HAL9000
fcf1306817
Fixed all issues identified in PR review: 1. Fixed self.history property/attribute conflict in ConversationStateManager: - Renamed internal snapshot list from self.history to self._snapshots - The @property history now correctly returns conversation messages - Added public snapshots property for time-travel snapshot access - Updated existing tests to use manager.snapshots instead of manager.history 2. Removed unused variable full_history and dead code self._node_executors noqa comment from graph.py execute() method 3. Fixed ambiguous NON-BREAKING HYPHEN in state.py docstring 4. Restored removed public methods to ConversationStateManager: - clear_history(), load_checkpoint(), get_latest_checkpoint() - time_travel(), replace_state() - Added state property for backward compatibility 5. Fixed reset() to properly restore metadata from initial_state 6. Added BDD feature file and step definitions for conversation state management covering all 6 changed files ISSUES CLOSED: #1
83 lines
4.1 KiB
Gherkin
83 lines
4.1 KiB
Gherkin
Feature: Conversation state management
|
|
The ConversationStateManager separates persistent conversation history from
|
|
ephemeral execution state. Conversation history survives across runs while
|
|
execution state (current_node, execution_count, error) resets each time.
|
|
|
|
Scenario: Conversation history persists across execution resets
|
|
Given a ConversationStateManager with no initial state
|
|
When I append messages "hello" and "world" to the conversation history
|
|
And I reset the execution state
|
|
Then the conversation history should still contain 2 messages
|
|
And the execution count should be 0 after reset
|
|
|
|
Scenario: reset_execution_state clears only execution fields
|
|
Given a ConversationStateManager with no initial state
|
|
When I update the state with current_node "agent_a" and execution_count 5
|
|
And I append a message "keep me" to the conversation history
|
|
And I reset the execution state
|
|
Then the conversation history should contain the message "keep me"
|
|
And the current_node should be None after reset
|
|
|
|
Scenario: get_full_history returns all persistent messages
|
|
Given a ConversationStateManager with no initial state
|
|
When I append messages "first" and "second" and "third" to the conversation history
|
|
Then get_full_history should return 3 messages
|
|
|
|
Scenario: history property returns persistent conversation messages
|
|
Given a ConversationStateManager with no initial state
|
|
When I append a message "persistent" to the conversation history
|
|
Then the history property should contain the message "persistent"
|
|
|
|
Scenario: update_state with messages key appends to conversation history
|
|
Given a ConversationStateManager with no initial state
|
|
When I call update_state with messages containing "new message"
|
|
Then the conversation history should contain the message "new message"
|
|
|
|
Scenario: GraphState.to_graph_state returns conversation_history and execution_state keys
|
|
Given a GraphState with messages and metadata
|
|
When I call to_graph_state on the GraphState
|
|
Then the result should contain a conversation_history key
|
|
And the result should contain an execution_state key with current_node
|
|
|
|
Scenario: ExecutionState model holds execution-level fields
|
|
Given an ExecutionState with current_node "node_x" and execution_count 3
|
|
Then the ExecutionState current_node should be "node_x"
|
|
And the ExecutionState execution_count should be 3
|
|
|
|
Scenario: StateManager backward-compat alias works as ConversationStateManager
|
|
Given a StateManager created via the backward-compat alias
|
|
When I append a message "compat" to the conversation history via StateManager
|
|
Then the StateManager history property should contain the message "compat"
|
|
|
|
Scenario: append_messages evicts oldest entries when MAX_HISTORY_SIZE is exceeded
|
|
Given a ConversationStateManager with no initial state
|
|
When I append 55 messages to the conversation history
|
|
Then the conversation history should contain at most 50 messages
|
|
And the most recent messages should be retained
|
|
|
|
Scenario: Full reset clears both conversation history and execution state
|
|
Given a ConversationStateManager with no initial state
|
|
When I append a message "before reset" to the conversation history
|
|
And I update the state with current_node "some_node" and execution_count 5
|
|
And I perform a full reset
|
|
Then the conversation history should be empty after full reset
|
|
And the execution count should be 0 after full reset
|
|
|
|
Scenario: Bridge default update mode is APPEND
|
|
Given a RxPyLangGraphBridge instance
|
|
Then the default update mode should be APPEND
|
|
|
|
Scenario: Session full_history property returns all messages
|
|
Given a Session with two messages appended
|
|
Then the full_history property should return 2 messages
|
|
|
|
Scenario: Session get_messages with no limit returns full thread
|
|
Given a Session with three messages appended
|
|
When I call get_messages with no limit
|
|
Then all 3 messages should be returned
|
|
|
|
Scenario: Session append_message preserves full history
|
|
Given a Session with one existing message
|
|
When I append a second message to the session
|
|
Then the session should have 2 messages in total
|