Pre-tool capability · PyPI 0.1.0 · Alpha · Apache-2.0

pydantic-ai-acgs for Pydantic AI.

Hooks into Pydantic AI tool validation and pre-execution lifecycle callbacks to enforce ACGS-Lite policy rules.

PyPI 0.1.0 GitHub repository Inspect src/pydantic_ai_acgs/capability.py#L30

Package Specification · Inspected HEAD b454b55

Framework maintainer
Pydantic · pydantic.dev
Upstream repository
pydantic/pydantic-ai
PyPI release
pydantic-ai-acgs==0.1.0
Python requirement
>=3.10
Framework constraint
pydantic-ai-slim>=2.16.0,<3
Governance engine pin
acgs-lite>=3.0.1,<4
Integration mechanic
Pre-tool veto / intervention hook
Hook entry point
ACGSCapability.after_tool_validate / before_tool_execute

01 · Local Setup & Quickstart

Attach governance to one tool-calling agent.

Install the pinned release from PyPI alongside acgs-lite>=3.0.1,<4. Rule evaluation uses deterministic keyword and regex matching locally in Python—no external model or network call is made by the governance check.

Caveat: Pre-execution gate after schema validation; does not wrap external side-effect settlement.

Shell · PyPI installation

pip install "pydantic-ai-acgs==0.1.0"

Python · Pydantic AI integration

src/pydantic_ai_acgs/capability.py#L30
from pydantic_ai import Agent, DeferredToolRequests
from pydantic_ai_acgs import ACGSCapability

agent = Agent(
    "openai:gpt-5.2",
    capabilities=[ACGSCapability()],
    output_type=[str, DeferredToolRequests],
)

02 · Verdict & Review Mapping

How ACGS-Lite verdicts map to Pydantic AI.

When a tool call is proposed, the adapter formats the tool name and arguments via format_tool_call and evaluates them against the configured Constitution. Blocking rules always take precedence over review rules.

Verdict.ALLOW

Passes validated tool arguments from after_tool_validate into before_tool_execute.

Verdict.DENY

Raises ToolFailed with the ACGS violation message without consuming the tool's retry budget.

Verdict.REVIEW

Raises ApprovalRequired so Pydantic AI emits a DeferredToolRequests output for human-in-the-loop approval.

03 · Public API Surface

Exported symbols and source entry points.

Each adapter keeps its public surface minimal and re-exports the shared GovernanceDecision, Verdict, and ToolCallFormatter types alongside the framework-specific hook or middleware class.

__all__ exports · pydantic-ai-acgs

src/pydantic_ai_acgs/capability.py#L30 ↗
ACGSCapabilityMETADATA_KEYGovernanceDecisionToolCallFormatterVerdictformat_tool_call

Primary hook surface: ACGSCapability.after_tool_validate / before_tool_execute. Custom ToolCallFormatter functions may be supplied if your tool arguments require custom canonical serialization before rule matching.

04 · Scope & Bypass Boundaries

What this adapter does and does not cover.

Verify these framework-specific boundaries when integrating pydantic-ai-acgs into an evaluation workflow.

All 13 adaptersReceipts, grants & EffectStore
  1. 01

    Include DeferredToolRequests in output_type if you want review-flagged tool calls to pause the run rather than fail.

  2. 02

    Pydantic AI 2.55+ requires Python 3.11+, while earlier 2.16+ releases support Python 3.10+.

  3. 03

    In the current 0.1.x adapter series, rule evaluation runs through acgs-lite's ToolCallGovernor path. If your workflow requires SignedReceipt, ExecutionGrant, or SQLite EffectStore reservations, wire those contracts explicitly at the tool executor boundary.

Other pre-tool capability packages

claude-agent-acgs0.1.0

Claude Agent SDK · Anthropic

openai-agents-acgs0.1.0

OpenAI Agents SDK · OpenAI

adk-acgs0.1.1

Google ADK · Google

crewai-acgs0.1.0

CrewAI · CrewAI

Put one real workflow under your control.

Choose one tool. Define what it may do. Inspect what happens when a request crosses the boundary.

Choose an evaluation Compare all products Request evaluation access Review Claims & Evidence