ag2-tealtiger 0.1.0

Deterministic governance adapter for AG2 (AutoGen fork) — policy enforcement, PII detection, cost tracking, per-agent kill switch, governed speaker selection, and structured audit evidence. No LLM in the governance path.

ag2-tealtiger

Deterministic governance adapter for AG2 (AutoGen fork) — policy enforcement, PII detection, cost tracking, per-agent kill switch, governed speaker selection, and structured audit evidence.

No LLM in the governance path. All policy evaluation is deterministic, adding <5ms latency.

Installation

pip install ag2-tealtiger

Quick Start

Zero-Config Observe Mode

from ag2_tealtiger import TealTigerGuard

guard = TealTigerGuard()  # No engine = observe mode
guard.attach(my_agent)
# Cost, PII, and tool usage tracked automatically. Nothing blocked.

Or the convenience subclass:

from ag2_tealtiger import TealTigerAuditAgent

agent = TealTigerAuditAgent(name="coder")  # Zero config, observe mode built-in

Policy Enforcement

from ag2_tealtiger import TealTigerGuard, GovernanceMode

guard = TealTigerGuard(engine=my_engine, mode=GovernanceMode.ENFORCE)
guard.attach(agent)
# DENY blocks the call with a structured denial message

Governed GroupChat

from ag2_tealtiger import GovernedGroupChat, TealTigerGuard, GovernanceMode

guard = TealTigerGuard(engine=engine, mode=GovernanceMode.ENFORCE)
group_chat = GovernedGroupChat(agents=[agent1, agent2, agent3], guard=guard)
speaker = group_chat.select_speaker(last_speaker=agent1)

Features

Feature	Observe	Monitor	Enforce
Cost tracking per agent	✅	✅	✅
PII detection in tool args	✅	✅	✅
Structured audit entries (TEEC)	✅	✅	✅
Per-agent freeze/unfreeze	✅	✅	✅
Policy evaluation	—	✅ (log)	✅ (block)
Budget enforcement	—	✅ (log)	✅ (block)
REFER escalation	—	✅	✅
Inter-agent message governance	✅	✅	✅
Governed speaker selection	✅	✅	✅
JSONL audit export	✅	✅	✅

License

Apache-2.0