agnt5 0.3.0a8__cp310-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl

agnt5/client.py

@@ -0,0 +1,1478 @@
+"""AGNT5 Client SDK for invoking components."""
+
+import json
+import os
+from typing import Any, AsyncIterator, Dict, Iterator, Optional
+from urllib.parse import urljoin
+
+import httpx
+
+from .events import Event, EventType
+
+# Environment variable for API key
+AGNT5_API_KEY_ENV = "AGNT5_API_KEY"
+
+
+def _parse_sse_to_event(event_type_str: str, data: Dict[str, Any]) -> Event:
+    """Convert SSE event type and data to typed Event object.
+
+    Args:
+        event_type_str: The event type string from SSE (e.g., "agent.started")
+        data: The parsed JSON data from the SSE data field
+
+    Returns:
+        Event object with typed event_type and data payload
+    """
+    try:
+        event_type = EventType(event_type_str)
+    except ValueError:
+        # Unknown event type - store as-is with a generic type
+        # This allows forward compatibility with new event types
+        return Event(
+            event_type=EventType.PROGRESS_UPDATE,
+            data={"_raw_event_type": event_type_str, **data},
+            content_index=data.get("index", 0),
+            sequence=data.get("sequence", 0),
+        )
+
+    return Event(
+        event_type=event_type,
+        data=data,
+        content_index=data.get("index", 0),
+        sequence=data.get("sequence", 0),
+    )
+
+
+class Client:
+    """Client for invoking AGNT5 components.
+
+    This client provides a simple interface for calling functions, workflows,
+    and other components deployed on AGNT5.
+
+    Example:
+        ```python
+        from agnt5 import Client
+
+        # Local development (no auth needed)
+        client = Client("http://localhost:34181")
+        result = client.run("greet", {"name": "Alice"})
+        print(result)  # {"message": "Hello, Alice!"}
+
+        # Production with API key
+        client = Client(
+            gateway_url="https://api.agnt5.com",
+            api_key="agnt5_sk_xxxxxxxxxxxxxxxxxxxxxxxxxxxx"
+        )
+
+        # Or use AGNT5_API_KEY environment variable
+        # export AGNT5_API_KEY=agnt5_sk_xxxxxxxxxxxxxxxxxxxxxxxxxxxx
+        client = Client(gateway_url="https://api.agnt5.com")
+        ```
+    """
+
+    def __init__(
+        self,
+        gateway_url: str = "http://localhost:34181",
+        timeout: float = 30.0,
+        api_key: Optional[str] = None,
+    ):
+        """Initialize the AGNT5 client.
+
+        Args:
+            gateway_url: Base URL of the AGNT5 gateway (default: http://localhost:34181)
+            timeout: Request timeout in seconds (default: 30.0)
+            api_key: Service key for authentication. If not provided, falls back to
+                     AGNT5_API_KEY environment variable. Keys start with "agnt5_sk_".
+        """
+        self.gateway_url = gateway_url.rstrip("/")
+        self.timeout = timeout
+        # Use provided api_key or fallback to environment variable
+        self.api_key = api_key or os.environ.get(AGNT5_API_KEY_ENV)
+        self._client = httpx.Client(timeout=timeout)
+
+    def _build_headers(
+        self,
+        session_id: Optional[str] = None,
+        user_id: Optional[str] = None,
+    ) -> Dict[str, str]:
+        """Build request headers with authentication and optional session/user context.
+
+        Args:
+            session_id: Session identifier for multi-turn conversations
+            user_id: User identifier for user-scoped memory
+
+        Returns:
+            Dictionary of HTTP headers
+        """
+        headers = {"Content-Type": "application/json"}
+        if self.api_key:
+            headers["X-API-KEY"] = self.api_key
+        if session_id:
+            headers["X-Session-ID"] = session_id
+        if user_id:
+            headers["X-User-ID"] = user_id
+        return headers
+
+    def run(
+        self,
+        component: str,
+        input_data: Optional[Dict[str, Any]] = None,
+        component_type: str = "function",
+        session_id: Optional[str] = None,
+        user_id: Optional[str] = None,
+    ) -> Dict[str, Any]:
+        """Execute a component synchronously and wait for the result.
+
+        This is a blocking call that waits for the component to complete execution.
