Ahy Governance

Open-source AI Agent governance. Deploy agents with confidence — conflict detection, cost control, audit trails, compliance ready.

中文文档

Recent Updates

• Web Dashboard — one-command launch, dark monitoring theme, 7 live-updating panels
• Transparent Proxy — OpenAI-compatible endpoint, zero code changes for existing agents
• Built-in Auth — email/password login + JWT + API Key management
• Docker Deploy — single command, supports Zeabur / EdgeOne Pages

What You Get

Before vs After

3 Lines to Integrate

from ahy_governance import ConflictDetector, CostTracker, AuditReporter

detector = ConflictDetector(); tracker = CostTracker(); auditor = AuditReporter()

Beyond LLM Observability

LangSmith and LangFuse are excellent LLM tracing tools, but they don't understand multi-agent orchestration. Ahy Governance is purpose-built for teams running 5+ agents that collaborate, conflict, and need coordination.

Who's It For

• You're building agents with LangChain / CrewAI and your boss asks "is this safe?"
• Compliance audit is coming and auditors want every AI decision traceable
• You manage 20+ agents and don't know who spent what or who's conflicting with whom
• You need SOC2 / ISO27001 evidence and Excel isn't cutting it anymore

Quick Start

pip install ahy-governance[web]
ahy-dashboard
# Open http://localhost:8080 — click "Demo Data" to populate

Or use individual modules:

from ahy_governance import ConflictDetector, CostTracker, AuditReporter

# Detect conflicts between agents
detector = ConflictDetector()
conflicts = detector.check(agent_outputs, dag_definition)

# Track costs per agent
tracker = CostTracker()
tracker.set_budget(limit_usd=100)
tracker.track("Planner", "claude-opus-4-7", tokens_in=15000, tokens_out=8000)

# Tamper-proof audit logging
auditor = AuditReporter()
auditor.log(AuditEventType.AGENT_START, "Planner", {"task": "plan"})

Web Dashboard

Launch with one command. 7 panels, dark theme, auto-refresh.

ahy-dashboard

Architecture

┌──────────────────────────────────────────────────────┐
│              Ahy Governance Dashboard                 │
│  ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌────────┐  │
│  │ Conflict │ │   Cost   │ │  Audit   │ │ Health │  │
│  │ Detector │ │ Tracker  │ │ Reporter │ │Monitor │  │
│  └──────────┘ └──────────┘ └──────────┘ └────────┘  │
├──────────────────────────────────────────────────────┤
│                 Governance Core                       │
│  ┌──────────┐ ┌──────────┐ ┌──────────────────────┐  │
│  │  Memory  │ │   RBAC   │ │    Prompt Guard      │  │
│  │ Sharing  │ │          │ │ (Injection + PII)    │  │
│  └──────────┘ └──────────┘ └──────────────────────┘  │
├──────────────────────────────────────────────────────┤
│      Existing Agent Core (not included — bring your own)  │
│      Orchestrator  │  TraceLogger  │  Router         │
└──────────────────────────────────────────────────────┘

Modules (7/7 complete)

384 tests, 0 failures.

SOC 2 / ISO 27001 Compliance

The Audit Reporter module provides cryptographic audit logging:

• SHA-256 hash chain — every audit entry is cryptographically linked to its predecessor
• SOC 2 export — one-click report covering Security, Availability, Confidentiality, Processing Integrity, and Privacy
• ISO 27001 export — Annex A controls with compliant/needs-review status per control

Contact

Questions or feedback? Open an issue or email 2115464137@qq.com.

Ecosystem

Community

• Discussions: GitHub Discussions — questions, ideas, feedback
• Issues: GitHub Issues — bug reports, feature requests
• Star the repo: If this is useful, a star helps others discover it

Contributing

PRs welcome. See CONTRIBUTING.md.

MIT License. Built by Leo-Ayh-Oday.