govagent 0.2.1__tar.gz → 0.2.2__tar.gz

govagent-0.2.2/PKG-INFO

@@ -0,0 +1,184 @@
+Metadata-Version: 2.4
+Name: govagent
+Version: 0.2.2
+Summary: The Governance-First Framework for Agentic AI
+Project-URL: Homepage, https://github.com/thekakodkar/govagent
+Author: Niraj Kakodkar
+Author-email: niraj.kakodkar@gmail.com
+License: MIT
+License-File: LICENSE
+Keywords: EU-AI-Act,agents,ai,enterprise,governance,roi
+Requires-Python: >=3.10
+Requires-Dist: httpx>=0.25.0
+Requires-Dist: pydantic>=2.0.0
+Requires-Dist: python-dotenv>=1.0.0
+Requires-Dist: pyyaml>=6.0.0
+Requires-Dist: slack-sdk>=3.0.0
+Provides-Extra: langchain
+Requires-Dist: langchain-core>=0.1.0; extra == 'langchain'
+Requires-Dist: langchain-openai>=0.1.0; extra == 'langchain'
+Requires-Dist: openai; extra == 'langchain'
+Description-Content-Type: text/markdown
+
+## govAgent: Enterprise-Grade AI Governance Framework
+
+**The Governance-First Framework for Production-Grade Autonomous Systems.**
+
+GovAgent provides a high-abstraction **Control Plane** for agentic AI. With a clear chain of accountability, this lightweight framework helps move autonomous systems from experimental sandboxes into governed, production environments.
+
+<img width="1097" height="479" alt="ImagegoV" src="https://github.com/user-attachments/assets/4e05d505-63d6-4f14-9475-cc00b4f20d73" />
+
+The v0.2.2 Update introduces a standardized LangChain Integration Layer and hardened Slack Socket Mode connectivity, ensuring that high-risk AI actions are always subject to verified human oversight.
+
+---
+
+## 🏗️ Core Pillars: The Chain of Accountability
+GovAgent replaces "Black Box" reasoning with a transparent, governed loop:
+
+1.  **Policy (The Law):** Declarative boundaries and "Rules of Engagement" defined by stakeholders in `policy.yaml`.
+2.  **Guards (The Enforcement):** Real-time circuit breakers that intercept agent intent *before* API execution to prevent budget or security breaches.
+3.  **HITL (The Judiciary):** **Synchronous** Human-in-the-Loop escalation. High-risk actions are physically blocked until an explicit "Approve" or "Reject" signal is received via Slack or CLI.
+4.  **Telemetry (The Evidence):** Forensic-grade audit trails providing an immutable ledger of compliance and real-world ROI.
+
+---
+
+## ⚖️ Regulatory Compliance: EU AI Act (Regulation 2024/1689)
+
+GovAgent is engineered to facilitate compliance for **High-Risk AI Systems** as defined under the EU AI Act. The framework provides the technical primitives required to satisfy key transparency and oversight mandates:
+
+*   **Article 14: Human Oversight**  
+    Native **Synchronous HITL** (Human-in-the-Loop) adapters ensure that high-risk tool execution is physically gated by natural persons in real-time.
+    
+*   **Article 9: Risk Management System**  
+    Automated, policy-driven enforcement (via `policy.yaml`) identifies and mitigates operational and financial risks *prior* to tool deployment.
+    
+*   **Article 12: Record-Keeping & Traceability**  
+    Immutable **Forensic Telemetry** captures the full "Chain of Accountability," including agent reasoning, tool parameters, and the identity of the human overseer who authorized the action.
+
+*   **Article 13: Transparency & Provision of Information**  
+    Automated generation of Execution Snapshots allows for the clear explanation of AI-driven outputs to end-users and regulators.
+    
+---
+
+## 🛠️ Key Capabilities (v0.2.2)
+* **🔗 LangChain Bridge:** Standardized `@langchain_tool` wrappers for seamless interception of LangChain agent intents.
+* **🔌 Socket Mode Judiciary:** Secure, persistent WebSocket connections for Slack oversight without exposing public endpoints (No Ngrok required).
+* **📦 Pydantic Telemetry:** Hardened serialization of agent tasks to ensure audit logs meet strict enterprise data schemas.
+* **⚠️ Constitutional Startup Check:** Refuses to boot if tool registry and policy permissions do not match, eliminating "Shadow AI."
+* **🛡️ Zero-Trust Guardrails:** Hardened whitelisting for all agent actions and financial disbursements.
+
+---
+## 🗺️ Strategic Roadmap
+
+### ✅ v0.2.2: Operational Stability (Current)