+
+        Args:
+            component: Name of the component to execute
+            input_data: Input data for the component (will be sent as JSON body)
+            component_type: Type of component - "function", "workflow", "agent", "tool" (default: "function")
+            session_id: Session identifier for multi-turn conversations (optional)
+            user_id: User identifier for user-scoped memory (optional)
+
+        Returns:
+            Dictionary containing the component's output
+
+        Raises:
+            RunError: If the component execution fails
+            httpx.HTTPError: If the HTTP request fails
+
+        Example:
+            ```python
+            # Simple function call (default)
+            result = client.run("greet", {"name": "Alice"})
+
+            # Workflow execution (explicit)
+            result = client.run("order_fulfillment", {"order_id": "123"}, component_type="workflow")
+
+            # Multi-turn conversation with session
+            result = client.run("chat", {"message": "Hello"}, session_id="session-123")
+
+            # User-scoped memory
+            result = client.run("assistant", {"message": "Help me"}, user_id="user-456")
+
+            # No input data
+            result = client.run("get_status")
+            ```
+        """
+        if input_data is None:
+            input_data = {}
+
+        # Build URL with component type
+        url = urljoin(self.gateway_url + "/", f"v1/run/{component_type}/{component}")
+
+        # Make request with auth and session headers
+        response = self._client.post(
+            url,
+            json=input_data,
+            headers=self._build_headers(session_id=session_id, user_id=user_id),
+        )
+
+        # Handle errors
+        if response.status_code == 404:
+            try:
+                error_data = response.json()
+                raise RunError(
+                    error_data.get("error", "Component not found"),
+                    run_id=error_data.get("runId"),
+                )
+            except ValueError:
+                # JSON parsing failed
+                raise RunError(f"Component '{component}' not found")
+
+        if response.status_code == 503:
+            error_data = response.json()
+            raise RunError(
+                f"Service unavailable: {error_data.get('error', 'Unknown error')}",
+                run_id=error_data.get("runId"),
+            )
+
+        if response.status_code == 504:
+            error_data = response.json()
+            raise RunError(
+                "Execution timeout",
+                run_id=error_data.get("runId"),
+            )
+
+        # Handle 500 errors with our RunResponse format
+        if response.status_code == 500:
+            try:
+                error_data = response.json()
+                raise RunError(
+                    error_data.get("error", "Unknown error"),
+                    run_id=error_data.get("runId"),
+                )
+            except ValueError:
+                # JSON parsing failed, fall through to raise_for_status
+                response.raise_for_status()
+        else:
+            # For other error codes, use standard HTTP error handling
+            response.raise_for_status()
+
+        # Parse response
+        data = response.json()
+
+        # Check execution status
+        if data.get("status") == "failed":
+            raise RunError(
+                data.get("error", "Unknown error"),
+                run_id=data.get("runId"),
+            )
+
+        # Return output
+        return data.get("output", {})
+
+    def submit(
+        self,
+        component: str,
+        input_data: Optional[Dict[str, Any]] = None,
+        component_type: str = "function",
+    ) -> str:
+        """Submit a component for async execution and return immediately.
+
+        This is a non-blocking call that returns a run ID immediately.
+        Use get_status() to check progress and get_result() to retrieve the output.
+
+        Args:
+            component: Name of the component to execute
+            input_data: Input data for the component (will be sent as JSON body)
+            component_type: Type of component - "function", "workflow", "agent", "tool" (default: "function")
+
+        Returns:
+            String containing the run ID
+
+        Raises:
+            httpx.HTTPError: If the HTTP request fails
+
+        Example:
+            ```python
+            # Submit async function (default)
+            run_id = client.submit("process_video", {"url": "https://..."})
+            print(f"Submitted: {run_id}")
+
+            # Submit workflow
+            run_id = client.submit("order_fulfillment", {"order_id": "123"}, component_type="workflow")
+
+            # Check status later
+            status = client.get_status(run_id)
+            if status["status"] == "completed":
+                result = client.get_result(run_id)
+            ```
+        """
+        if input_data is None:
+            input_data = {}
+
+        # Build URL with component type
+        url = urljoin(self.gateway_url + "/", f"v1/submit/{component_type}/{component}")
+
+        # Make request with auth headers
+        response = self._client.post(
+            url,
+            json=input_data,
+            headers=self._build_headers(),
+        )
+
+        # Handle errors
+        response.raise_for_status()
+
+        # Parse response and extract run ID
+        data = response.json()
+        return data.get("runId", "")
+
+    def get_status(self, run_id: str) -> Dict[str, Any]:
+        """Get the current status of a run.
+
+        Args:
+            run_id: The run ID returned from submit()
+
+        Returns:
+            Dictionary containing status information:
+            {
+                "runId": "...",
+                "status": "pending|running|completed|failed|cancelled",
+                "submittedAt": 1234567890,
+                "startedAt": 1234567891,  // optional
+                "completedAt": 1234567892 // optional
+            }
+
+        Raises:
+            httpx.HTTPError: If the HTTP request fails
+
+        Example:
+            ```python
+            status = client.get_status(run_id)
+            print(f"Status: {status['status']}")
+            ```
+        """
+        url = urljoin(self.gateway_url + "/", f"v1/status/{run_id}")
+
+        response = self._client.get(url, headers=self._build_headers())
+        response.raise_for_status()
+
+        return response.json()
+
+    def get_result(self, run_id: str) -> Dict[str, Any]:
+        """Get the result of a completed run.
+
+        This will raise an error if the run is not yet complete.
+
+        Args:
+            run_id: The run ID returned from submit()
+
+        Returns:
+            Dictionary containing the component's output
+
+        Raises:
+            RunError: If the run failed or is not yet complete
+            httpx.HTTPError: If the HTTP request fails
+
+        Example:
+            ```python
+            try:
+                result = client.get_result(run_id)
+                print(result)
+            except RunError as e:
+                if "not complete" in str(e):
+                    print("Run is still in progress")
+                else:
+                    print(f"Run failed: {e}")
+            ```
+        """
+        url = urljoin(self.gateway_url + "/", f"v1/result/{run_id}")
+
+        response = self._client.get(url, headers=self._build_headers())
+
+        # Handle 404 - run not complete or not found
+        if response.status_code == 404:
+            error_data = response.json()
+            error_msg = error_data.get("error", "Run not found or not complete")
+            current_status = error_data.get("status", "unknown")
+            raise RunError(f"{error_msg} (status: {current_status})", run_id=run_id)
+
+        # Handle other errors
+        response.raise_for_status()
+
+        # Parse response
+        data = response.json()
+
+        # Check if run failed
+        if data.get("status") == "failed":
+            raise RunError(
+                data.get("error", "Unknown error"),
+                run_id=run_id,
+            )
+
+        # Return output
+        return data.get("output", {})
+
+    def wait_for_result(
+        self,
+        run_id: str,
+        timeout: float = 300.0,
+        poll_interval: float = 1.0,
+    ) -> Dict[str, Any]:
+        """Wait for a run to complete and return the result.
