maivn-shared 0.4.0

Shared contracts, models, and utilities for the mAIvn platform

maivn-shared

Shared contracts, models, utilities, and logging primitives for the mAIvn platform.

Overview

maivn-shared is the public shared package used by the mAIvn SDK and service repos. It provides the shared domain entities, configuration models, serialization helpers, event contracts, and logging abstractions that need to stay consistent across the platform.

Key Design Principles:

• ✅ Environment-agnostic (no environment variable access)
• ✅ Immutable configuration objects
• ✅ Type-safe Pydantic models
• ✅ Proper dependency inversion
• ✅ Zero external service dependencies

Installation

# Using uv (recommended)
uv add maivn-shared

# Using pip
pip install maivn-shared

Architecture

maivn-shared/
├── domain/          # Pure business entities (no external dependencies)
│   └── entities/    # Pydantic models for tasks, tools, sessions, etc.
├── core/            # Application services
│   ├── config/      # Configuration management
│   ├── data/        # Data utilities (UUID generation)
│   └── http/        # HTTP client
└── infrastructure/  # Factories and adapters
    └── factories/   # Object creation logic

Domain Layer

Pure business entities with no external dependencies:

• Tasks: Planning-time task models (Task, TaskList, TaskToolCall)
• Tools: Tool specifications and execution models (ToolSpec, ToolExecutionCall)
• Sessions: Session request/response models
• Messages: LangChain message wrappers
• Dependencies: Tool dependency models (agent, tool, data, user)
• Status: Lifecycle status enumeration

Core Layer

Application services and utilities:

• ConfigManager: Factory for creating HTTPConfig and ServerConfig
• HTTPClient: Unified HTTP client with proper error handling
• UUID utilities: Deterministic and random UUID generation
• ArgumentResolver: Resolves TaskReferences, DataReferences, and ArgValues in task arguments
• MessageFilter: Filters messages and events for client consumption
• constants: Shared non-sensitive constants (timeouts, separators, etc.)

Infrastructure Layer

Factories and adapters:

• TaskExecutionCallFactory: Creates TaskExecutionCall instances from tasks

Usage Examples

Configuration Management

from maivn_shared.core import config

# Get HTTP configuration with defaults
http_config = config.get_http_config()

# Override specific values
http_config = config.get_http_config(
    timeout=30.0,
    max_retries=5
)

# Get server configuration
server_config = config.get_server_config(
    base_url="https://api.example.com"
)

HTTP Client

from maivn_shared.core import HTTPClient

client = HTTPClient(timeout=10.0)

# Make POST request
response = client.post(
    "https://api.example.com/endpoint",
    payload={"key": "value"}
)

# Make GET request
response = client.get("https://api.example.com/data")

Task Models

from maivn_shared.domain.entities import (
    Task,
    TaskList,
    TaskToolCall,
    ArgValue,
    TaskReference,
    Status
)

# Create a task
task = Task(
    index=0,
    status=Status.PENDING,
    tool_call=TaskToolCall(
        name="search_web",
        args={"query": ArgValue(value="python tutorial")}
    )
)

# Create task list
task_list = TaskList(items=[task])

# Check dependencies
deps = task.get_dependency_indices()
can_run = task.can_start(completed_task_indices=set())

Task Execution Calls

from maivn_shared import ToolSpec
from maivn_shared.domain.entities import Task, TaskList
from maivn_shared.infrastructure.factories import TaskExecutionCallFactory

# Define available tools
tools = [
    ToolSpec(
        tool_id="tool_1",
        agent_id="agent_1",
        name="search_web",
        description="Search the web",
        args_schema={"query": {"type": "string"}}
    )
]

# Create factory
factory = TaskExecutionCallFactory(tools=tools)

# Build execution call from task
execution_call = factory.build(task)

# Build from entire task list
call_list = factory.build_from_tasklist(task_list)

Argument Resolution

from maivn_shared import ArgumentResolver, TaskReference, DataReference, ArgValue

# Create resolver
resolver = ArgumentResolver()

# Resolve TaskReference
result = resolver.resolve(
    TaskReference(task_index=0),
    task_results={0: "search results"}
)
# Result: "search results"

# Resolve DataReference
result = resolver.resolve(
    DataReference(key_name="user_id"),
    private_data={"user_id": "user_123"}
)
# Result: "user_123"