+* **LangChain Integration:** Standardized tool-guarding for LangChain ecosystems.
+* **Synchronous HITL:** Stabilized Slack Socket Mode and CLI adapters for real-time intervention.
+* **Intent Serialization:** Resolved telemetry validation errors via JSON-based intent snapshots.
+
+### 🚀 v0.3.0: Enterprise Connectivity (Next)
+* **Fiscal Ceilings:** Recursive approval for multi-agent sub-tasks and "Total Cost of Operation" (TCO) guardrails.
+* **Cloud Exporters:** Native integrations for enterprise logging stacks (AWS CloudWatch / Azure Monitor).
+* **Dynamic Budgeting:** Real-time API pricing integration for penny-accurate cost tracking.
+---
+
+## 📖 Usage Example: Governed LangChain Tool (Simplified API)
+
+GovAgent v0.2.2 introduces a streamlined import structure and native support for LangChain tool interception. The following example demonstrates how to gate a high-risk financial transaction with a synchronous Slack Judiciary.
+
+```python
+import os
+import asyncio
+from langchain_core.tools import tool as langchain_tool
+from govagent import ExecutiveAgent, Policy, HITLManager, SlackJudiciaryAdapter
+
+@langchain_tool
+async def healthcare_payment_tool(amount: float) -> str:
+    """Authorizes payments for healthcare claims. Input: amount."""
+    
+    # 1. Initialize the Judiciary Adapter (Slack Socket Mode)
+    # This establishes a secure, persistent connection for real-time oversight.
+    adapter = SlackJudiciaryAdapter(
+        bot_token=os.getenv("SLACK_BOT_TOKEN"),
+        app_token=os.getenv("SLACK_APP_TOKEN"),
+        channel_id=os.getenv("SLACK_CHANNEL_ID")
+    )
+    adapter.start() 
+    
+    # 2. Initialize Governance Layer via Flat API
+    manager = HITLManager(adapter=adapter)
+    policy = Policy.from_yaml("policies/healthcare_policy.yaml")
+    
+    # 3. ARTICLE 14: Synchronous Human Oversight
+    # Execution physically pauses here until a signal is received from Slack.
+    print(f"⚖️ [GovAgent] Intercepting request for ${amount}...")
+    
+    approved = await manager.secure_approval(
+        agent_id="Healthcare-Director-v0.2.2",
+        reason=f"High-risk authorization required for GuardedPayment (${amount}).",
+        context={
+            "amount": f"${amount}",
+            "compliance_check": "EU-AI-ACT-HIGH-RISK"
+        }
+    )
+
+    if approved:
+        # Business logic proceeds only after explicit human authorization
+        return f"SUCCESS: Payment of ${amount} authorized via Slack."
+    
+    return "REJECTED: Transaction denied by Human Judiciary."
+
+# Example Invocation
+async def main():
+    result = await healthcare_payment_tool.ainvoke({"amount": 1200.0})
+    print(result)
+
+if __name__ == "__main__":
+    asyncio.run(main())
+```
+---
+## ⚙️ Installation
+
+GovAgent is designed to be lightweight and modular. You can install the core framework or include specific integrations as needed.
+
+### 1. Core Installation (Lightweight)
+Recommended for users building custom agents or those who only require the Judiciary and Policy layers.
+```bash
+pip install govagent
+```
+### 2. Full Integration (With LangChain)
+Includes all dependencies required to run governed LangChain sessions, including the langchain_tool wrappers and OpenAI clients.
+
+```bash
+pip install "govagent[langchain]"
+```
+### 3. Development Installation
+If you are contributing to the framework or running the examples in this repository, install in editable mode:
+
+```bash
+git clone [https://github.com/thekakodkar/govagent.git](https://github.com/thekakodkar/govagent.git)
+cd govagent
+pip install -e ".[langchain]"
+
+```
+### 🚀 Quick Setup
+Ensure your .env file is configured with the necessary tokens for the Judiciary Layer to function:
+
+Code snippet
+# Slack Credentials (Socket Mode)
+SLACK_BOT_TOKEN=xoxb-your-token
+SLACK_APP_TOKEN=xapp-your-token
+SLACK_CHANNEL_ID=C12345678
+
+# Model Provider
+OPENAI_API_KEY=sk-your-key
+
+---
+
+**"Governance is not a constraint; it is the catalyst for enterprise AI adoption."**
+
+---
+
+### Author Stamp
+*   **Framework:** GovAgent v0.2.2 (Stable)
+*   **Status:** Active / Open-Source Standard
+*   **Compliance:** Designed for Enterprise-Grade Accountability

