Feature: Application Configuration, Templating, and Lifecycle
  As a developer
  I want the application to initialize with templates, register agents and routes, process single-shot and interactive sessions, and dispose cleanly
  So that the ReactiveCleverAgentsApp lifecycle works correctly from startup to shutdown

  Background:
    Given the application test environment is initialized

  Scenario: Application initialization with prompt template processing
    Given I have a configuration with prompt templates:
      """
      cleveragents:
        template_engine: JINJA2

      agents:
        template_agent:
          type: llm
          config:
            provider: openai
            model: gpt-3.5-turbo

      prompts:
        greeting:
          content: "Hello {{name}}"
        simple_prompt: "Just a string"
        dict_prompt:
          content: "Dict content"
          metadata: "extra"
      """
    When I load the configuration
    Then the application should initialize successfully
    And prompt templates should be registered correctly
    And both string and dict prompts should be processed

  Scenario: Enhanced template registry initialization
    Given I have a configuration with advanced complex templates requiring preprocessing:
      """
      agents:
        test_agent:
          type: llm
          config:
            provider: openai
            model: gpt-3.5-turbo

      templates:
        agents:
          complex_agent:
            _needs_preprocessing: true
            _raw_template: "{{agent_type}}"
            type: "{{agent_type}}"
        graphs:
          simple_graph:
            __jinja_template__: true
            type: graph
        streams:
          template_stream:
            __is_template__: true
            type: stream
      """
    When I load the configuration
    Then the enhanced template registry should be used
    And complex templates should be processed correctly

  Scenario: Regular template registry for simple templates
    Given I have a configuration with simple templates:
      """
      agents:
        simple_agent:
          type: llm
          config:
            provider: openai
            model: gpt-3.5-turbo

      templates:
        agents:
          basic_agent:
            type: llm
            config:
              provider: openai
      """
    When I load the configuration
    Then the regular template registry should be used
    And simple templates should be registered

  Scenario: Agent creation with template instances
    Given I have a configuration with agent template instances:
      """
      agents:
        simple_agent:
          type: llm
          config:
            provider: openai
            model: gpt-3.5-turbo

        template_instance_agent:
          type: template_instance
          config:
            template: basic_agent
            params:
              model: gpt-4

      templates:
        agents:
          basic_agent:
            type: llm
            config:
              provider: openai
              model: gpt-4
      """
    When I load the configuration
    Then agents should be created from templates
    And template instances should be processed correctly

  Scenario: Agent creation with enhanced registry template instances
    Given I have a configuration requiring enhanced registry with template instances:
      """
      agents:
        enhanced_template_agent:
          type: template_instance
          config:
            agent_template: complex_agent
            params:
              agent_type: tool

      templates:
        agents:
          complex_agent:
            _needs_preprocessing: true
            _raw_template: "tool"
            type: tool
            config:
              tools: ["echo"]
      """
    When I load the configuration
    Then the enhanced template registry should instantiate agents
    And complex template parameters should be applied

  Scenario: Bridge route configuration
    Given I have a configuration with bridge routes:
      """
      agents:
        bridge_agent:
          type: llm
          config:
            provider: openai
            model: gpt-3.5-turbo

      routes:
        bridge_route:
          type: bridge
          source: input_stream
          target: output_stream
      """
    When I load the configuration
    Then bridge routes should be registered
    And the route bridge should be initialized

  Scenario: Graph route with state class resolution
    Given I have a configuration with graph routes and state class:
      """
      agents:
        graph_agent:
          type: llm
          config:
            provider: openai
            model: gpt-3.5-turbo

      routes:
        state_graph:
          type: graph
          state_class: "builtins.dict"
          nodes:
            start:
              agent: graph_agent
          edges:
            - source: start
              target: END
          entry_point: start
      """
    When I load the configuration
    Then the state class should be resolved correctly
    And the graph should be created with the state class

  Scenario: Graph route with invalid state class
    Given I have a configuration with invalid state class:
      """
      agents:
        graph_agent:
          type: llm
          config:
            provider: openai
            model: gpt-3.5-turbo

      routes:
        invalid_state_graph:
          type: graph
          state_class: "nonexistent.module.Class"
          nodes:
            start:
              agent: graph_agent
          edges:
            - source: start
              target: END
      """
    When I load the configuration
    Then a warning should be logged about the invalid state class
    And the graph should still be created

  Scenario: Route template instantiation
    Given I have a configuration with route templates:
      """
      agents:
        template_route_agent:
          type: llm
          config:
            provider: openai
            model: gpt-3.5-turbo

      routes:
        templated_route:
          template_config:
            template: stream_template
            params:
              agent_name: template_route_agent

      templates:
        streams:
          stream_template:
            type: stream
            stream_type: cold
            agents:
              - "{{agent_name}}"
      """
    When I load the configuration
    Then route templates should be instantiated
    And template parameters should be applied to routes

  Scenario: Stream operations setup with merges and splits
    Given I have a configuration with merge and split operations:
      """
      agents:
        merge_agent:
          type: llm
          config:
            provider: openai
            model: gpt-3.5-turbo

      routes:
        source1:
          type: stream
          stream_type: cold
        source2:
          type: stream
          stream_type: cold
        target:
          type: stream
          stream_type: cold
          operators:
            - type: map
              params:
                agent: merge_agent
        split_stream:
          type: stream
          stream_type: cold

      merges:
        - sources: [source1, source2]
          target: target

      splits:
        - source: split_stream
          targets:
            positive: "content.startswith('good')"
            negative: "content.startswith('bad')"
      """
    When I load the configuration
    Then merges should be set up correctly
    And splits should be configured properly
    And subscriptions should be re-setup after operations

