Files
makelore/tests/unit/task-visualization.test.ts
inman 80e8386fa6
Some checks failed
Electron E2E / Electron E2E (macos-latest) (push) Has been cancelled
Electron E2E / Electron E2E (ubuntu-latest) (push) Has been cancelled
Electron E2E / Electron E2E (windows-latest) (push) Has been cancelled
feat: update Makelore modules and conversations
2026-07-31 10:08:41 +08:00

673 lines
20 KiB
TypeScript

import { describe, expect, it } from 'vitest';
import { deriveTaskSteps, parseSubagentCompletionInfo } from '@/pages/Chat/task-visualization';
import {
formatVisibleChatTranscript,
getMessageVisibilityKey,
getToolVisibilityId,
} from '@/pages/Chat/chat-transcript';
import { stripProcessMessagePrefix } from '@/pages/Chat/message-utils';
import { hydrateOpenCodeSession } from '@/lib/opencode-session-state';
import type { RawMessage, ToolStatus } from '@/types/chat';
describe('deriveTaskSteps', () => {
it('uses canonical Part order to keep pre-subtask text in the execution trace', () => {
const transcript = hydrateOpenCodeSession('ses_native', [{
info: { id: 'msg_native', role: 'assistant', sessionID: 'ses_native' },
parts: [
{ id: 'part_narration', type: 'text', text: 'Inspecting the project.' },
{ id: 'part_subtask', type: 'subtask', sessionID: 'ses_child' },
{ id: 'part_answer', type: 'text', text: 'The project is ready.' },
],
}]);
const steps = deriveTaskSteps({
messages: [{
id: 'msg_native',
role: 'assistant',
content: [
{ type: 'text', id: 'part_narration', text: 'Inspecting the project.' },
{ type: 'text', id: 'part_answer', text: 'The project is ready.' },
],
}],
nativeTranscript: transcript,
streamingMessage: null,
streamingTools: [],
});
expect(steps).toEqual(expect.arrayContaining([
expect.objectContaining({ kind: 'message', detail: 'Inspecting the project.' }),
expect.objectContaining({ kind: 'system', id: 'native-part_subtask' }),
]));
expect(steps.some((step) => step.detail === 'The project is ready.')).toBe(false);
});
it('builds running steps from streaming thinking and tool status', () => {
const streamingTools: ToolStatus[] = [
{
name: 'web_search',
status: 'running',
updatedAt: Date.now(),
summary: 'Searching docs',
},
];
const steps = deriveTaskSteps({
messages: [],
streamingMessage: {
role: 'assistant',
content: [
{ type: 'thinking', thinking: 'Compare a few approaches before coding.' },
{ type: 'tool_use', id: 'tool-1', name: 'web_search', input: { query: 'opencode task list' } },
],
},
streamingTools,
});
expect(steps).toEqual([
expect.objectContaining({
id: 'stream-thinking-0',
label: 'Thinking',
status: 'running',
kind: 'thinking',
}),
expect.objectContaining({
label: 'web_search',
status: 'running',
kind: 'tool',
visibilityId: getToolVisibilityId('assistant-streaming', 'tool-1', 0),
}),
]);
});
it('keeps completed tool steps visible while a later tool is still streaming', () => {
const steps = deriveTaskSteps({
messages: [
{
role: 'assistant',
id: 'assistant-history',
content: [
{ type: 'tool_use', id: 'tool-read', name: 'read', input: { filePath: '/tmp/a.md' } },
],
},
],
streamingMessage: {
role: 'assistant',
content: [
{ type: 'tool_use', id: 'tool-grep', name: 'grep', input: { pattern: 'TODO' } },
],
},
streamingTools: [
{
toolCallId: 'tool-grep',
name: 'grep',
status: 'running',
updatedAt: Date.now(),
summary: 'Scanning files',
},
],
});
expect(steps).toEqual([
expect.objectContaining({
id: 'tool-read',
label: 'read',
status: 'completed',
kind: 'tool',
}),
expect.objectContaining({
id: 'tool-grep',
label: 'grep',
status: 'running',
kind: 'tool',
}),
]);
});
it('upgrades a completed historical tool step when streaming status reports a later state', () => {
const steps = deriveTaskSteps({
messages: [
{
role: 'assistant',
id: 'assistant-history',
content: [
{ type: 'tool_use', id: 'tool-read', name: 'read', input: { filePath: '/tmp/a.md' } },
],
},
],
streamingMessage: null,
streamingTools: [
{
toolCallId: 'tool-read',
name: 'read',
status: 'error',
updatedAt: Date.now(),
summary: 'Permission denied',
},
],
});
expect(steps).toEqual([
expect.objectContaining({
id: 'tool-read',
label: 'read',
status: 'error',
kind: 'tool',
detail: expect.stringContaining('"filePath": "/tmp/a.md"'),
}),
]);
});
it('keeps all steps when the execution graph exceeds the previous max length', () => {
const messages: RawMessage[] = Array.from({ length: 9 }, (_, index) => ({
role: 'assistant',
id: `assistant-${index}`,
content: [
