Files
copilotkit__copilotkit/showcase/aimock/d6/langgraph-fastapi/frontend-tools-async.json
Lukas Moschitz 1dde897af3 fix(showcase): stop langgraph-fastapi frontend-tools-async D6 loop
fixture[8] (userMessage:"project planning", turnIndex:0) returned a
query_notes toolCall. aimock matches userMessage by substring, so
"project planning" also matched the test prompt "Find my notes about
project planning"; as a turnIndex:0 'scripted turn' it beat the correct
terminal fixture[2] (toolCallId-keyed, content-only) on the follow-up turn,
so the agent kept getting another tool call and looped ~30x, never emitting
the done signal (D6 done-signal-missing).

Remove turnIndex:0 from fixture[8] so it's a plain fallback that loses to
the terminal fixture by registration order (matching the langgraph-typescript
shape). D6 now green: two real ~6s runs (was a 120s timeout).

Note: langgraph-python has the identical bug (still red) — same one-line fix
applies there; tracked separately.
2026-07-24 15:04:04 +02:00

128 lines
5.6 KiB
JSON

{
"_meta": {
"description": "D6 fixtures for langgraph-fastapi / frontend-tools-async",
"sourceFile": "d5-all.json",
"created": "2026-05-21"
},
"fixtures": [
{
"match": {
"userMessage": "auth check turn 1",
"turnIndex": 0,
"context": "langgraph-fastapi"
},
"response": {
"content": "Authenticated session is active. The runtime accepted your request because the Authorization header carried the demo bearer token."
}
},
{
"match": {
"userMessage": "fetch the async metric",
"turnIndex": 0,
"context": "langgraph-fastapi"
},
"response": {
"content": "The async tool resolved with the requested metric. The frontend handler awaited completion before forwarding the result back to the agent — async-streaming behavior confirmed."
}
},
{
"_comment": "frontend-tools-async — project-planning pill, follow-up content keyed on the prior tool's id. Comes BEFORE the tool-emitting fixture so iteration 2 (after the async handler returns) hits this branch instead of re-emitting query_notes. hasToolResult gates were dropped because they broke after the user clicked another tool-using pill earlier in the same thread.",
"match": {
"userMessage": "Find my notes about project planning",
"toolCallId": "call_d5_query_notes_project_planning_001",
"context": "langgraph-fastapi"
},
"response": {
"content": "You have notes on project planning: \"Q2 project planning kickoff\" and \"Project planning retrospective notes\". Let me know if you want a summary of either."
}
},
{
"_comment": "frontend-tools-async — project-planning pill. 1st turn: query_notes tool call. Specific match wins over the broad 'plan' fixture in feature-parity.json (d5-all.json loads first). turnIndex gate dropped — the multi-pill sequential e2e (frontend-tools-async.spec.ts:149) clicks earlier pills first, leaving prior tool/assistant turns in the thread, so this user message is no longer at turnIndex 0. The toolCallId-keyed follow-up fixture above still wins for iteration 2 (last message is tool with that id); this fixture only fires when last-message.role=user.",
"match": {
"userMessage": "Find my notes about project planning",
"context": "langgraph-fastapi"
},
"response": {
"toolCalls": [
{
"id": "call_d5_query_notes_project_planning_001",
"name": "query_notes",
"arguments": "{\"keyword\":\"project planning\"}"
}
]
}
},
{
"_comment": "frontend-tools-async — auth pill, follow-up content keyed on the prior tool's id. Must come BEFORE the tool-emitting fixture.",
"match": {
"userMessage": "Search my notes for anything related to auth",
"toolCallId": "call_d5_query_notes_auth_001",
"context": "langgraph-fastapi"
},
"response": {
"content": "You have one note related to auth: \"Planning: migrate auth to passkeys\" — it covers WebAuthn library options and a fallback for unsupported browsers."
}
},
{
"_comment": "frontend-tools-async — auth pill. 1st turn: query_notes tool call. Beats the showcase-assistant catch-all in feature-parity.json by virtue of d5-all.json's load order and a longer specific substring match. turnIndex gate dropped — multi-pill sequential e2e leaves prior turns in the thread before this pill is clicked, so the prompt is no longer at turnIndex 0. The toolCallId-keyed follow-up above still wins for iteration 2.",
"match": {
"userMessage": "Search my notes for anything related to auth",
"context": "langgraph-fastapi"
},
"response": {
"toolCalls": [
{
"id": "call_d5_query_notes_auth_001",
"name": "query_notes",
"arguments": "{\"keyword\":\"auth\"}"
}
]
}
},
{
"_comment": "frontend-tools-async — reading pill, follow-up content keyed on the prior tool's id. Must come BEFORE the tool-emitting fixture. Locked narration leading phrase per spec test #4 assertion.",
"match": {
"userMessage": "Do I have any notes tagged reading",
"toolCallId": "call_d5_query_notes_reading_001",
"context": "langgraph-fastapi"
},
"response": {
"content": "You have a note titled \"Book recommendations\" that is tagged with \"reading.\" It includes the following books: Thinking Fast and Slow by Daniel Kahneman, The Design of Everyday Things by Don Norman."
}
},
{
"_comment": "frontend-tools-async — reading pill. 1st turn: query_notes tool call. turnIndex gate dropped — multi-pill sequential e2e leaves prior turns in the thread before this pill is clicked, so the prompt is no longer at turnIndex 0. The toolCallId-keyed follow-up above still wins for iteration 2.",
"match": {
"userMessage": "Do I have any notes tagged reading",
"context": "langgraph-fastapi"
},
"response": {
"toolCalls": [
{
"id": "call_d5_query_notes_reading_001",
"name": "query_notes",
"arguments": "{\"keyword\":\"reading\"}"
}
]
}
},
{
"match": {
"userMessage": "project planning",
"context": "langgraph-fastapi"
},
"response": {
"content": "I searched your notes and found a few results.",
"toolCalls": [
{
"name": "query_notes",
"arguments": {
"keyword": "project planning"
}
}
]
}
}
]
}