govagent-0.2.2/README.md

@@ -0,0 +1,162 @@
+## govAgent: Enterprise-Grade AI Governance Framework
+
+**The Governance-First Framework for Production-Grade Autonomous Systems.**
+
+GovAgent provides a high-abstraction **Control Plane** for agentic AI. With a clear chain of accountability, this lightweight framework helps move autonomous systems from experimental sandboxes into governed, production environments.
+
+<img width="1097" height="479" alt="ImagegoV" src="https://github.com/user-attachments/assets/4e05d505-63d6-4f14-9475-cc00b4f20d73" />
+
+The v0.2.2 Update introduces a standardized LangChain Integration Layer and hardened Slack Socket Mode connectivity, ensuring that high-risk AI actions are always subject to verified human oversight.
+
+---
+
+## 🏗️ Core Pillars: The Chain of Accountability
+GovAgent replaces "Black Box" reasoning with a transparent, governed loop:
+
+1.  **Policy (The Law):** Declarative boundaries and "Rules of Engagement" defined by stakeholders in `policy.yaml`.
+2.  **Guards (The Enforcement):** Real-time circuit breakers that intercept agent intent *before* API execution to prevent budget or security breaches.
+3.  **HITL (The Judiciary):** **Synchronous** Human-in-the-Loop escalation. High-risk actions are physically blocked until an explicit "Approve" or "Reject" signal is received via Slack or CLI.
+4.  **Telemetry (The Evidence):** Forensic-grade audit trails providing an immutable ledger of compliance and real-world ROI.
+
+---
+
+## ⚖️ Regulatory Compliance: EU AI Act (Regulation 2024/1689)
+
+GovAgent is engineered to facilitate compliance for **High-Risk AI Systems** as defined under the EU AI Act. The framework provides the technical primitives required to satisfy key transparency and oversight mandates:
+
+*   **Article 14: Human Oversight**  
+    Native **Synchronous HITL** (Human-in-the-Loop) adapters ensure that high-risk tool execution is physically gated by natural persons in real-time.
+    
+*   **Article 9: Risk Management System**  
+    Automated, policy-driven enforcement (via `policy.yaml`) identifies and mitigates operational and financial risks *prior* to tool deployment.
+    
+*   **Article 12: Record-Keeping & Traceability**  
+    Immutable **Forensic Telemetry** captures the full "Chain of Accountability," including agent reasoning, tool parameters, and the identity of the human overseer who authorized the action.
+
+*   **Article 13: Transparency & Provision of Information**  
+    Automated generation of Execution Snapshots allows for the clear explanation of AI-driven outputs to end-users and regulators.
+    
+---
+
+## 🛠️ Key Capabilities (v0.2.2)
+* **🔗 LangChain Bridge:** Standardized `@langchain_tool` wrappers for seamless interception of LangChain agent intents.
+* **🔌 Socket Mode Judiciary:** Secure, persistent WebSocket connections for Slack oversight without exposing public endpoints (No Ngrok required).
+* **📦 Pydantic Telemetry:** Hardened serialization of agent tasks to ensure audit logs meet strict enterprise data schemas.
+* **⚠️ Constitutional Startup Check:** Refuses to boot if tool registry and policy permissions do not match, eliminating "Shadow AI."
+* **🛡️ Zero-Trust Guardrails:** Hardened whitelisting for all agent actions and financial disbursements.
+
+---
+## 🗺️ Strategic Roadmap
+
+### ✅ v0.2.2: Operational Stability (Current)
+* **LangChain Integration:** Standardized tool-guarding for LangChain ecosystems.
+* **Synchronous HITL:** Stabilized Slack Socket Mode and CLI adapters for real-time intervention.
+* **Intent Serialization:** Resolved telemetry validation errors via JSON-based intent snapshots.
+
+### 🚀 v0.3.0: Enterprise Connectivity (Next)
+* **Fiscal Ceilings:** Recursive approval for multi-agent sub-tasks and "Total Cost of Operation" (TCO) guardrails.
+* **Cloud Exporters:** Native integrations for enterprise logging stacks (AWS CloudWatch / Azure Monitor).
