Merge pull request #36 from Contextable/issue-24-cleanup-agent-registration

feat: eliminate AgentRegistry and implement direct agent embedding

Author: contextablemark

typescript-sdk/integrations/adk-middleware/CHANGELOG.md

@@ -7,6 +7,29 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.5.0] - 2025-08-05
+
+### Breaking Changes
+- **BREAKING**: ADKAgent constructor now requires `adk_agent` parameter instead of `agent_id` for direct agent embedding
+- **BREAKING**: Removed AgentRegistry dependency - agents are now directly embedded in middleware instances
+- **BREAKING**: Removed `agent_id` parameter from `ADKAgent.run()` method
+- **BREAKING**: Endpoint registration no longer extracts agent_id from URL path
+- **BREAKING**: AgentRegistry class removed from public API
+
+### Architecture Improvements
+- **ARCHITECTURE**: Eliminated AgentRegistry entirely - simplified architecture by embedding ADK agents directly
+- **ARCHITECTURE**: Cleaned up agent registration/instantiation redundancy (issue #24)
+- **ARCHITECTURE**: Removed confusing indirection where endpoint agent didn't determine execution
+- **ARCHITECTURE**: Each ADKAgent instance now directly holds its ADK agent instance
+- **ARCHITECTURE**: Simplified method signatures and removed agent lookup overhead
+
+### Fixed
+- **FIXED**: All 271 tests now pass with new simplified architecture
+- **TESTS**: Updated all test fixtures to match new ADKAgent.run(input_data) signature without agent_id parameter
+- **TESTS**: Fixed test expectations in test_endpoint.py to work with direct agent embedding architecture
+- **TESTS**: Updated all test fixtures to work with new agent embedding pattern
+- **EXAMPLES**: Updated examples to demonstrate direct agent embedding pattern
+
 ### Added
 - **NEW**: SystemMessage support for ADK agents (issue #22) - SystemMessages as first message are now appended to agent instructions
 - **NEW**: Comprehensive tests for SystemMessage functionality including edge cases

typescript-sdk/integrations/adk-middleware/LOGGING.md

@@ -38,7 +38,7 @@ import logging
 logging.getLogger().setLevel(logging.DEBUG)
 
 # Or configure specific components
-components = ['adk_agent', 'event_translator', 'endpoint', 'session_manager', 'agent_registry']
+components = ['adk_agent', 'event_translator', 'endpoint', 'session_manager']
 for component in components:
     logging.getLogger(component).setLevel(logging.DEBUG)
 ```
@@ -51,7 +51,6 @@ for component in components:
 | `endpoint` | HTTP endpoint responses | WARNING |
 | `adk_agent` | Main agent logic | INFO |
 | `session_manager` | Session management | WARNING |
-| `agent_registry` | Agent registration | WARNING |
 
 ## Python API
 
@@ -128,8 +127,7 @@ components = {
     'adk_agent': os.getenv('LOG_ADK_AGENT', 'INFO'),
     'event_translator': os.getenv('LOG_EVENT_TRANSLATOR', 'WARNING'),
     'endpoint': os.getenv('LOG_ENDPOINT', 'WARNING'),
-    'session_manager': os.getenv('LOG_SESSION_MANAGER', 'WARNING'),
-    'agent_registry': os.getenv('LOG_AGENT_REGISTRY', 'WARNING')
+    'session_manager': os.getenv('LOG_SESSION_MANAGER', 'WARNING')
 }
 
 for component, level in components.items():

typescript-sdk/integrations/adk-middleware/README.md

@@ -7,7 +7,7 @@ This Python middleware enables Google ADK agents to be used with the AG-UI Proto
 - ⚠️ Full event translation between AG-UI and ADK (partial - full support coming soon)
 - ✅ Automatic session management with configurable timeouts
 - ✅ Automatic session memory option - expired sessions automatically preserved in ADK memory service
-- ✅ Support for multiple agents with centralized registry
+- ✅ Support for multiple agents with direct embedding
 - ❌ State synchronization between protocols (coming soon)
 - ✅ **Complete bidirectional tool support** - Enable AG-UI Protocol tools within Google ADK agents
 - ✅ Streaming responses with SSE
@@ -49,7 +49,7 @@ Although this is a Python integration, it lives in `typescript-sdk/integrations/
 
 ### Option 1: Direct Usage
 ```python
-from adk_middleware import ADKAgent, AgentRegistry
+from adk_middleware import ADKAgent
 from google.adk.agents import Agent
 
