string prefix
2025-11-14 06:49:44 -08:00
Hanzo Dev
67592ef116
Add comprehensive test suite with runtime-generated secrets
...
- 70+ tests covering regex-based sanitization
- API keys: OpenAI, Anthropic, GitHub, HuggingFace, Stripe, SendGrid, Google, AWS
- Crypto: Private keys, ETH/BTC addresses, seed phrases
- Financial: Credit cards, SSN
- PII: Email addresses
- Context preservation tests
- Edge cases and JSONL processing
Security: All test secrets generated at runtime to avoid GitHub secret scanning
- No hardcoded secrets in test files
- Uses random generators for realistic test patterns
- Follows security best practices
CI: GitHub Actions workflow for multi-OS/Python testing with coverage
2025-11-14 06:46:51 -08:00
Hanzo Dev
4dd83c3cce
v0.2.1: Add comprehensive API key protection
...
Additional API key formats protected:
- GitHub: 40-char hex tokens, ghs_* secrets
- Google: AIza* API keys, ya29.* OAuth tokens
- SendGrid: SG.* format
- Stripe: sk_live/test/rk_live keys
- Square: sq0atp/sqOatp tokens
- Facebook: EAACEdEose0c* access tokens
- Twitter: Bearer tokens, API keys
- Slack: xoxe-* app tokens
- Generic: key-* patterns
Ensures comprehensive protection for all major platforms!
2025-11-14 06:00:53 -08:00
Hanzo Dev
c8b092e197
v0.2.0: Comprehensive security sanitization
...
Major security enhancements:
- 🔐 All major API keys/tokens (OpenAI, Anthropic, GitHub, HF, Slack, AWS)
- 💰 Financial data (credit cards, SSN, crypto seed phrases, private keys)
- 🪙 Crypto addresses (Ethereum, Bitcoin, Bech32)
- 🔑 Authentication (PEM keys, JWT, DB passwords, OAuth)
- 👤 PII (emails, phone numbers, IP addresses, file paths)
Protects users from accidentally leaking:
- BIP39 seed phrases (12/24 word patterns)
- 64-char hex private keys
- Credit card numbers
- Social security numbers
- All common API key formats
Safe for contribution to public datasets!
2025-11-14 05:59:01 -08:00
Hanzo Dev
d2bad09534
docs: add install instructions for PyPI
2025-11-13 13:18:02 -08:00
Hanzo Dev
16a3c80f9b
feat: full agentic data extraction
...
- Include ALL tool calls and metadata
- Preserve agent invocations
- Keep full context for agentic training
- Minimal sanitization (only critical PII)
- Extract 4B+ tokens with all structure
Perfect for training coding agents that use tools.
2025-11-13 13:03:41 -08:00
Hanzo Dev
1c726c4710
Initial commit: Claude Collector v0.1.0
...
- Extract agent conversations
- Auto-sanitize PII
- Token counting
- Training dataset export
- Works with uvx/uv tool
2025-11-13 13:00:50 -08:00
Hanzo AI