+* **Dynamic Budgeting:** Real-time API pricing integration for penny-accurate cost tracking.
+---
+
+## 📖 Usage Example: Governed LangChain Tool (Simplified API)
+
+GovAgent v0.2.2 introduces a streamlined import structure and native support for LangChain tool interception. The following example demonstrates how to gate a high-risk financial transaction with a synchronous Slack Judiciary.
+
+```python
+import os
+import asyncio
+from langchain_core.tools import tool as langchain_tool
+from govagent import ExecutiveAgent, Policy, HITLManager, SlackJudiciaryAdapter
+
+@langchain_tool
+async def healthcare_payment_tool(amount: float) -> str:
+    """Authorizes payments for healthcare claims. Input: amount."""
+    
+    # 1. Initialize the Judiciary Adapter (Slack Socket Mode)
+    # This establishes a secure, persistent connection for real-time oversight.
+    adapter = SlackJudiciaryAdapter(
+        bot_token=os.getenv("SLACK_BOT_TOKEN"),
+        app_token=os.getenv("SLACK_APP_TOKEN"),
+        channel_id=os.getenv("SLACK_CHANNEL_ID")
+    )
+    adapter.start() 
+    
+    # 2. Initialize Governance Layer via Flat API
+    manager = HITLManager(adapter=adapter)
+    policy = Policy.from_yaml("policies/healthcare_policy.yaml")
+    
+    # 3. ARTICLE 14: Synchronous Human Oversight
+    # Execution physically pauses here until a signal is received from Slack.
+    print(f"⚖️ [GovAgent] Intercepting request for ${amount}...")
+    
+    approved = await manager.secure_approval(
+        agent_id="Healthcare-Director-v0.2.2",
+        reason=f"High-risk authorization required for GuardedPayment (${amount}).",
+        context={
+            "amount": f"${amount}",
+            "compliance_check": "EU-AI-ACT-HIGH-RISK"
+        }
+    )
+
+    if approved:
+        # Business logic proceeds only after explicit human authorization
+        return f"SUCCESS: Payment of ${amount} authorized via Slack."
+    
+    return "REJECTED: Transaction denied by Human Judiciary."
+
+# Example Invocation
+async def main():
+    result = await healthcare_payment_tool.ainvoke({"amount": 1200.0})
+    print(result)
+
+if __name__ == "__main__":
+    asyncio.run(main())
+```
+---
+## ⚙️ Installation
+
+GovAgent is designed to be lightweight and modular. You can install the core framework or include specific integrations as needed.
+
+### 1. Core Installation (Lightweight)
+Recommended for users building custom agents or those who only require the Judiciary and Policy layers.
+```bash
+pip install govagent
+```
+### 2. Full Integration (With LangChain)
+Includes all dependencies required to run governed LangChain sessions, including the langchain_tool wrappers and OpenAI clients.
+
+```bash
+pip install "govagent[langchain]"
+```
+### 3. Development Installation
+If you are contributing to the framework or running the examples in this repository, install in editable mode:
+
+```bash
+git clone [https://github.com/thekakodkar/govagent.git](https://github.com/thekakodkar/govagent.git)
+cd govagent
+pip install -e ".[langchain]"
+
+```
+### 🚀 Quick Setup
+Ensure your .env file is configured with the necessary tokens for the Judiciary Layer to function:
+
+Code snippet
+# Slack Credentials (Socket Mode)
+SLACK_BOT_TOKEN=xoxb-your-token
+SLACK_APP_TOKEN=xapp-your-token
+SLACK_CHANNEL_ID=C12345678
+
+# Model Provider
+OPENAI_API_KEY=sk-your-key
+
+---
+
+**"Governance is not a constraint; it is the catalyst for enterprise AI adoption."**
+
+---
+
+### Author Stamp
+*   **Framework:** GovAgent v0.2.2 (Stable)
+*   **Status:** Active / Open-Source Standard
+*   **Compliance:** Designed for Enterprise-Grade Accountability

{govagent-0.2.1 → govagent-0.2.2}/changelog.md

@@ -2,6 +2,16 @@
 
 All notable changes to the **GovAgent** framework are documented in this file. This project adheres to a "Governance-First" versioning strategy, prioritizing human oversight and fiscal accountability.
 