+
+        This polls the status endpoint until the run completes or times out.
+
+        Args:
+            run_id: The run ID returned from submit()
+            timeout: Maximum time to wait in seconds (default: 300)
+            poll_interval: How often to check status in seconds (default: 1.0)
+
+        Returns:
+            Dictionary containing the component's output
+
+        Raises:
+            RunError: If the run fails or times out
+            httpx.HTTPError: If the HTTP request fails
+
+        Example:
+            ```python
+            # Submit and wait for result
+            run_id = client.submit("long_task", {"data": "..."})
+            try:
+                result = client.wait_for_result(run_id, timeout=600)
+                print(result)
+            except RunError as e:
+                print(f"Failed: {e}")
+            ```
+        """
+        import time
+
+        start_time = time.time()
+
+        while True:
+            # Check timeout
+            elapsed = time.time() - start_time
+            if elapsed >= timeout:
+                raise RunError(
+                    f"Timeout waiting for run to complete after {timeout}s",
+                    run_id=run_id,
+                )
+
+            # Get current status
+            status = self.get_status(run_id)
+            current_status = status.get("status", "")
+
+            # Check if complete
+            if current_status in ("completed", "failed", "cancelled"):
+                # Get result (will raise if failed)
+                return self.get_result(run_id)
+
+            # Wait before next poll
+            time.sleep(poll_interval)
+
+    def stream(
+        self,
+        component: str,
+        input_data: Optional[Dict[str, Any]] = None,
+    ):
+        """Stream responses from a component using Server-Sent Events (SSE).
+
+        This method yields chunks as they arrive from the component.
+        Perfect for LLM token streaming and incremental responses.
+
+        Args:
+            component: Name of the component to execute
+            input_data: Input data for the component (will be sent as JSON body)
+
+        Yields:
+            String chunks as they arrive from the component
+
+        Raises:
+            RunError: If the component execution fails
+            httpx.HTTPError: If the HTTP request fails
+
+        Example:
+            ```python
+            # Stream LLM tokens
+            for chunk in client.stream("generate_text", {"prompt": "Write a story"}):
+                print(chunk, end="", flush=True)
+            ```
+        """
+        if input_data is None:
+            input_data = {}
+
+        # Build URL
+        url = urljoin(self.gateway_url + "/", f"v1/stream/{component}")
+
+        # Use streaming request with auth headers
+        with self._client.stream(
+            "POST",
+            url,
+            json=input_data,
+            headers=self._build_headers(),
+            timeout=300.0,  # 5 minute timeout for streaming
+        ) as response:
+            # Check for errors
+            if response.status_code != 200:
+                # For streaming responses, we can't read the full text
+                # Just raise an HTTP error
+                raise RunError(
+                    f"HTTP {response.status_code}: Streaming request failed",
+                    run_id=None,
+                )
+
+            # Parse SSE stream
+            current_event = None
+            for line in response.iter_lines():
+                line = line.strip()
+
+                # Skip empty lines and comments
+                if not line or line.startswith(":"):
+                    continue
+
+                # Parse event type: "event: output.delta"
+                if line.startswith("event: "):
+                    current_event = line[7:]  # Remove "event: " prefix
+                    continue
+
+                # Parse SSE format: "data: {...}"
+                if line.startswith("data: "):
+                    data_str = line[6:]  # Remove "data: " prefix
+
+                    try:
+                        data = json.loads(data_str)
+
+                        # Check for completion
+                        if data.get("done") or current_event == "done":
+                            return
+
+                        # Check for error
+                        if "error" in data:
+                            raise RunError(
+                                data.get("error"),
+                                run_id=data.get("runId"),
+                            )
+
+                        # Yield chunk from output.delta events
+                        if current_event == "output.delta" and "content" in data:
+                            yield data["content"]
+                        # Also support legacy "chunk" format
+                        elif "chunk" in data:
+                            yield data["chunk"]
+
+                    except json.JSONDecodeError:
+                        # Skip malformed JSON
+                        continue
+
+    def stream_events(
+        self,
+        component: str,
+        input_data: Optional[Dict[str, Any]] = None,
+        component_type: str = "function",
+        session_id: Optional[str] = None,
+        user_id: Optional[str] = None,
+        timeout: float = 300.0,
+    ) -> Iterator[Event]:
+        """Stream typed Event objects from a component execution.
+
+        This method yields Event objects as they arrive from the component,
+        providing full access to the event taxonomy including agent lifecycle,
+        LM streaming, tool calls, and workflow events.
