LLM Tool Manager Documentation

Overview

The LLMToolManager provides specialized management for LLM tools in the official tools registry. It handles adding, removing, and updating LLM tools with automatic A2A model card creation and cryptographic identity management.

File: tools/llm_tool_manager.py Class: LLMToolManager Version: 1.0.0 Status: ✅ Active

Architecture

Design Principles

LLM-Specific: Focused on LLM tool management

Model Card Creation: Automatic A2A model card generation

Identity Management: Cryptographic identity integration

Registry Integration: Direct registry modification

Error Handling: Comprehensive error handling

Core Components

class LLMToolManager(BaseTool):
    - memory_agent: MemoryAgent - For workspace management
    - tool_registry_path: Path - Tools registry file path
    - model_cards_path: Path - Model cards directory

Available Actions

1. add

Adds a new LLM tool to the registry.

Parameters:

action (str, required): "add"

tool_id (str, required): Tool identifier

tool_config (Dict, required): Tool configuration

Returns:

Tuple[bool, str]  # (success, message)

2. remove

Removes an LLM tool from the registry.

Parameters:

action (str, required): "remove"

tool_id (str, required): Tool to remove

Returns:

Tuple[bool, str]  # (success, message)

3. update

Updates an existing LLM tool in the registry.

Parameters:

action (str, required): "update"

tool_id (str, required): Tool to update

tool_config (Dict, required): Updated configuration

Returns:

Tuple[bool, str]  # (success, message)

Usage

Add LLM Tool

from tools.llm_tool_manager import LLMToolManager
from agents.memory_agent import MemoryAgent

manager = LLMToolManager(memory_agent=memory_agent)
Add LLM tool
success, message = await manager.execute(
    action="add",
    tool_id="custom_llm_tool",
    tool_config={
        "name": "Custom LLM Tool",
        "description": "Specialized LLM tool",
        "version": "1.0.0",
        "enabled": True,
        "commands": [...],
        "access_control": {...}
    }
)

Update LLM Tool

# Update LLM tool
success, message = await manager.execute(
    action="update",
    tool_id="custom_llm_tool",
    tool_config={
        "name": "Updated LLM Tool",
        "version": "2.0.0"
    }
)

Model Card Creation

Automatic Model Card Generation

When adding or updating LLM tools:

Gets or creates identity via IDManagerAgent

Creates A2A model card

Signs the model card

Saves to model cards directory

Model Card Structure

{
    "id": "tool_id",
    "name": "Tool Name",
    "description": "Tool description",
    "type": "tool",
    "version": "1.0.0",
    "enabled": true,
    "commands": [...],
    "access_control": {...},
    "identity": {
        "public_address": "0x...",
        "signature": "..."
    },
    "a2a_endpoint": "https://mindx.internal/{tool_id}/a2a"
}

Features

1. LLM Tool Focus

Specialized for LLM tools:

LLM-specific configurations

Model card generation

Identity management

2. Model Card Integration

Creates A2A model cards:

Cryptographic identity

Endpoint definition

Tool metadata

3. Identity Management

Integrates with IDManagerAgent:

Gets existing identity

Creates new identity if needed

Signs model cards

Limitations

Current Limitations

LLM-Specific Only: Only for LLM tools

No Validation: Limited config validation

No Rollback: No rollback capability

Basic Operations: Simple add/remove/update

No Versioning: No version history

Recommended Improvements

Enhanced Validation: Comprehensive validation

Rollback Support: Rollback failed operations

Version Control: Version history

LLM Testing: Test LLM tools

Model Management: LLM model management

Performance Tracking: Track LLM tool performance

Consolidation: Consider merging with RegistryManagerTool

Integration

With IDManagerAgent

Manages identities:

id_manager = await IDManagerAgent.get_instance()
identity = await id_manager.get_identity(item_id)

With Memory Agent

Manages model cards:

self.model_cards_path = self.memory_agent.get_agent_data_directory("a2a_model_cards")

Examples

Complete LLM Tool Lifecycle

# 1. Add LLM tool
await manager.execute("add", "llm_analyzer", {...})

2. Update LLM tool
await manager.execute("update", "llm_analyzer", {...})

3. Remove LLM tool
await manager.execute("remove", "llm_analyzer")

Technical Details

Dependencies

core.id_manager_agent.IDManagerAgent: Identity management

agents.memory_agent.MemoryAgent: Workspace management

core.bdi_agent.BaseTool: Base tool class

Registry File

Tools registry located at:

data/config/official_tools_registry.json

Model Cards Directory

Model cards stored in:

data/memory/agent_workspaces/a2a_model_cards/

Comparison with RegistryManagerTool

FeatureLLMToolManagerRegistryManagerTool ScopeLLM tools onlyAll tools + agents Model CardsYesYes IdentityYesYes Use CaseLLM tool managementGeneral registry

Future Enhancements

LLM Testing: Test LLM tools

Model Management: Manage LLM models

Performance Tracking: Track performance

Enhanced Validation: Better validation

Consolidation: Merge with RegistryManagerTool

Version Control: Version history

API Integration: REST API access

---