{ type: 'tool_use', id: `tool-${index}`, name: `read_${index}`, input: { filePath: `/tmp/${index}.md` } },
],
}));
const steps = deriveTaskSteps({
messages,
streamingMessage: {
role: 'assistant',
content: [
{ type: 'tool_use', id: 'tool-live', name: 'grep_live', input: { pattern: 'TODO' } },
],
},
streamingTools: [
{
toolCallId: 'tool-live',
name: 'grep_live',
status: 'running',
updatedAt: Date.now(),
summary: 'Scanning current workspace',
},
],
});
expect(steps).toHaveLength(10);
expect(steps[0]).toEqual(expect.objectContaining({
id: 'tool-0',
label: 'read_0',
status: 'completed',
}));
expect(steps.at(-1)).toEqual(expect.objectContaining({
id: 'tool-live',
label: 'grep_live',
status: 'running',
}));
});
it('keeps recent completed steps from assistant history', () => {
const messages: RawMessage[] = [
{
role: 'assistant',
id: 'assistant-1',
content: [
{ type: 'thinking', thinking: 'Reviewing the code path.' },
{ type: 'tool_use', id: 'tool-2', name: 'read_file', input: { path: 'src/App.tsx' } },
],
},
];
const steps = deriveTaskSteps({
messages,
streamingMessage: null,
streamingTools: [],
});
expect(steps).toEqual([
expect.objectContaining({
id: 'history-thinking-assistant-1-0',
label: 'Thinking',
status: 'completed',
kind: 'thinking',
}),
expect.objectContaining({
id: 'tool-2',
visibilityId: getToolVisibilityId('assistant-1', 'tool-2', 0),
label: 'read_file',
status: 'completed',
kind: 'tool',
}),
]);
});
it('removes historical and streaming thinking at the ExecutionGraph derivation boundary', () => {
const steps = deriveTaskSteps({
messages: [{
id: 'history',
role: 'assistant',
content: [
{ type: 'thinking', thinking: 'historical thought' },
{ type: 'text', text: 'historical narration' },
{ type: 'tool_use', id: 'read-1', name: 'read', input: { file: 'README.md' } },
],
}],
streamingMessage: {
role: 'assistant',
content: [
{ type: 'thinking', thinking: 'stream thought' },
{ type: 'text', text: 'stream narration' },
],
},
streamingTools: [],
includeThinking: false,
});
expect(steps.some((step) => step.kind === 'thinking')).toBe(false);
expect(steps.some((step) => step.kind === 'message')).toBe(true);
expect(steps.some((step) => step.kind === 'tool' && step.id === 'read-1')).toBe(true);
});
it('uses the transcript-wide fallback index for historical tools without IDs', () => {
const steps = deriveTaskSteps({
messages: [{
role: 'assistant',
content: [{
type: 'tool_use',
id: '',
name: 'read',
input: { file: 'README.md' },
}],
}],
messageIndexOffset: 4,
streamingMessage: null,
streamingTools: [],
});
expect(steps).toEqual([
expect.objectContaining({
visibilityId: getToolVisibilityId('assistant-4', undefined, 0),
}),
]);
});
it('shares one fallback visibility ID between an ID-less streaming graph and transcript export', () => {
const streamingMessage: RawMessage = {
role: 'assistant',
content: [{
type: 'tool_use',
id: '',
name: 'read',
input: { visible: 'stream detail' },
}],
};
const streamingMessageKey = getMessageVisibilityKey(streamingMessage, 0);
const steps = deriveTaskSteps({
messages: [],
streamingMessage,
streamingMessageKey,
streamingTools: [],
});
const visibilityId = steps.find((step) => step.kind === 'tool')?.visibilityId;
expect(visibilityId).toBe(getToolVisibilityId(streamingMessageKey, undefined, 0));
expect(formatVisibleChatTranscript([streamingMessage], {
showThinking: false,
showTimestamps: false,
expandedToolIds: new Set(visibilityId ? [visibilityId] : []),
})).toContain('stream detail');
});
it('uses the matched streaming tool input as the shared visible graph and export detail', () => {
const streamingMessage: RawMessage = {
id: 'assistant-stream',
role: 'assistant',
content: [{
type: 'tool_use',
id: 'tool-read',
name: 'read',
input: { marker: 'RAW-UNSEEN-TOKEN' },
}],
};
const messageKey = getMessageVisibilityKey(streamingMessage, 0);
const visibilityId = getToolVisibilityId(messageKey, 'tool-read', 0);
const steps = deriveTaskSteps({
messages: [],
streamingMessage,
streamingMessageKey: messageKey,
streamingTools: [{
toolCallId: 'tool-read',
name: 'read',
status: 'running',
updatedAt: 1,
summary: 'Visible status title',
}],
});
const graphDetail = steps.find((step) => step.visibilityId === visibilityId)?.detail;
const markdown = formatVisibleChatTranscript([streamingMessage], {
showThinking: false,
showTimestamps: false,
expandedToolIds: new Set([visibilityId]),
});
expect(graphDetail).toContain('RAW-UNSEEN-TOKEN');
expect(graphDetail).not.toContain('Visible status title');