+
+        Args:
+            component: Name of the component to execute
+            input_data: Input data for the component (will be sent as JSON body)
+            component_type: Type of component - "function", "workflow", "agent", "tool"
+            session_id: Session identifier for multi-turn conversations (optional)
+            user_id: User identifier for user-scoped memory (optional)
+            timeout: Stream timeout in seconds (default: 300.0 / 5 minutes)
+
+        Yields:
+            Event objects as they arrive from the stream
+
+        Raises:
+            RunError: If the component execution fails
+            httpx.HTTPError: If the HTTP request fails
+
+        Example:
+            ```python
+            from agnt5 import Client, EventType
+
+            client = Client()
+
+            # Stream agent events
+            for event in client.stream_events("my_agent", {"message": "Hi"}, "agent"):
+                if event.event_type == EventType.AGENT_STARTED:
+                    print(f"Agent started: {event.data['agent_name']}")
+                elif event.event_type == EventType.LM_MESSAGE_DELTA:
+                    print(event.data['content'], end='', flush=True)
+                elif event.event_type == EventType.AGENT_COMPLETED:
+                    print(f"\\nDone: {event.data['output']}")
+            ```
+        """
+        if timeout <= 0:
+            raise ValueError("timeout must be a positive number")
+
+        if input_data is None:
+            input_data = {}
+
+        # Build URL with component type (using v2 streaming endpoint)
+        url = urljoin(self.gateway_url + "/", f"v1/streamv2/{component_type}/{component}")
+
+        # Use streaming request with auth and session headers
+        with self._client.stream(
+            "POST",
+            url,
+            json=input_data,
+            headers=self._build_headers(session_id=session_id, user_id=user_id),
+            timeout=timeout,
+        ) as response:
+            # Check for errors
+            if response.status_code != 200:
+                # Try to get error details from response body
+                try:
+                    error_body = response.read().decode("utf-8")
+                    error_data = json.loads(error_body)
+                    error_msg = error_data.get("error", f"HTTP {response.status_code}")
+                    run_id = error_data.get("runId")
+                except (json.JSONDecodeError, UnicodeDecodeError):
+                    error_msg = f"HTTP {response.status_code}: Streaming request failed"
+                    run_id = None
+                raise RunError(error_msg, run_id=run_id)
+
+            # Parse SSE stream
+            current_event_type: Optional[str] = None
+            for line in response.iter_lines():
+                line = line.strip()
+
+                # Skip empty lines and comments (keep-alive)
+                if not line or line.startswith(":"):
+                    continue
+
+                # Parse event type: "event: agent.started"
+                if line.startswith("event: "):
+                    current_event_type = line[7:]  # Remove "event: " prefix
+                    continue
+
+                # Parse SSE data: "data: {...}"
+                if line.startswith("data: "):
+                    data_str = line[6:]  # Remove "data: " prefix
+
+                    try:
+                        data = json.loads(data_str)
+
+                        # Check for completion signal
+                        if data.get("done") or current_event_type == "done":
+                            return
+
+                        # Check for error event
+                        if current_event_type == "error" or "error" in data:
+                            error_msg = data.get("error", "Unknown streaming error")
+                            raise RunError(error_msg, run_id=data.get("runId"))
+
+                        # Yield typed Event object
+                        if current_event_type:
+                            yield _parse_sse_to_event(current_event_type, data)
+
+                    except json.JSONDecodeError:
+                        # Skip malformed JSON
+                        continue
+
+    def entity(self, entity_type: str, key: str) -> "EntityProxy":
+        """Get a proxy for calling methods on a durable entity.
+
+        This provides a fluent API for entity method invocations with key-based routing.
+
+        Args:
+            entity_type: The entity class name (e.g., "Counter", "ShoppingCart")
+            key: The entity instance key (e.g., "user-123", "cart-alice")
+
+        Returns:
+            EntityProxy that allows method calls on the entity
+
+        Example:
+            ```python
+            # Call entity method
+            result = client.entity("Counter", "user-123").increment(amount=5)
+            print(result)  # 5
+
+            # Shopping cart
+            result = client.entity("ShoppingCart", "user-alice").add_item(
+                item_id="item-123",
+                quantity=2,
+                price=29.99
+            )
+            ```
+        """
+        return EntityProxy(self, entity_type, key)
+
+    def workflow(self, workflow_name: str) -> "WorkflowProxy":
+        """Get a proxy for invoking a workflow with fluent API.
+
+        This provides a convenient API for workflow invocations, including
+        a chat() method for multi-turn conversation workflows.
+
+        Args:
+            workflow_name: Name of the workflow to invoke
+
+        Returns:
+            WorkflowProxy that provides workflow-specific methods
+
+        Example:
+            ```python
+            # Standard workflow execution
+            result = client.workflow("order_process").run(order_id="123")
+
+            # Chat workflow with session
+            response = client.workflow("support_bot").chat(
+                message="My order hasn't arrived",
+                session_id="user-123",
+            )
+
+            # Continue conversation
+            response = client.workflow("support_bot").chat(
+                message="Can you track it?",
+                session_id="user-123",
+            )
+            ```
+        """
+        return WorkflowProxy(self, workflow_name)
+
+    def session(self, session_type: str, key: str) -> "SessionProxy":
+        """Get a proxy for a session entity (OpenAI/ADK-style API).
+
+        This is a convenience wrapper around entity() specifically for SessionEntity subclasses,
+        providing a familiar API for developers coming from OpenAI Agents SDK or Google ADK.
+
+        Args:
+            session_type: The session entity class name (e.g., "Conversation", "ChatSession")
+            key: The session instance key (typically user ID or session ID)
+
+        Returns:
+            SessionProxy that provides session-specific methods
+
+        Example:
+            ```python
+            # Create a conversation session
+            session = client.session("Conversation", "user-alice")
+
+            # Chat with the session
+            response = session.chat("Hello! How are you?")
+            print(response)
+
+            # Get conversation history
+            history = session.get_history()
+            for msg in history:
+                print(f"{msg['role']}: {msg['content']}")
+            ```
+        """
+        return SessionProxy(self, session_type, key)
+
+    def close(self):
+        """Close the underlying HTTP client."""
