sinan-agentic-core 0.16.1__py3-none-any.whl

sinan_agentic_core/__init__.py

@@ -0,0 +1,186 @@
+"""Sinan - Generic agent orchestration framework.
+
+This package provides a reusable framework for building multi-agent systems
+using the OpenAI Agents SDK with code orchestration.
+
+Core Components:
+- AgentSession: Manages conversation history
+- AgentContext: Stores workflow state and data
+- AgentRegistry: Declarative agent definitions
+- ToolRegistry: Declarative tool definitions
+- AgentOrchestrator: Workflow orchestration logic
+
+Quick Start:
+    from sinan_agentic_core import (
+        AgentSession,
+        AgentContext,
+        AgentOrchestrator,
+        register_agent,
+        register_tool,
+        AgentDefinition
+    )
+
+    # Define your agent
+    my_agent = AgentDefinition(
+        name="my_agent",
+        description="Does something",
+        instructions="You are a helpful assistant...",
+        tools=["my_tool"]
+    )
+    register_agent(my_agent)
+
+    # Run orchestrator
+    orchestrator = AgentOrchestrator()
+    result = await orchestrator.run(
+        user_query="Do something",
+        context_data={"database_connector": db}
+    )
+"""
+
+from .core import (
+    BaseAgentRunner,
+    Capability,
+    ModelRetryConfig,
+    RetryBackoffConfig,
+    RetryTrigger,
+    RunErrorKind,
+    ToolErrorRecovery,
+    ToolOutputTrimConfig,
+    ToolTracer,
+    TurnBudget,
+    TurnBudgetConfig,
+    build_error_handlers,
+    build_run_config,
+    classify_run_error,
+    recover_invalid_final_output,
+)
+from .instructions import InstructionBuilder
+from .llm import (
+    AzureOpenAIProviderConfig,
+    LLMProviderConfig,
+    OpenAIProviderConfig,
+    configure_llm_provider,
+    load_llm_provider_config,
+    parse_llm_provider_config,
+    resolve_openai_client,
+)
+from .models.context import AgentContext
+from .models.outputs import ChatResponse, ToolOutput
+from .orchestrator import AgentOrchestrator
+from .registry import (
+    AgentCatalog,
+    AgentDefinition,
+    AgentRegistry,
+    AgentYamlEntry,
+    CapabilityFactory,
+    CapabilityNotFoundError,
+    CapabilityRef,
+    CapabilityRegistry,
+    GuardrailCategory,
+    GuardrailDefinition,
+    GuardrailRegistry,
+    ResolvedGuardrails,
+    ToolCatalog,
+    ToolDefinition,
+    ToolRegistry,
+    ToolYamlEntry,
+    create_agent_from_registry,
+    get_agent_registry,
+    get_capability_registry,
+    get_guardrail_registry,
+    get_tool_registry,
+    load_agent_catalog,
+    load_tool_catalog,
+    register_agent,
+    register_capability,
+    register_guardrail,
+    register_tool,
+)
+from .services import (
+    StreamingRunHooks,
+    chat,
+    chat_streamed,
+    chat_with_hooks,
+)
+from .session import AgentSession, ConversationHistory, SQLiteSessionStore
+from .utils import tool_error, tool_response, unwrap_context
+
+# MCP support (optional — requires 'sinan_agentic_core[mcp]')
+# Lazy import to avoid requiring FastMCP for non-MCP users.
+# Usage: from sinan_agentic_core.mcp import build_mcp_server, MCPContextFactory
+
+__all__ = [
+    # Core
+    "BaseAgentRunner",
+    "build_error_handlers",
+    "build_run_config",
+    "Capability",
+    "classify_run_error",
+    "ModelRetryConfig",
+    "recover_invalid_final_output",
+    "RetryBackoffConfig",
+    "RetryTrigger",
+    "RunErrorKind",
+    "ToolErrorRecovery",
+    "ToolOutputTrimConfig",
+    "ToolTracer",
+    "TurnBudget",
+    "TurnBudgetConfig",
+    # Instructions
+    "InstructionBuilder",
+    # LLM providers
+    "AzureOpenAIProviderConfig",
+    "LLMProviderConfig",
+    "OpenAIProviderConfig",
+    "configure_llm_provider",
+    "load_llm_provider_config",
+    "parse_llm_provider_config",
+    "resolve_openai_client",
+    # Session
+    "AgentSession",
+    "ConversationHistory",
+    "SQLiteSessionStore",
+    # Models
+    "AgentContext",
+    "ToolOutput",
+    "ChatResponse",
+    # Registry
+    "AgentCatalog",
+    "AgentYamlEntry",
+    "CapabilityFactory",
+    "CapabilityNotFoundError",
+    "CapabilityRef",
+    "CapabilityRegistry",
+    "get_capability_registry",
+    "register_capability",
+    "load_agent_catalog",
+    "AgentDefinition",
+    "AgentRegistry",
+    "get_agent_registry",
+    "register_agent",
+    "create_agent_from_registry",
+    "ToolCatalog",
+    "ToolYamlEntry",
+    "load_tool_catalog",
+    "ToolDefinition",
+    "ToolRegistry",
+    "get_tool_registry",
+    "register_tool",
+    "GuardrailCategory",
+    "GuardrailDefinition",
+    "GuardrailRegistry",
+    "ResolvedGuardrails",
+    "get_guardrail_registry",
+    "register_guardrail",
+    # Services
+    "StreamingRunHooks",
+    "chat",
+    "chat_with_hooks",
+    "chat_streamed",
+    # Orchestrator
+    "AgentOrchestrator",
+    # Utils
+    "unwrap_context",
+    "tool_response",
+    "tool_error",
+]