+## [0.2.2] - 2026-05-05
+### Added
+- **LangChain Integration**: Standardized @langchain_tool bridging logic for GovAgent interception.
+- **Socket Mode Support**: Transitioned Slack Judiciary to persistent Socket Mode for increased security.
+- **Intent Serialization**: JSON-based serialization for tool inputs to ensure telemetry data integrity.
+
+### Fixed
+- **Validation Schema**: Resolved Pydantic string-type constraints in ExecutionSnapshot.
+- **Registry Mapping**: Fixed namespace mismatches between Python tool decorators and YAML policy names.
+
 ---
 
 ## [0.2.0] - 2026-05-04 (Stable)

govagent-0.2.2/examples/langchain_integration.py

@@ -0,0 +1,67 @@
+import asyncio
+import json
+import os
+from pathlib import Path
+from dotenv import load_dotenv
+
+# LangChain & GovAgent Imports
+from langchain_core.tools import tool as langchain_tool
+
+# GovAgent v0.2.2 Flat API
+from govagent import ExecutiveAgent, Policy, HITLManager, SlackJudiciaryAdapter
+
+load_dotenv()
+BASE_DIR = Path(__file__).parent
+
+@langchain_tool
+async def healthcare_payment_tool(amount: float) -> str:
+    """Authorizes payments for healthcare claims. Input: amount."""
+    
+    # 1. Setup Slack Adapter (Mirroring your working test)
+    adapter = SlackJudiciaryAdapter(
+        bot_token=os.getenv("SLACK_BOT_TOKEN"),
+        app_token=os.getenv("SLACK_APP_TOKEN"),
+        channel_id=os.getenv("SLACK_CHANNEL_ID")
+    )
+    adapter.start() # Start Socket Mode listener
+    
+    # 2. Setup Manager & Policy
+    manager = HITLManager(adapter=adapter)
+    policy = Policy.from_yaml(BASE_DIR / "../policies/langchain_integration_sample_policy.yaml")
+    
+    # 3. Initialize Agent
+    agent = ExecutiveAgent(
+        persona="Healthcare Billing Director",
+        policy=policy,
+        model_client=None,
+        hitl_manager=manager
+    )
+
+    print(f"\n⚖️ [GovAgent] Intercepting request for ${amount}...")
+
+    # 4. ARTICLE 14: Use the 'secure_approval' method from your working test
+    # We call the manager directly to ensure the Slack handshake happens
+    approved = await manager.secure_approval(
+        agent_id="Healthcare-Director-v0.3.0",
+        reason=f"Authorization required for GuardedPayment (${amount}).",
+        context={
+            "amount": f"${amount}",
+            "tool": "GuardedPayment",
+            "compliance_check": "EU-AI-ACT-HIGH-RISK"
+        }
+    )
+
+    if approved:
+        # In a real app, this would call your business logic
+        return f"SUCCESS: Payment of ${amount} was approved via Slack."
+    else:
+        return "REJECTED: The transaction was denied by the Human Judiciary in Slack."
+
+async def main():
+    print("🚀 Starting Governed LangChain Session (Slack Mode)...")
+    # Simulate a high-risk tool call
+    result = await healthcare_payment_tool.ainvoke({"amount": 1200.0})
+    print(f"\n🏁 Final System Output: {result}")
+
+if __name__ == "__main__":
+    asyncio.run(main())

govagent-0.2.2/policies/langchain_integration_sample_policy.yaml

@@ -0,0 +1,30 @@
+# govAgent Policy Configuration v0.3.0
+version: "0.3.0"
+
+metadata:
+  org_name: "DecodeAIWithNiraj"
+  compliance_level: "EU-AI-ACT-HIGH-RISK"
+
+# 1. DEFINE TOOLS & RISK LEVELS
+tools:
+  - name: "GuardedPayment"
+    risk_level: "HIGH"
+    require_human_approval: true  # This triggers the Slack Judiciary
+    description: "Financial transaction tool for procurement"
+    max_per_transaction: 2000.0
+
+  - name: "SearchTool"
+    risk_level: "LOW"
+    require_human_approval: false # This will execute automatically
+    description: "General web search for information gathering"
+
+# 2. GLOBAL FISCAL GUARDRAILS (v0.3.0 Feature)
+global_limits:
+  daily_budget_usd: 100.0
+  max_tokens_per_run: 4000
+
+# 3. JUDICIARY SETTINGS
+judiciary:
+  primary_channel: "slack"
+  fallback: "cli"              # If Slack is down, ask in the terminal
+  timeout_seconds: 300         # Wait 5 mins for Director approval before auto-reject

