dapr default better devex

dapr_agents/agents/durableagent/agent.py

@@ -120,10 +120,6 @@ def model_post_init(self, __context: Any) -> None:
 
         self.register_agentic_system()
 
-        # Start the runtime if it's not already running
-        logger.info("Starting workflow runtime...")
-        self.start_runtime()
-
     async def run(self, input_data: Union[str, Dict[str, Any]]) -> Any:
         """
         Fire up the workflow, wait for it to complete, then return the final serialized_output.

dapr_agents/llm/dapr/chat.py

@@ -15,7 +15,6 @@
 )
 
 from pydantic import BaseModel, Field
-
 from dapr_agents.llm.chat import ChatClientBase
 from dapr_agents.llm.dapr.client import DaprInferenceClientBase
 from dapr_agents.llm.utils import RequestHandler, ResponseHandler
@@ -80,6 +79,8 @@ class DaprChatClient(DaprInferenceClientBase, ChatClientBase):
 
     component_name: Optional[str] = None
 
+    component_name: Optional[str] = None
+
     # Only function_call–style structured output is supported
     SUPPORTED_STRUCTURED_MODES: ClassVar[set[str]] = {"function_call"}
 
@@ -92,8 +93,8 @@ def model_post_init(self, __context: Any) -> None:
         if not self._llm_component:
             self._llm_component = os.environ.get("DAPR_LLM_COMPONENT_DEFAULT")
         if not self._llm_component:
-            raise ValueError(
-                "You must provide a component_name or set DAPR_LLM_COMPONENT_DEFAULT in the environment."
+            logger.debug(
+                "No LLM component provided and no default component found in the environment. Will try to get it from the metadata at runtime."
             )
         super().model_post_init(__context)
 
@@ -327,16 +328,11 @@ def generate(
                 )
             # get metadata information from the dapr client
             metadata = self.client.dapr_client.get_metadata()
-            extended_metadata = metadata.extended_metadata
-            dapr_runtime_version = extended_metadata.get("daprRuntimeVersion", None)
-            if dapr_runtime_version is not None:
-                # Allow only versions >=1.16.0 and <2.0.0 for Alpha2 Chat Client
-                if not is_version_supported(
-                    str(dapr_runtime_version), ">=1.16.0, edge, <2.0.0"
-                ):
-                    raise DaprRuntimeVersionNotSupportedError(
-                        f"!!!!! Dapr Runtime Version {dapr_runtime_version} is not supported with Alpha2 Dapr Chat Client. Only Dapr runtime versions >=1.16.0, edge,and <2.0.0 are supported."
-                    )
+            _check_dapr_runtime_support(metadata)
+
+            llm_component = llm_component or self._llm_component
+            if not llm_component:
+                llm_component = _get_llm_component(metadata)
 
             raw = self.client.chat_completion_alpha2(
                 llm=llm_component or self._llm_component,
@@ -365,3 +361,36 @@ def generate(
             structured_mode=structured_mode,
             stream=False,
         )
+
+
+def _check_dapr_runtime_support(metadata: "GetMetadataResponse"):  # noqa: F821
+    """Check if the Dapr runtime version is supported for Alpha2 Chat Client."""
+    extended_metadata = metadata.extended_metadata
+    dapr_runtime_version = extended_metadata.get("daprRuntimeVersion", None)
+    if dapr_runtime_version is not None:
+        # Allow only versions >=1.16.0, edge, and <2.0.0 for Alpha2 Chat Client
+        if not is_version_supported(
+            str(dapr_runtime_version), ">=1.16.0, edge, <2.0.0"
+        ):
+            raise DaprRuntimeVersionNotSupportedError(
+                f"!!!!! Dapr Runtime Version {dapr_runtime_version} is not supported with Alpha2 Dapr Chat Client. Only Dapr runtime versions >=1.16.0, edge, and <2.0.0 are supported."
+            )
+
+
+def _get_llm_component(metadata: "GetMetadataResponse") -> str:  # noqa: F821
+    """Get the LLM component from the metadata."""
+    conversation_components = [
+        component
+        for component in metadata.registered_components
+        if component.type.startswith("conversation.")
+    ]
+    if len(conversation_components) == 1:
+        return conversation_components[0].name
+    elif len(conversation_components) > 1:
+        raise ValueError(
+            "Multiple LLM components found in the metadata. Please provide the component name explicitly (e.g. llm = DaprChatClient(component_name='openai')) or environment variable DAPR_LLM_COMPONENT_DEFAULT."
+        )
+    else:
+        raise ValueError(
+            "No LLM component provided and no default component found in the metadata."
+        )

