Verifies that langchain_messages_to_copilotkit always emits the
assistant message even when content is empty (OpenAI-style tool-call-only
responses), ensuring no orphaned parentMessageId references.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The original test only tested the _merge_emit_state helper in isolation
using mocks. Replace with integration tests that simulate the actual
state-tracking loop from _stream_events, including:
- Sequential emits with different keys preserve all keys (core bug)
- Proof that the bug manifests without current_graph_state.update
- Three sequential emits accumulate correctly
- Same key emitted twice uses latest value
- Non-dict emit does not corrupt current_graph_state
- Initial state reference is not mutated
The original fix only covered _emit_state_sync_event but missed the
get_state() code path, which also returns state containing Pydantic
BaseModel instances to callers that will JSON-serialize downstream.
Also fixes filter_state_on_schema_keys returning None implicitly when
schema keys are not set (the except branch returned state but the
non-matching if branch did not).
Adds 17 tests covering _serialize_state, _emit_state_sync_event, and
get_state with Pydantic models (nested, lists, plain dicts, empty).
Verifies langchain_messages_to_copilotkit correctly concatenates all text
parts from list-style content blocks (Anthropic models). Includes
regression tests for the exact scenario from #1748 (text + image blocks)
and tests for string lists, mixed content, empty lists, and edge cases.
LangGraph 1.x can return string or dict resume values from interrupt(),
not just lists. The code now type-checks the response: strings are
returned directly, dicts are JSON-serialized, and lists use the
existing [-1].content path.
Sequential emit_state calls with different keys now preserve all keys
in the snapshot. Previously, each call would replace the entire
manually_emitted_state, losing keys from earlier calls.
- Add TestAGUIStyleEventIntegration to test_agui_agent.py: proves that AG-UI-style
unprefixed events (manually_emit_message, manually_emit_tool_call, exit) flow
correctly through CopilotKit's _handle_single_event to the AG-UI base handler
without being suppressed or double-converted by the CopilotKit layer.
- Replace LangGraphAgent with LangGraphAGUIAgent in both example files
(canvas/gemini and v1/_legacy/saas-dynamic-dashboards) to close the concern
about removed APIs still being referenced externally.
- Update poetry.lock to reflect ag_ui_langgraph 0.0.32 and dependency updates.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Expands test suite for LangGraphAGUIAgent:
- TestLanggraphDefaultMergeState: verify no duplicates in copilotkit.actions
and that tool ordering is preserved in the merged result
- TestReasoningContentPreservation: verify unknown custom events pass through,
empty-string messages still emit the full TEXT_MESSAGE_* sequence, and
empty-args tool calls still emit the full TOOL_CALL_* sequence
Expands test suite for emit filtering:
- TestMissingOrNoneRawEvent: events with rawEvent=None should pass through
to super() without crashing
- TestNoneEmitMetadataValues: None and 0 values for emit metadata keys
should NOT filter (only exactly False suppresses events)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Python SDK: 18 new tests for LangGraphAGUIAgent (custom event handling,
emit filtering, state merging, copilotkit namespace)
- TypeScript SDK: 25 new tests for copilotkitCustomizeConfig and
convertActionsToDynamicStructuredTools
- TypeScript Runtime: 27 new tests for event-source helpers
(shouldEmitToolCall, getCurrentMessageId, getCurrentContent, etc.)
- TypeScript Runtime: expanded dispatch-event-filtering tests with
custom event dispatch (manually_emit_message/tool_call/state, exit)
and langGraphDefaultMergeState tests
- Dead code annotations: LangGraphAgent class and use_function_call=True
branch annotated with TODO(ran-review) for Ran to verify
https://claude.ai/code/session_01BPMn7zhadhapfyyD8kYeAH
Tests the two bugs fixed in this PR:
1. Dict raw_event metadata must be read with .get(), not getattr()
2. Filtered events must return None (not ""), and run() strips them
Adds pytest as a dev dependency.
Maxim