+        self._client.close()
+
+    def __enter__(self):
+        """Context manager entry."""
+        return self
+
+    def __exit__(self, exc_type, exc_val, exc_tb):
+        """Context manager exit."""
+        self.close()
+
+
+class EntityProxy:
+    """Proxy for calling methods on a durable entity instance.
+
+    This class enables fluent method calls on entities using Python's
+    attribute access. Any method call is translated to an HTTP request
+    to /entity/:type/:key/:method.
+
+    Example:
+        ```python
+        counter = client.entity("Counter", "user-123")
+        result = counter.increment(amount=5)  # Calls /entity/Counter/user-123/increment
+        ```
+    """
+
+    def __init__(self, client: "Client", entity_type: str, key: str):
+        """Initialize entity proxy.
+
+        Args:
+            client: The AGNT5 client instance
+            entity_type: The entity class name
+            key: The entity instance key
+        """
+        self._client = client
+        self._entity_type = entity_type
+        self._key = key
+
+    def __getattr__(self, method_name: str):
+        """Dynamic method lookup that creates entity method callers.
+
+        Args:
+            method_name: The entity method to call
+
+        Returns:
+            Callable that executes the entity method
+        """
+
+        def method_caller(*args, **kwargs) -> Any:
+            """Call an entity method with the given parameters.
+
+            Args:
+                *args: Positional arguments (not recommended, use kwargs)
+                **kwargs: Method parameters as keyword arguments
+
+            Returns:
+                The method's return value
+
+            Raises:
+                RunError: If the method execution fails
+                ValueError: If both positional and keyword arguments are provided
+            """
+            # Convert positional args to kwargs if provided
+            if args and kwargs:
+                raise ValueError(
+                    f"Cannot mix positional and keyword arguments when calling entity method '{method_name}'. "
+                    "Please use keyword arguments only."
+                )
+
+            # If positional args provided, we can't convert them without knowing parameter names
+            # Raise helpful error
+            if args:
+                raise ValueError(
+                    f"Entity method '{method_name}' requires keyword arguments, but got {len(args)} positional arguments. "
+                    f"Example: .{method_name}(param1=value1, param2=value2)"
+                )
+
+            # Build URL: /v1/entity/:entityType/:key/:method
+            url = urljoin(
+                self._client.gateway_url + "/",
+                f"v1/entity/{self._entity_type}/{self._key}/{method_name}",
+            )
+
+            # Make request with method parameters as JSON body and auth headers
+            response = self._client._client.post(
+                url,
+                json=kwargs,
+                headers=self._client._build_headers(),
+            )
+
+            # Handle errors
+            if response.status_code == 504:
+                error_data = response.json()
+                raise RunError(
+                    "Execution timeout",
+                    run_id=error_data.get("run_id"),
+                )
+
+            if response.status_code == 500:
+                try:
+                    error_data = response.json()
+                    raise RunError(
+                        error_data.get("error", "Unknown error"),
+                        run_id=error_data.get("run_id"),
+                    )
+                except ValueError:
+                    response.raise_for_status()
+            else:
+                response.raise_for_status()
+
+            # Parse response
+            data = response.json()
+
+            # Check execution status
+            if data.get("status") == "failed":
+                raise RunError(
+                    data.get("error", "Unknown error"),
+                    run_id=data.get("run_id"),
+                )
+
+            # Return output
+            return data.get("output")
+
+        return method_caller
+
+
+class SessionProxy(EntityProxy):
+    """Proxy for session entities with conversation-specific helper methods.
+
+    This extends EntityProxy to provide familiar APIs for session-based
+    conversations, similar to OpenAI Agents SDK and Google ADK.
+
+    Example:
+        ```python
+        # Create a session
+        session = client.session("Conversation", "user-alice")
+
+        # Chat
+        response = session.chat("Tell me about AI")
+
+        # Get history
+        history = session.get_history()
+        ```
+    """
+
+    def chat(self, message: str, **kwargs) -> str:
+        """Send a message to the conversation session.
+
+        This is a convenience method that calls the `chat` method on the
+        underlying SessionEntity and returns just the response text.
+
+        Args:
+            message: The user's message
+            **kwargs: Additional parameters to pass to the chat method
+
+        Returns:
+            The assistant's response as a string
+
+        Example:
+            ```python
+            response = session.chat("What is the weather today?")
+            print(response)
+            ```
+        """
+        # Call the chat method via the entity proxy
+        result = self.__getattr__("chat")(message=message, **kwargs)
+
+        # SessionEntity.chat() returns a dict with 'response' key
+        if isinstance(result, dict) and "response" in result:
+            return result["response"]
+
+        # If it's already a string, return as-is
+        return str(result)
+
+    def get_history(self) -> list:
+        """Get the conversation history for this session.
+
+        Returns:
+            List of message dictionaries with 'role' and 'content' keys
+
+        Example:
+            ```python
+            history = session.get_history()
+            for msg in history:
+                print(f"{msg['role']}: {msg['content']}")
+            ```
+        """
+        return self.__getattr__("get_history")()
+
+    def add_message(self, role: str, content: str) -> dict:
+        """Add a message to the conversation history.
+
+        Args:
+            role: Message role ('user', 'assistant', or 'system')
+            content: Message content
+
+        Returns:
+            Dictionary confirming the message was added
+
+        Example:
+            ```python
+            session.add_message("system", "You are a helpful assistant")
+            session.add_message("user", "Hello!")