dapr_agents/workflow/base.py

@@ -3,6 +3,7 @@
 import inspect
 import json
 import logging
+import time
 import sys
 import uuid
 from datetime import datetime, timezone
@@ -516,6 +517,9 @@ def start_runtime(self):
             self.wf_runtime.start()
             self.wf_runtime_is_running = True
 
+            logger.info("Sleeping for 5 seconds to ensure runtime is started.")
+            time.sleep(5)
+
             # Sync database state with Dapr workflow status after runtime starts
             # This ensures our database reflects the actual state of resumed workflows
             self._sync_workflow_state_after_startup()

quickstarts/01-hello-world/03_durable_agent.py

@@ -9,9 +9,10 @@
 import asyncio
 import logging
 
+
 from typing import List
 from pydantic import BaseModel, Field
-from dapr_agents import tool, DurableAgent, DaprChatClient
+from dapr_agents import tool, DurableAgent
 from dapr_agents.memory import ConversationDaprStateMemory
 from dotenv import load_dotenv
 
@@ -38,11 +39,17 @@ def search_flights(destination: str) -> List[FlightOption]:
     ]
 
 
-# one can use the environment variable to set the default component name
+# ----------------------------------------------------------------------------------------------------------------------
+# There are three ways to set the LLM component with DaprChatClient:
+#
+# 1. Explicitly instantiate the DaprChatClient with the component name
+# llm = DaprChatClient(component_name="openai")
+#
+# 2. Use the environment variable DAPR_LLM_COMPONENT_DEFAULT
 # os.environ.setdefault("DAPR_LLM_COMPONENT_DEFAULT", "openai")
-
-# or directly pass the component name
-llm = DaprChatClient(component_name="openai")
+#
+# 3. If there is only one conversation component in the resources folder, it will be used by default
+# ----------------------------------------------------------------------------------------------------------------------
 
 
 async def main():
@@ -66,7 +73,7 @@ async def main():
             memory=ConversationDaprStateMemory(
                 store_name="conversationstore", session_id="my-unique-id"
             ),
-            llm=llm,
+            # llm=llm, # if you don't set the llm attribute, it will be by default set to DaprChatClient()
         )
 
         await travel_planner.run("I want to find flights to Paris")

quickstarts/01-hello-world/README.md

@@ -4,15 +4,25 @@ This quickstart provides a hands-on introduction to Dapr Agents through simple e
 
 ## Prerequisites
 
-- Python 3.10 (recommended)
-- pip package manager
+- Python 3.10+ (recommended)
+- [uv package manager](https://docs.astral.sh/uv/getting-started/installation/)
 - OpenAI API key (you can put in an .env file or directly in the `openai.yaml` file, but we recommend the .env file that is gitignored)
 
 ## Environment Setup
 
 <details open>
 <summary><strong>Option 1: Using uv (Recommended)</strong></summary>
 
+<!-- We include setting up the venv as part of the first step to make sure the venv is created and activated before the examples are run.-->
+
+<!-- STEP
+name: Run basic LLM example
+expected_stdout_lines:
+  - "Got response:"
+timeout_seconds: 30
+output_match_mode: substring
+-->
+
 ```bash
 # Create and activate virtual environment
 uv venv .venv
@@ -27,7 +37,7 @@ uv pip install -r requirements.txt
 <details>
 <summary><strong>Option 2: Using pip</strong></summary>
 
-```bash
+```a shell [not setting type to avoid mechanical markdown execution]
 # Create a virtual environment
 python3.10 -m venv .venv
 
@@ -39,17 +49,19 @@ source .venv/bin/activate
 
 # Install dependencies
 pip install -r requirements.txt
-
 ```
 
 </details>
 
 
-## Configuration
+## OpenAI API Key
 
 > **Warning**
 > The examples will not work if you do not have a OpenAI API key exported in the environment.
 