sinan_agentic_core/core/__init__.py

@@ -0,0 +1,47 @@
+"""Core agent execution components.
+
+This module contains fundamental components for agent execution:
+- BaseAgentRunner: Abstract base class for all agent runners
+- Capability: Pluggable agent behavior primitive
+- TurnBudget: Soft turn budget with self-extension (Capability)
+- ToolErrorRecovery: Intelligent tool error tracking (Capability)
+- ToolTracer: Observation-only tool-call tracer (Capability)
+- recover_invalid_final_output: SDK handler that salvages malformed structured output
+- build_error_handlers: Run error handlers a built agent needs
+- build_run_config: Run-level SDK settings a built agent needs
+- ModelRetryConfig: Declarative opt-in retry policy for model calls
+- ToolOutputTrimConfig: Declarative opt-in trimming of oversized tool outputs
+- TurnBudgetConfig: Declarative opt-in soft turn budget for one agent
+- RunErrorKind / classify_run_error: Typed classification of agent-run failures
+"""
+
+from .base_runner import BaseAgentRunner
+from .capabilities import Capability
+from .errors import structured_tool_error
+from .model_retry import ModelRetryConfig, RetryBackoffConfig, RetryTrigger
+from .output_recovery import build_error_handlers, recover_invalid_final_output
+from .run_config import build_run_config
+from .run_errors import RunErrorKind, classify_run_error
+from .tool_error_recovery import ToolErrorRecovery
+from .tool_output_trim import ToolOutputTrimConfig
+from .tool_tracer import ToolTracer
+from .turn_budget import TurnBudget, TurnBudgetConfig
+
+__all__ = [
+    "BaseAgentRunner",
+    "build_error_handlers",
+    "build_run_config",
+    "Capability",
+    "classify_run_error",
+    "ModelRetryConfig",
+    "recover_invalid_final_output",
+    "RetryBackoffConfig",
+    "RetryTrigger",
+    "RunErrorKind",
+    "structured_tool_error",
+    "ToolErrorRecovery",
+    "ToolOutputTrimConfig",
+    "ToolTracer",
+    "TurnBudget",
+    "TurnBudgetConfig",
+]