+            ```
+        """
+        return self.__getattr__("add_message")(role=role, content=content)
+
+    def clear_history(self) -> dict:
+        """Clear the conversation history for this session.
+
+        Returns:
+            Dictionary confirming the history was cleared
+
+        Example:
+            ```python
+            session.clear_history()
+            ```
+        """
+        return self.__getattr__("clear_history")()
+
+
+class WorkflowProxy:
+    """Proxy for invoking workflows with a fluent API.
+
+    Provides convenient methods for workflow execution, including
+    a chat() method for multi-turn conversation workflows.
+
+    Example:
+        ```python
+        # Standard workflow
+        result = client.workflow("order_process").run(order_id="123")
+
+        # Chat workflow
+        response = client.workflow("support_bot").chat(
+            message="Help me",
+            session_id="user-123",
+        )
+        ```
+    """
+
+    def __init__(self, client: "Client", workflow_name: str):
+        """Initialize workflow proxy.
+
+        Args:
+            client: The AGNT5 client instance
+            workflow_name: Name of the workflow
+        """
+        self._client = client
+        self._workflow_name = workflow_name
+
+    def run(
+        self,
+        session_id: Optional[str] = None,
+        user_id: Optional[str] = None,
+        **kwargs,
+    ) -> Dict[str, Any]:
+        """Execute the workflow synchronously.
+
+        Args:
+            session_id: Session identifier for multi-turn workflows (optional)
+            user_id: User identifier for user-scoped memory (optional)
+            **kwargs: Input parameters for the workflow
+
+        Returns:
+            Dictionary containing the workflow's output
+
+        Example:
+            ```python
+            result = client.workflow("order_process").run(
+                order_id="123",
+                customer_id="cust-456",
+            )
+            ```
+        """
+        return self._client.run(
+            component=self._workflow_name,
+            input_data=kwargs,
+            component_type="workflow",
+            session_id=session_id,
+            user_id=user_id,
+        )
+
+    def chat(
+        self,
+        message: str,
+        session_id: Optional[str] = None,
+        user_id: Optional[str] = None,
+        **kwargs,
+    ) -> Dict[str, Any]:
+        """Send a message to a chat-enabled workflow.
+
+        This is a convenience method for multi-turn conversation workflows.
+        The message is passed as the 'message' input parameter.
+
+        Args:
+            message: The user's message
+            session_id: Session identifier for conversation continuity (recommended)
+            user_id: User identifier for user-scoped memory (optional)
+            **kwargs: Additional input parameters for the workflow
+
+        Returns:
+            Dictionary containing the workflow's response (typically has 'response' key)
+
+        Example:
+            ```python
+            # First message
+            result = client.workflow("support_bot").chat(
+                message="My order hasn't arrived",
+                session_id="session-123",
+            )
+            print(result.get("response"))
+
+            # Continue conversation
+            result = client.workflow("support_bot").chat(
+                message="Can you track it?",
+                session_id="session-123",
+            )
+            ```
+        """
+        # Merge message into kwargs
+        input_data = {"message": message, **kwargs}
+
+        return self._client.run(
+            component=self._workflow_name,
+            input_data=input_data,
+            component_type="workflow",
+            session_id=session_id,
+            user_id=user_id,
+        )
+
+    def stream_events(
+        self,
+        session_id: Optional[str] = None,
+        user_id: Optional[str] = None,
+        timeout: float = 300.0,
+        **kwargs,
+    ) -> Iterator[Event]:
+        """Stream typed Event objects from workflow execution.
+
+        This method yields Event objects as they arrive from the workflow,
+        including nested events from agents and functions called within the workflow.
+
+        Args:
+            session_id: Session identifier for multi-turn workflows (optional)
+            user_id: User identifier for user-scoped memory (optional)
+            timeout: Stream timeout in seconds (default: 300.0 / 5 minutes)
+            **kwargs: Input parameters for the workflow
+
+        Yields:
+            Event objects as they arrive from the stream
+
+        Example:
+            ```python
+            from agnt5 import Client, EventType
+
+            # Stream workflow events
+            for event in client.workflow("research_workflow").stream_events(query="AI"):
+                if event.event_type == EventType.WORKFLOW_STEP_STARTED:
+                    print(f"Step started: {event.data.get('step_name')}")
+                elif event.event_type == EventType.LM_MESSAGE_DELTA:
+                    print(event.data['content'], end='', flush=True)
+                elif event.event_type == EventType.WORKFLOW_STEP_COMPLETED:
+                    print(f"\\nStep done: {event.data.get('step_name')}")
+            ```
+        """
+        return self._client.stream_events(
+            component=self._workflow_name,
+            input_data=kwargs,
+            component_type="workflow",
+            session_id=session_id,
+            user_id=user_id,
+            timeout=timeout,
+        )
+
+    def submit(self, **kwargs) -> str:
+        """Submit the workflow for async execution.
+
+        Args:
+            **kwargs: Input parameters for the workflow
+
+        Returns:
+            Run ID for tracking the execution
+
+        Example:
+            ```python
+            run_id = client.workflow("long_process").submit(data="...")
+            # Check status later
+            status = client.get_status(run_id)
+            ```
+        """
+        return self._client.submit(
+            component=self._workflow_name,
+            input_data=kwargs,
+            component_type="workflow",
+        )
+
+
+class AsyncClient:
+    """Async client for invoking AGNT5 components.