govagent-0.2.2/pyproject.toml

@@ -0,0 +1,36 @@
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[project]
+name = "govagent"
+version = "0.2.2"
+description = "The Governance-First Framework for Agentic AI"
+readme = "README.md"
+requires-python = ">=3.10"
+license = {text = "MIT"}
+authors = [
+    {name = "Niraj Kakodkar"}, 
+    {email = "niraj.kakodkar@gmail.com"}
+]
+keywords = ["ai", "agents", "governance", "roi", "enterprise", "EU-AI-Act"]
+
+# CORE DEPENDENCIES: Kept minimal to reduce the footprint
+dependencies = [
+    "pydantic>=2.0.0",
+    "pyyaml>=6.0.0",
+    "httpx>=0.25.0",
+    "slack-sdk>=3.0.0",
+    "python-dotenv>=1.0.0",
+]
+
+# OPTIONAL DEPENDENCIES: Users install via pip install "govagent[langchain]"
+[project.optional-dependencies]
+langchain = [
+    "langchain-core>=0.1.0",
+    "langchain-openai>=0.1.0",
+    "openai"
+]
+
+[project.urls]
+"Homepage" = "https://github.com/thekakodkar/govagent"

govagent-0.2.2/src/govagent/__init__.py

@@ -0,0 +1,13 @@
+# src/govagent/__init__.py
+
+from .agent import ExecutiveAgent
+from .policy import Policy
+from .hitl.manager import HITLManager
+from .hitl.slack_adapter import SlackJudiciaryAdapter
+
+__all__ = [
+    "ExecutiveAgent",
+    "Policy",
+    "HITLManager",
+    "SlackJudiciaryAdapter"
+]