# Resolve nested structure
args = {
    "query": TaskReference(task_index=0),
    "limit": 10,
    "user": DataReference(key_name="user_id")
}
resolved = resolver.resolve(
    args,
    task_results={0: "python tutorial"},
    private_data={"user_id": "user_123"}
)
# Result: {"query": "python tutorial", "limit": 10, "user": "user_123"}

# Convenience function for resolving all args
from maivn_shared import resolve_arguments
resolved_args = resolve_arguments(
    args,
    task_results={0: "value"},
    private_data={"key": "value"}
)

Message Filtering

from maivn_shared import MessageFilter, HumanMessage, AIMessage, SystemMessage

messages = [
    HumanMessage(content="Hello"),
    SystemMessage(content="System prompt"),
    AIMessage(content="Hi there!")
]

# Filter for client messages only (human + AI)
filtered = MessageFilter.filter_client_messages(messages)
# Result: [HumanMessage(...), AIMessage(...)]

# Filter by specific types
filtered = MessageFilter.filter_messages_by_type(
    messages,
    allowed_types={"human", "system"}
)

# Filter event data
event = {
    "results": ["result1"],
    "messages": [...],
    "metadata": {...},
    "_internal": {...}  # filtered out
}
filtered_event = MessageFilter.filter_update_event(event)

Constants

from maivn_shared import constants

# Use shared constants
timeout = constants.DEFAULT_TASK_TIMEOUT_SECONDS
separator = constants.INPUT_SEPARATOR

Logging

from maivn_shared import get_logger

# Initialize with file logging (first call only)
logger = get_logger(log_file_path="/path/to/logs/agent_execution.log")

# Or initialize with console-only logging
logger = get_logger()  # No file logging, only console output

# Use logger
logger.log_session_start(
    session_id="session_123",
    assistant_id="assistant_1",
    thread_id="thread_1"
)

logger.log_task_execution(
    phase="completed",
    task_idx=0,
    tool_name="search_web",
    result={"results": [...]}
)

logger.log_session_end(session_id="session_123")

API Reference

Core Exports

from maivn_shared import (
    # Utilities
    ArgumentResolver,
    resolve_arguments,
    MessageFilter,
    constants,

    # Configuration
    ArgsSchema,
    HTTPConfig,
    ServerConfig,

    # Session
    SessionResponse,

    # Tools
    ToolExecutionCall,
    ToolSpec,
    ToolType,

    # Logging
    MaivnLogger,
    get_logger,
    get_optional_logger,
)

Domain Entities

from maivn_shared.domain.entities import (
    # Tasks
    Task,
    TaskList,
    TaskToolCall,
    TaskResult,
    ArgValue,
    DataReference,
    TaskReference,

    # Tools
    ToolSpec,
    ToolType,
    ToolExecutionCall,
    ToolExecutionCallList,

    # Sessions
    SessionRequest,
    SessionResponse,
    SessionStartupResponse,

    # Messages
    BaseMessage,
    HumanMessage,
    AIMessage,
    SystemMessage,
    ToolMessage,

    # Dependencies
    BaseDependency,
    AgentDependency,
    ToolDependency,
    DataDependency,
    InterruptDependency,

    # Status
    Status,
)

Core Services

from maivn_shared.core import (
    ArgumentResolver,    # Argument resolution utility
    resolve_arguments,   # Convenience function for arg resolution
    MessageFilter,       # Message/event filtering
    constants,           # Shared constants module
    config,              # Global ConfigManager instance
    ConfigManager,       # Configuration factory
    HTTPClient,          # HTTP client
    HTTPClientError,     # Base HTTP exception
    create_uuid,         # UUID generation
)

Infrastructure

from maivn_shared.infrastructure.factories import (
    TaskExecutionCallFactory,
)

Configuration

maivn-shared is intentionally environment-agnostic and does not read deployment-specific environment variables or configuration files. Applications should inject those values explicitly.

Development

Requirements

• Python 3.10+
• langchain-core >= 0.3.76

Code Quality

# Format code
ruff format .

# Lint code
ruff check .

# Type checking
pyright

Releases

• CI runs on pull requests and pushes to main.
• Publish PyPI runs on version tags that match v*.
• Configure PyPI Trusted Publishing for this repository before the first release.
• See DEPLOYMENT.md for the full GitHub and PyPI release procedure.

License

See LICENSE file in the repository root.

Related Packages

• maivn: Public SDK that depends on maivn-shared
• maivn-tools: Optional connector layer for the mAIvn SDK