+
+    This client provides an async interface for calling functions, workflows,
+    and other components deployed on AGNT5. Use this when you need to stream
+    events in an async context or integrate with async frameworks.
+
+    Example:
+        ```python
+        import asyncio
+        from agnt5 import AsyncClient, EventType
+
+        async def main():
+            async with AsyncClient() as client:
+                # Stream agent events asynchronously
+                async for event in client.stream_events("my_agent", {"msg": "Hi"}, "agent"):
+                    if event.event_type == EventType.LM_MESSAGE_DELTA:
+                        print(event.data['content'], end='', flush=True)
+
+        asyncio.run(main())
+        ```
+    """
+
+    def __init__(
+        self,
+        gateway_url: str = "http://localhost:34181",
+        timeout: float = 30.0,
+        api_key: Optional[str] = None,
+    ):
+        """Initialize the async AGNT5 client.
+
+        Args:
+            gateway_url: Base URL of the AGNT5 gateway (default: http://localhost:34181)
+            timeout: Request timeout in seconds (default: 30.0)
+            api_key: Service key for authentication. If not provided, falls back to
+                     AGNT5_API_KEY environment variable. Keys start with "agnt5_sk_".
+        """
+        self.gateway_url = gateway_url.rstrip("/")
+        self.timeout = timeout
+        # Use provided api_key or fallback to environment variable
+        self.api_key = api_key or os.environ.get(AGNT5_API_KEY_ENV)
+        self._client: Optional[httpx.AsyncClient] = None
+
+    def _build_headers(
+        self,
+        session_id: Optional[str] = None,
+        user_id: Optional[str] = None,
+    ) -> Dict[str, str]:
+        """Build request headers with authentication and optional session/user context.
+
+        Args:
+            session_id: Session identifier for multi-turn conversations
+            user_id: User identifier for user-scoped memory
+
+        Returns:
+            Dictionary of HTTP headers
+        """
+        headers = {"Content-Type": "application/json"}
+        if self.api_key:
+            headers["X-API-KEY"] = self.api_key
+        if session_id:
+            headers["X-Session-ID"] = session_id
+        if user_id:
+            headers["X-User-ID"] = user_id
+        return headers
+
+    async def __aenter__(self) -> "AsyncClient":
+        """Async context manager entry."""
+        self._client = httpx.AsyncClient(timeout=self.timeout)
+        return self
+
+    async def __aexit__(self, exc_type, exc_val, exc_tb) -> None:
+        """Async context manager exit."""
+        if self._client:
+            await self._client.aclose()
+            self._client = None
+
+    async def _ensure_client(self) -> httpx.AsyncClient:
+        """Ensure async client is available."""
+        if self._client is None:
+            self._client = httpx.AsyncClient(timeout=self.timeout)
+        return self._client
+
+    async def close(self) -> None:
+        """Close the underlying async HTTP client."""
+        if self._client:
+            await self._client.aclose()
+            self._client = None
+
+    async def run(
+        self,
+        component: str,
+        input_data: Optional[Dict[str, Any]] = None,
+        component_type: str = "function",
+        session_id: Optional[str] = None,
+        user_id: Optional[str] = None,
+    ) -> Dict[str, Any]:
+        """Execute a component asynchronously and wait for the result.
+
+        Args:
+            component: Name of the component to execute
+            input_data: Input data for the component
+            component_type: Type of component - "function", "workflow", "agent", "tool"
+            session_id: Session identifier for multi-turn conversations
+            user_id: User identifier for user-scoped memory
+
+        Returns:
+            Dictionary containing the component's output
+
+        Raises:
+            RunError: If the component execution fails
+            httpx.HTTPError: If the HTTP request fails
+        """
+        if input_data is None:
+            input_data = {}
+
+        client = await self._ensure_client()
+        url = urljoin(self.gateway_url + "/", f"v1/run/{component_type}/{component}")
+
+        response = await client.post(
+            url,
+            json=input_data,
+            headers=self._build_headers(session_id=session_id, user_id=user_id),
+        )
+
+        if response.status_code == 404:
+            try:
+                error_data = response.json()
+                raise RunError(
+                    error_data.get("error", "Component not found"),
+                    run_id=error_data.get("runId"),
+                )
+            except ValueError:
+                raise RunError(f"Component '{component}' not found")
+
+        if response.status_code in (500, 503, 504):
+            try:
+                error_data = response.json()
+                raise RunError(
+                    error_data.get("error", "Unknown error"),
+                    run_id=error_data.get("runId"),
+                )
+            except ValueError:
+                response.raise_for_status()
+        else:
+            response.raise_for_status()
+
+        data = response.json()
+        if data.get("status") == "failed":
+            raise RunError(data.get("error", "Unknown error"), run_id=data.get("runId"))
+
+        return data.get("output", {})
+
+    async def stream_events(
+        self,
+        component: str,
+        input_data: Optional[Dict[str, Any]] = None,
+        component_type: str = "function",
+        session_id: Optional[str] = None,
+        user_id: Optional[str] = None,
+        timeout: float = 300.0,
+    ) -> AsyncIterator[Event]:
+        """Async stream typed Event objects from a component execution.
+
+        This method yields Event objects as they arrive from the component,
+        providing full access to the event taxonomy including agent lifecycle,
+        LM streaming, tool calls, and workflow events.