+<details open>
+<summary><strong>Option 1: Using .env file</strong></summary>
+
 Create a `.env` file in the project root and add your OpenAI API key:
 
 ```env
@@ -59,23 +71,32 @@ OPENAI_API_KEY=your_api_key_here
 Replace `your_api_key_here` with your actual OpenAI API key.
 
 Export the environment variables from the .env file to your shell:
-```bash
-export $(grep -v '^#' .env | xargs) # or if .env is in the root directory, you can just run `export $(grep -v '^#' ../../.env | xargs)`
+```a shell [not setting type to avoid mechanical markdown execution]
+export $(grep -v '^#' .env | xargs) 
+
+# or if .env is in the root directory of the repository, 
+# export $(grep -v '^#' ../../.env | xargs)
+```
+
+</details>
+
+<details>
+<summary><strong>Option 2: Exporting the OpenAI API Key directly to the shell</strong></summary>
+
+```a shell [not setting type to avoid mechanical markdown execution]
+export OPENAI_API_KEY=your_api_key_here
 ```
 
+Replace `your_api_key_here` with your actual OpenAI API key.
+
+</details>
+
 ## Examples
 
 ### 1. Basic LLM Usage
 
 Run the basic LLM example to see how to interact with OpenAI's language models:
 
-<!-- STEP
-name: Run basic LLM example
-expected_stdout_lines:
-  - "Got response:"
-timeout_seconds: 30
-output_match_mode: substring
--->
 ```bash
 python 01_ask_llm.py
 ```
@@ -167,6 +188,9 @@ A stateful agent that uses Dapr Workflows to ensure durability and persistence o
 We are using the Dapr ChatClient to interact with the OpenAI API. In the components folder, we have a `openai.yaml` file that contains the configuration for the OpenAI API.
 You need to replace the `{YOUR_OPENAI_API_KEY}` with your actual OpenAI API key.
 
+We are using the Dapr ChatClient to interact with the OpenAI API. In the components folder, we have a `openai.yaml` file that contains the configuration for the OpenAI API.
+You need to replace the `{YOUR_OPENAI_API_KEY}` with your actual OpenAI API key.
+
 Make sure Dapr is initialized on your system:
 
 ```bash
@@ -180,14 +204,16 @@ name: Run basic LLM example
 expected_stdout_lines:
   - "I want to find flights to Paris"
   - "TravelBuddy"
-timeout_seconds: 30
+timeout_seconds: 60
 output_match_mode: substring
 -->
 
 
-We are using the `resolve_env_templates.py` script to resolve the environment variables in the components folder and substitute them with the actual values in your .env file, like the OpenAI API key.
+We are using the `resolve_env_templates.py` script to resolve the environment variables in the components folder and substitute them with the actual values in your environment, like the OpenAI API key.
 
 ```bash
+source .venv/bin/activate
+
 dapr run --app-id stateful-llm --dapr-http-port 3500 --resources-path $(../resolve_env_templates.py ./components) -- python 03_durable_agent.py
 ```
 
@@ -321,6 +347,8 @@ expected_stdout_lines:
 output_match_mode: substring
 -->
 ```bash
+source .venv/bin/activate
+
 dapr run --app-id dapr-agent-wf --resources-path $(../resolve_env_templates.py ./components) -- python 04_chain_tasks.py
 ```
 <!-- END_STEP -->
@@ -402,6 +430,8 @@ expected_stdout_lines:
 output_match_mode: substring
 -->
 ```bash
+source .venv/bin/activate
+
 python 05_agent_with_vectorstore.py
 ```
 <!-- END_STEP -->
@@ -541,6 +571,7 @@ if __name__ == "__main__":
 
 ## Key Concepts
 
+- **DaprChatClient**: The interface for interacting with Dapr's LLMs
 - **OpenAIChatClient**: The interface for interacting with OpenAI's LLMs
 - **Agent**: A class that combines an LLM with tools and instructions
 - **@tool decorator**: A way to create tools that agents can use