govagent-0.2.1/PKG-INFO

@@ -1,105 +0,0 @@
-Metadata-Version: 2.4
-Name: govagent
-Version: 0.2.1
-Summary: The Governance-First Framework for Agentic AI
-Project-URL: Homepage, https://github.com/thekakodkar/govagent
-Author: Niraj Kakodkar
-Author-email: niraj.kakodkar@gmail.com
-License: MIT
-License-File: LICENSE
-Keywords: agents,ai,enterprise,governance,roi
-Requires-Python: >=3.10
-Requires-Dist: httpx>=0.25.0
-Requires-Dist: openai
-Requires-Dist: pydantic>=2.0.0
-Requires-Dist: pyyaml>=6.0.0
-Requires-Dist: slack-sdk
-Description-Content-Type: text/markdown
-
-## govAgent: Enterprise-Grade AI Governance Framework
-
-**The Governance-First Framework for Production-Grade Autonomous Systems.**
-
-GovAgent provides a high-abstraction **Control Plane** for agentic AI. With a clear chain of accountability, this lightweight framework helps move autonomous systems from experimental sandboxes into governed, production environments.
-
-<img width="1097" height="479" alt="ImagegoV" src="https://github.com/user-attachments/assets/4e05d505-63d6-4f14-9475-cc00b4f20d73" />
-
-
-The **v0.2.0 Stable Release** introduces a persistent, bi-directional Judiciary layer, ensuring that high-risk AI actions are always subject to human verification before execution.
-
----
-
-## 🏗️ Core Pillars: The Chain of Accountability
-GovAgent replaces "Black Box" reasoning with a transparent, governed loop:
-
-1.  **Policy (The Law):** Declarative boundaries and "Rules of Engagement" defined by stakeholders in `policy.yaml`.
-2.  **Guards (The Enforcement):** Real-time circuit breakers that intercept agent intent *before* API execution to prevent budget or security breaches.
-3.  **HITL (The Judiciary):** **Synchronous** Human-in-the-Loop escalation. High-risk actions are physically blocked until an explicit "Approve" or "Reject" signal is received via Slack or CLI.
-4.  **Telemetry (The Evidence):** Forensic-grade audit trails providing an immutable ledger of compliance and real-world ROI.
-
----
-
-## 🇪🇺 EU AI Act Alignment
-GovAgent is architected to meet the rigorous transparency and oversight standards for **High-Risk AI Systems**:
-
-*   **Human Oversight (Article 14):** Native HITL adapters ensure high-risk systems are overseen by natural persons in real-time.
-*   **Risk Management (Article 9):** Automated policy enforcement identifies and mitigates operational risks before they manifest.
-*   **Traceability (Article 12):** Forensic telemetry captures who approved an action and when, satisfying the "Chain of Accountability" required for regulatory audits.
-
----
-
-## 🛠️ Key Capabilities (v0.2.0 Stable)
-*   **@tool Registry:** A type-safe decorator that auto-maps Python functions to policy permissions (risk level, category, signature).
-*   **Socket Mode Handshake:** Secure, persistent WebSocket connections for judiciary oversight without exposing public endpoints.
-*   **Financial Circuit Breakers:** Real-time monitoring of session spend with automated halting when budget caps are reached.
-*   **Constitutional Startup Check:** Refuses to boot if tool code and policy permissions do not match, eliminating "Shadow AI".
-*   **Zero-Trust Guardrails:** Hardened whitelisting for all agent actions and web domain access.
-
----
-
-## 🗺️ Strategic Roadmap
-
-### ✅ v0.2.0: Operational Safety (Current Stable)
-*   **Synchronous HITL:** Full implementation of Slack and CLI adapters for real-time intervention.
-*   **Legislative Registry:** Stabilized @tool decorator for type-safe permission mapping.
-*   **Forensic Telemetry:** Real-time ROI and audit trail generation.
-
-### 🚀 v0.3.0: Enterprise Connectivity (Next)
-*   **Fiscal Ceilings:** Recursive approval for multi-agent sub-tasks and "Total Cost of Operation" (TCO) guardrails.
-*   **Cloud Exporters:** Native integrations for enterprise logging stacks like AWS CloudWatch and Azure Monitor.
-*   **Dynamic Budgeting:** Real-time API pricing integration for penny-accurate cost tracking.
-
----
-
-## 📖 Usage Example: Controlled Execution
-
-```python
-from govagent.agent import ExecutiveAgent
-from govagent.policy import Policy
-
-# Load Hardened Policy (The Law)
-policy = Policy.from_yaml("policies/enterprise_audit_policy.yaml")
-
-# Initialize Executive Agent
-agent = ExecutiveAgent(
-    persona="Technology Director",
-    policy=policy,
-    model_client=YourModelClient()
-)
-
-# Execute Governed Task
-# High-risk tools (e.g., payments) will pause and alert Slack.
-report = await agent.execute("Analyze Q4 market shifts and authorize $5k payment.")
-print(f"Status: {report.status} | ROI: ${report.estimated_cost_usd}")
-```
-
----
-
-**"Governance is not a constraint; it is the catalyst for enterprise AI adoption."**
-
----
-
-### Author Stamp
-*   **Framework:** GovAgent v0.2.0 (Stable)
-*   **Status:** Active / Open-Source Standard
-*   **Compliance:** Designed for Enterprise-Grade Accountability