+
+        Args:
+            component: Name of the component to execute
+            input_data: Input data for the component
+            component_type: Type of component - "function", "workflow", "agent", "tool"
+            session_id: Session identifier for multi-turn conversations
+            user_id: User identifier for user-scoped memory
+            timeout: Stream timeout in seconds (default: 300.0 / 5 minutes)
+
+        Yields:
+            Event objects as they arrive from the stream
+
+        Raises:
+            RunError: If the component execution fails
+            httpx.HTTPError: If the HTTP request fails
+
+        Example:
+            ```python
+            async with AsyncClient() as client:
+                async for event in client.stream_events("my_agent", {"msg": "Hi"}, "agent"):
+                    if event.event_type == EventType.AGENT_STARTED:
+                        print(f"Agent started: {event.data['agent_name']}")
+                    elif event.event_type == EventType.LM_MESSAGE_DELTA:
+                        print(event.data['content'], end='', flush=True)
+            ```
+        """
+        if timeout <= 0:
+            raise ValueError("timeout must be a positive number")
+
+        if input_data is None:
+            input_data = {}
+
+        client = await self._ensure_client()
+        url = urljoin(self.gateway_url + "/", f"v1/streamv2/{component_type}/{component}")
+
+        async with client.stream(
+            "POST",
+            url,
+            json=input_data,
+            headers=self._build_headers(session_id=session_id, user_id=user_id),
+            timeout=timeout,
+        ) as response:
+            if response.status_code != 200:
+                # Try to get error details from response body
+                try:
+                    error_body = (await response.aread()).decode("utf-8")
+                    error_data = json.loads(error_body)
+                    error_msg = error_data.get("error", f"HTTP {response.status_code}")
+                    run_id = error_data.get("runId")
+                except (json.JSONDecodeError, UnicodeDecodeError):
+                    error_msg = f"HTTP {response.status_code}: Streaming request failed"
+                    run_id = None
+                raise RunError(error_msg, run_id=run_id)
+
+            current_event_type: Optional[str] = None
+            async for line in response.aiter_lines():
+                line = line.strip()
+
+                # Skip empty lines and comments (keep-alive)
+                if not line or line.startswith(":"):
+                    continue
+
+                # Parse event type: "event: agent.started"
+                if line.startswith("event: "):
+                    current_event_type = line[7:]
+                    continue
+
+                # Parse SSE data: "data: {...}"
+                if line.startswith("data: "):
+                    data_str = line[6:]
+
+                    try:
+                        data = json.loads(data_str)
+
+                        # Check for completion signal
+                        if data.get("done") or current_event_type == "done":
+                            return
+
+                        # Check for error event
+                        if current_event_type == "error" or "error" in data:
+                            error_msg = data.get("error", "Unknown streaming error")
+                            raise RunError(error_msg, run_id=data.get("runId"))
+
+                        # Yield typed Event object
+                        if current_event_type:
+                            yield _parse_sse_to_event(current_event_type, data)
+
+                    except json.JSONDecodeError:
+                        continue
+
+    async def submit(
+        self,
+        component: str,
+        input_data: Optional[Dict[str, Any]] = None,
+        component_type: str = "function",
+    ) -> str:
+        """Submit a component for async execution and return immediately.
+
+        Args:
+            component: Name of the component to execute
+            input_data: Input data for the component
+            component_type: Type of component
+
+        Returns:
+            String containing the run ID
+        """
+        if input_data is None:
+            input_data = {}
+
+        client = await self._ensure_client()
+        url = urljoin(self.gateway_url + "/", f"v1/submit/{component_type}/{component}")
+
+        response = await client.post(
+            url,
+            json=input_data,
+            headers=self._build_headers(),
+        )
+        response.raise_for_status()
+
+        data = response.json()
+        return data.get("runId", "")
+
+    async def get_status(self, run_id: str) -> Dict[str, Any]:
+        """Get the current status of a run.
+
+        Args:
+            run_id: The run ID returned from submit()
+
+        Returns:
+            Dictionary containing status information
+        """
+        client = await self._ensure_client()
+        url = urljoin(self.gateway_url + "/", f"v1/status/{run_id}")
+
+        response = await client.get(url, headers=self._build_headers())
+        response.raise_for_status()
+
+        return response.json()
+
+    async def get_result(self, run_id: str) -> Dict[str, Any]:
+        """Get the result of a completed run.
+
+        Args:
+            run_id: The run ID returned from submit()
+
+        Returns:
+            Dictionary containing the component's output
+
+        Raises:
+            RunError: If the run failed or is not yet complete
+        """
+        client = await self._ensure_client()
+        url = urljoin(self.gateway_url + "/", f"v1/result/{run_id}")
+
+        response = await client.get(url, headers=self._build_headers())
+
+        if response.status_code == 404:
+            error_data = response.json()
+            error_msg = error_data.get("error", "Run not found or not complete")
+            current_status = error_data.get("status", "unknown")
+            raise RunError(f"{error_msg} (status: {current_status})", run_id=run_id)
+
+        response.raise_for_status()
+        data = response.json()
+
+        if data.get("status") == "failed":
+            raise RunError(data.get("error", "Unknown error"), run_id=run_id)
+
+        return data.get("output", {})
+
+
+class RunError(Exception):
+    """Raised when a component run fails on AGNT5.
+
+    Attributes:
+        message: Error message describing what went wrong
+        run_id: The unique run ID associated with this execution (if available)
+    """
+
+    def __init__(self, message: str, run_id: Optional[str] = None):
+        super().__init__(message)
+        self.run_id = run_id
+        self.message = message
+
+    def __str__(self):
+        if self.run_id:
+            return f"{self.message} (run_id: {self.run_id})"
+        return self.message