 # 1. Create your ADK agent
@@ -58,30 +58,38 @@ my_agent = Agent(
     instruction="You are a helpful assistant."
 )
 
-# 2. Register the agent
-registry = AgentRegistry.get_instance()
-registry.set_default_agent(my_agent)
-
-# 3. Create the middleware
-agent = ADKAgent(app_name="my_app", user_id="user123")
+# 2. Create the middleware with direct agent embedding
+agent = ADKAgent(
+    adk_agent=my_agent,
+    app_name="my_app", 
+    user_id="user123"
+)
 
-# 4. Use directly with AG-UI RunAgentInput
+# 3. Use directly with AG-UI RunAgentInput
 async for event in agent.run(input_data):
     print(f"Event: {event.type}")
 ```
 
 ### Option 2: FastAPI Server
 ```python
 from fastapi import FastAPI
-from adk_middleware import ADKAgent, AgentRegistry, add_adk_fastapi_endpoint
+from adk_middleware import ADKAgent, add_adk_fastapi_endpoint
 from google.adk.agents import Agent
 
-# Set up agent and registry (same as above)
-registry = AgentRegistry.get_instance()
-registry.set_default_agent(my_agent)
-agent = ADKAgent(app_name="my_app", user_id="user123")
+# 1. Create your ADK agent
+my_agent = Agent(
+    name="assistant",
+    instruction="You are a helpful assistant."
+)
+
+# 2. Create the middleware with direct agent embedding
+agent = ADKAgent(
+    adk_agent=my_agent,
+    app_name="my_app", 
+    user_id="user123"
+)
 
-# Create FastAPI app
+# 3. Create FastAPI app
 app = FastAPI()
 add_adk_fastapi_endpoint(app, agent, path="/chat")
 
@@ -90,32 +98,15 @@ add_adk_fastapi_endpoint(app, agent, path="/chat")
 
 ## Configuration Options
 
-### Agent Registry
-
-The `AgentRegistry` provides flexible agent mapping:
-
-```python
-registry = AgentRegistry.get_instance()
-
-# Option 1: Default agent for all requests
-registry.set_default_agent(my_agent)
-
-# Option 2: Map specific agent IDs
-registry.register_agent("support", support_agent)
-registry.register_agent("coder", coding_agent)
-
-# Option 3: Dynamic agent creation
-def create_agent(agent_id: str) -> BaseAgent:
-    return Agent(name=agent_id, instruction="You are a helpful assistant.")
-
-registry.set_agent_factory(create_agent)
-```
-
 ### App and User Identification
 
 ```python
 # Static app name and user ID (single-tenant apps)
-agent = ADKAgent(app_name="my_app", user_id="static_user")
+agent = ADKAgent(
+    adk_agent=my_agent,
+    app_name="my_app", 
+    user_id="static_user"
+)
 
 # Dynamic extraction from context (recommended for multi-tenant)
 def extract_app(input: RunAgentInput) -> str:
@@ -133,6 +124,7 @@ def extract_user(input: RunAgentInput) -> str:
     return f"anonymous_{input.thread_id}"
 
 agent = ADKAgent(
+    adk_agent=my_agent,
     app_name_extractor=extract_app,
     user_id_extractor=extract_user
 )
@@ -219,19 +211,16 @@ my_agent = Agent(
     tools=[adk_tools.preload_memory_tool.PreloadMemoryTool()]  # Add memory tools here
 )
 
-# Register the agent
-registry = AgentRegistry.get_instance()
-registry.set_default_agent(my_agent)
-
-# Create middleware WITHOUT tools parameter - THIS IS CORRECT  
+# Create middleware with direct agent embedding
 adk_agent = ADKAgent(
+    adk_agent=my_agent,
     app_name="my_app",
     user_id="user123",
     memory_service=shared_memory_service  # Memory service enables automatic session memory
 )
 ```
 