govagent-0.2.1/README.md

@@ -1,87 +0,0 @@
-## govAgent: Enterprise-Grade AI Governance Framework
-
-**The Governance-First Framework for Production-Grade Autonomous Systems.**
-
-GovAgent provides a high-abstraction **Control Plane** for agentic AI. With a clear chain of accountability, this lightweight framework helps move autonomous systems from experimental sandboxes into governed, production environments.
-
-<img width="1097" height="479" alt="ImagegoV" src="https://github.com/user-attachments/assets/4e05d505-63d6-4f14-9475-cc00b4f20d73" />
-
-
-The **v0.2.0 Stable Release** introduces a persistent, bi-directional Judiciary layer, ensuring that high-risk AI actions are always subject to human verification before execution.
-
----
-
-## 🏗️ Core Pillars: The Chain of Accountability
-GovAgent replaces "Black Box" reasoning with a transparent, governed loop:
-
-1.  **Policy (The Law):** Declarative boundaries and "Rules of Engagement" defined by stakeholders in `policy.yaml`.
-2.  **Guards (The Enforcement):** Real-time circuit breakers that intercept agent intent *before* API execution to prevent budget or security breaches.
-3.  **HITL (The Judiciary):** **Synchronous** Human-in-the-Loop escalation. High-risk actions are physically blocked until an explicit "Approve" or "Reject" signal is received via Slack or CLI.
-4.  **Telemetry (The Evidence):** Forensic-grade audit trails providing an immutable ledger of compliance and real-world ROI.
-
----
-
-## 🇪🇺 EU AI Act Alignment
-GovAgent is architected to meet the rigorous transparency and oversight standards for **High-Risk AI Systems**:
-
-*   **Human Oversight (Article 14):** Native HITL adapters ensure high-risk systems are overseen by natural persons in real-time.
-*   **Risk Management (Article 9):** Automated policy enforcement identifies and mitigates operational risks before they manifest.
-*   **Traceability (Article 12):** Forensic telemetry captures who approved an action and when, satisfying the "Chain of Accountability" required for regulatory audits.
-
----
-
-## 🛠️ Key Capabilities (v0.2.0 Stable)
-*   **@tool Registry:** A type-safe decorator that auto-maps Python functions to policy permissions (risk level, category, signature).
-*   **Socket Mode Handshake:** Secure, persistent WebSocket connections for judiciary oversight without exposing public endpoints.
-*   **Financial Circuit Breakers:** Real-time monitoring of session spend with automated halting when budget caps are reached.
-*   **Constitutional Startup Check:** Refuses to boot if tool code and policy permissions do not match, eliminating "Shadow AI".
-*   **Zero-Trust Guardrails:** Hardened whitelisting for all agent actions and web domain access.
-
----
-
-## 🗺️ Strategic Roadmap
-
-### ✅ v0.2.0: Operational Safety (Current Stable)
-*   **Synchronous HITL:** Full implementation of Slack and CLI adapters for real-time intervention.
-*   **Legislative Registry:** Stabilized @tool decorator for type-safe permission mapping.
-*   **Forensic Telemetry:** Real-time ROI and audit trail generation.
-
-### 🚀 v0.3.0: Enterprise Connectivity (Next)
-*   **Fiscal Ceilings:** Recursive approval for multi-agent sub-tasks and "Total Cost of Operation" (TCO) guardrails.
-*   **Cloud Exporters:** Native integrations for enterprise logging stacks like AWS CloudWatch and Azure Monitor.
-*   **Dynamic Budgeting:** Real-time API pricing integration for penny-accurate cost tracking.
-
----
-
-## 📖 Usage Example: Controlled Execution
-
-```python
-from govagent.agent import ExecutiveAgent
-from govagent.policy import Policy
-
-# Load Hardened Policy (The Law)
-policy = Policy.from_yaml("policies/enterprise_audit_policy.yaml")
-
-# Initialize Executive Agent
-agent = ExecutiveAgent(
-    persona="Technology Director",
-    policy=policy,
-    model_client=YourModelClient()
-)
-
-# Execute Governed Task
-# High-risk tools (e.g., payments) will pause and alert Slack.
-report = await agent.execute("Analyze Q4 market shifts and authorize $5k payment.")
-print(f"Status: {report.status} | ROI: ${report.estimated_cost_usd}")
-```
-
----
-
-**"Governance is not a constraint; it is the catalyst for enterprise AI adoption."**
-
----
-
-### Author Stamp
-*   **Framework:** GovAgent v0.2.0 (Stable)
-*   **Status:** Active / Open-Source Standard
-*   **Compliance:** Designed for Enterprise-Grade Accountability

govagent-0.2.1/pyproject.toml

@@ -1,26 +0,0 @@
-[build-system]
-requires = ["hatchling"]
-build-backend = "hatchling.build"
-
-[project]
-name = "govagent"
-version = "0.2.1"
-description = "The Governance-First Framework for Agentic AI"
-readme = "README.md"
-requires-python = ">=3.10"
-license = {text = "MIT"}
-authors = [
-    {name = "Niraj Kakodkar"},
-    {email = "niraj.kakodkar@gmail.com"}
-]
-keywords = ["ai", "agents", "governance", "roi", "enterprise"]
-dependencies = [
-    "pydantic>=2.0.0",
-    "pyyaml>=6.0.0",
-    "httpx>=0.25.0",  # Async-ready for modern LLM APIs
-    "slack-sdk",
-    "openai" # or your specific model client
-]
-
-[project.urls]
-"Homepage" = "https://github.com/thekakodkar/govagent"