  Scenario: Hybrid pipeline setup
    Given I have a configuration with hybrid pipelines:
      """
      agents:
        pipeline_agent:
          type: llm
          config:
            provider: openai
            model: gpt-3.5-turbo

      pipelines:
        hybrid_pipeline:
          name: test_pipeline
          stages:
            - type: stream
              config:
                operators:
                  - type: map
                    params:
                      agent: pipeline_agent
            - type: graph
              config:
                nodes:
                  process:
                    agent: pipeline_agent
          metadata:
            description: "Test hybrid pipeline"
      """
    When I load the configuration
    Then hybrid pipelines should be created
    And the pipeline should be registered with the bridge

  @skip
  Scenario: Single-shot processing with timeout
    Given I have a loaded application for single-shot testing
    When I run single-shot processing with a slow operation
    Then the operation should timeout after 3 seconds
    And a timeout error should be raised

  Scenario: Single-shot processing with None message handling
    Given I have a loaded application that returns None messages
    When I run single-shot processing with prompt "test"
    Then the result should handle None messages gracefully
    And return an empty string result

  Scenario: Single-shot processing error handling
    Given I have a loaded application for error testing
    When I run single-shot processing that causes an error
    Then the error should be wrapped in CleverAgentsException
    And the original error should be preserved

  Scenario: Interactive session with help command
    Given I have a loaded application for interactive testing
    When I start an interactive session and request help
    Then help information should be displayed
    And available commands should be shown

  Scenario: Interactive session with stream commands
    Given I have a loaded application with named streams
    When I use stream commands in interactive session
    Then messages should be sent to specific streams
    And stream command errors should be handled

  Scenario: Interactive session with graph commands
    Given I have a loaded application with graph routes
    When I use graph commands in interactive session
    Then graphs should be executed with messages
    And graph results should be displayed

  Scenario: Interactive session with unknown streams and graphs
    Given I have a loaded application for command testing
    When I use commands with unknown streams or graphs
    Then appropriate error messages should be shown
    And available options should be listed

  Scenario: Interactive session error handling
    Given I have a loaded application for interactive error testing
    When errors occur during interactive session
    Then errors should be caught and displayed
    And the session should continue running

  Scenario: Network visualization with different formats
    Given I have a complex application configuration
    When I request network visualization in mermaid format
    Then a mermaid network diagram should be generated
    And agents, streams, and graphs should be shown

  Scenario: Network visualization with unsupported format
    Given I have a loaded application for visualization
    When I request network visualization in unsupported format
    Then an unsupported format message should be returned

  Scenario: Application disposal and cleanup
    Given I have a running application with active streams
    When I perform application cleanup
    Then all streams should be disposed
    And cleanup should be logged
    And resources should be freed

  Scenario: Configuration without agents or routes
    Given I have an empty configuration
    When I try to load the configuration
    Then the application should handle empty configuration
    And no errors should occur for missing sections

  Scenario: Agent factory error handling
    Given I have a configuration that causes agent factory errors
    When I try to load the configuration
    Then agent creation errors should be properly handled
    And meaningful error messages should be provided

  Scenario: Agent type registration with built-in types
    Given I have a configuration with various agent types
    When I load the configuration
    Then built-in agent types should be registered
    And LLM and tool agents should be available

  Scenario: Template renderer with different template types
    Given I have a configuration with mixed template content types
    When the templates are processed
    Then string templates should be handled correctly
    And dict templates should extract content properly
    And invalid templates should be skipped

  Scenario: Configuration to dict conversion
    Given I have a reactive configuration
    When the configuration is converted to dictionary format
    Then agents should be properly mapped
    And global context should be included
    And prompts should be preserved

  Scenario: Error stream subscription in interactive mode
    Given I have an application with error handling
    When I start interactive session with error streams
    Then error observers should be set up
    And errors should be displayed to user

  Scenario: Tool command processing with valid JSON
    Given I have an application with tool agents configured
    And I have content with tool execution commands:
      """
      Here is the result: [TOOL_EXECUTE:echo]{"text": "hello world"}[/TOOL_EXECUTE]
      """
    When I process tool commands in the content
    Then tool commands should be executed successfully
    And the result should contain "hello world"

  # Note: Tool command processing tests removed due to environment compatibility issues
  # These tests require async execution context that conflicts with the test runner

  Scenario: Dispose application with agent cleanup
    Given an application with agents having cleanup
    When disposing the application
    Then all agent cleanup methods are called

  Scenario: Dispose application with cleanup failure
    Given an application with failing agent cleanup
    When disposing the application
    Then cleanup errors are logged as warnings
    And disposal completes successfully

  @skip
  Scenario: Deferred initialization in run_single_shot
    Given an app with deferred initialization
    When calling run_single_shot
    Then routes are initialized before execution

  @skip
  Scenario: Deferred initialization in interactive session
    Given an app with deferred initialization
    When starting interactive session
    Then routes are initialized before session