-**⚠️ Important**: The `tools` parameter belongs to the ADK agent (like `Agent` or `LlmAgent`), **not** to the `ADKAgent` middleware. The middleware automatically handles any tools defined on the registered agents.
+**⚠️ Important**: The `tools` parameter belongs to the ADK agent (like `Agent` or `LlmAgent`), **not** to the `ADKAgent` middleware. The middleware automatically handles any tools defined on the embedded agents.
 
 ### Memory Testing Configuration
 
@@ -377,7 +366,7 @@ toolset = ClientProxyToolset(
 ### Tool Configuration
 
 ```python
-from adk_middleware import ADKAgent, AgentRegistry
+from adk_middleware import ADKAgent
 from google.adk.agents import LlmAgent
 from ag_ui.core import RunAgentInput, UserMessage, Tool
 
@@ -431,11 +420,9 @@ agent = LlmAgent(
     Use calculate for math operations and get_weather for weather information."""
 )
 
-registry = AgentRegistry.get_instance()
-registry.set_default_agent(agent)
-
 # 3. Create middleware with hybrid execution configuration
 adk_agent = ADKAgent(
+    adk_agent=agent,
     user_id="user123",
     tool_timeout_seconds=60,       # Timeout for blocking tools only
     execution_timeout_seconds=300, # Overall execution timeout
@@ -664,18 +651,19 @@ This will start a FastAPI server that connects your ADK middleware to the Dojo a
 
 ```python
 import asyncio
-from adk_middleware import ADKAgent, AgentRegistry
+from adk_middleware import ADKAgent
 from google.adk.agents import Agent
 from ag_ui.core import RunAgentInput, UserMessage
 
 async def main():
     # Setup
-    registry = AgentRegistry.get_instance()
-    registry.set_default_agent(
-        Agent(name="assistant", instruction="You are a helpful assistant.")
-    )
+    my_agent = Agent(name="assistant", instruction="You are a helpful assistant.")
 
-    agent = ADKAgent(app_name="demo_app", user_id="demo")
+    agent = ADKAgent(
+        adk_agent=my_agent,
+        app_name="demo_app", 
+        user_id="demo"
+    )
 
     # Create input
     input = RunAgentInput(
@@ -702,17 +690,33 @@ asyncio.run(main())
 ### Multi-Agent Setup
 
 ```python
-# Register multiple agents
-registry = AgentRegistry.get_instance()
-registry.register_agent("general", general_agent)
-registry.register_agent("technical", technical_agent)
-registry.register_agent("creative", creative_agent)
+# Create multiple agent instances with different ADK agents
+general_agent_wrapper = ADKAgent(
+    adk_agent=general_agent,
+    app_name="demo_app",
+    user_id="demo"
+)
 
-# The middleware uses the default agent from the registry
-agent = ADKAgent(
+technical_agent_wrapper = ADKAgent(
+    adk_agent=technical_agent,
+    app_name="demo_app",
+    user_id="demo"
+)
+
+creative_agent_wrapper = ADKAgent(
+    adk_agent=creative_agent,
     app_name="demo_app",
-    user_id="demo"  # Or use user_id_extractor for dynamic extraction
+    user_id="demo"
 )
+
+# Use different endpoints for each agent
+from fastapi import FastAPI
+from adk_middleware import add_adk_fastapi_endpoint
+
+app = FastAPI()
+add_adk_fastapi_endpoint(app, general_agent_wrapper, path="/agents/general")
+add_adk_fastapi_endpoint(app, technical_agent_wrapper, path="/agents/technical")
+add_adk_fastapi_endpoint(app, creative_agent_wrapper, path="/agents/creative")
 ```
 
 ## Event Translation