expect(markdown).toContain('RAW-UNSEEN-TOKEN');
});
it('matches reordered same-name streaming tools by call ID before position', () => {
const streamingMessage: RawMessage = {
id: 'message-same-name',
role: 'assistant',
content: [
{ type: 'tool_use', id: 'call-a', name: 'read', input: { marker: 'A-INPUT' } },
{ type: 'tool_use', id: 'call-b', name: 'read', input: { marker: 'B-INPUT' } },
],
};
const steps = deriveTaskSteps({
messages: [],
streamingMessage,
streamingMessageKey: 'message-same-name',
streamingTools: [
{
toolCallId: 'call-b',
name: 'read',
status: 'running',
updatedAt: 2,
},
{
toolCallId: 'call-a',
name: 'read',
status: 'running',
updatedAt: 1,
},
],
}).filter((step) => step.kind === 'tool');
expect(steps).toHaveLength(2);
expect(steps.find((step) =>
step.visibilityId === getToolVisibilityId('message-same-name', 'call-a', 0))
?.detail).toContain('A-INPUT');
expect(steps.find((step) =>
step.visibilityId === getToolVisibilityId('message-same-name', 'call-b', 1))
?.detail).toContain('B-INPUT');
});
it('uses the matched message call ID when status has only a different event ID', () => {
const steps = deriveTaskSteps({
messages: [],
streamingMessage: {
id: 'message-event-id',
role: 'assistant',
content: [{
type: 'tool_use',
id: 'call-read',
name: 'read',
input: { marker: 'READ-INPUT' },
}],
},
streamingMessageKey: 'message-event-id',
streamingTools: [{
id: 'part-read',
name: 'read',
status: 'running',
updatedAt: 1,
}],
}).filter((step) => step.kind === 'tool');
expect(steps).toHaveLength(1);
expect(steps[0]).toEqual(expect.objectContaining({
id: 'call-read',
visibilityId: getToolVisibilityId('message-event-id', 'call-read', 0),
detail: expect.stringContaining('READ-INPUT'),
}));
});
it('preserves meaningful whitespace in the shared tool input detail', () => {
const streamingMessage: RawMessage = {
id: 'message-whitespace',
role: 'assistant',
content: [{
type: 'tool_use',
id: 'call-write',
name: 'write',
input: {
code: 'const answer = 42;',
path: 'Private Notes/file.txt',
},
}],
};
const visibilityId = getToolVisibilityId(
'message-whitespace',
'call-write',
0,
);
const detail = deriveTaskSteps({
messages: [],
streamingMessage,
streamingMessageKey: 'message-whitespace',
streamingTools: [{
toolCallId: 'call-write',
name: 'write',
status: 'running',
updatedAt: 1,
}],
}).find((step) => step.visibilityId === visibilityId)?.detail;
const markdown = formatVisibleChatTranscript([streamingMessage], {
showThinking: false,
showTimestamps: false,
expandedToolIds: new Set([visibilityId]),
});
expect(detail).toContain('const answer = 42;');
expect(detail).toContain('Private Notes/file.txt');
expect(markdown).toContain('const answer = 42;');
expect(markdown).toContain('Private Notes/file.txt');
});
it('preserves leading and trailing whitespace in string tool input details', () => {
const detail = deriveTaskSteps({
messages: [],
streamingMessage: {
id: 'message-string-whitespace',
role: 'assistant',
content: [{
type: 'tool_use',
id: 'call-write',
name: 'write',
input: ' indented code ',
}],
},
streamingMessageKey: 'message-string-whitespace',
streamingTools: [],
}).find((step) => step.kind === 'tool')?.detail;
expect(detail).toBe(' indented code ');
});
it('uses the matched message-tool index when ID-less streaming statuses arrive out of order', () => {
const streamingMessage: RawMessage = {
role: 'assistant',
content: [
{ type: 'tool_use', id: '', name: 'read', input: { file: 'read.md' } },
{ type: 'tool_use', id: '', name: 'write', input: { file: 'write.md' } },
],
};
const streamingMessageKey = getMessageVisibilityKey(streamingMessage, 0);
const steps = deriveTaskSteps({
messages: [],
streamingMessage,
streamingMessageKey,
streamingTools: [
{ name: 'write', status: 'running', updatedAt: 2 },
{ name: 'read', status: 'running', updatedAt: 1 },
],
});
expect(steps.find((step) => step.label === 'write')?.visibilityId)
.toBe(getToolVisibilityId(streamingMessageKey, undefined, 1));
expect(steps.find((step) => step.label === 'read')?.visibilityId)
.toBe(getToolVisibilityId(streamingMessageKey, undefined, 0));
});
it('splits cumulative streaming thinking into separate execution steps', () => {
const steps = deriveTaskSteps({
messages: [],
streamingMessage: {
role: 'assistant',
content: [
{ type: 'thinking', thinking: 'Reviewing X.' },
{ type: 'thinking', thinking: 'Reviewing X. Comparing Y.' },
{ type: 'thinking', thinking: 'Reviewing X. Comparing Y. Drafting answer.' },
],
},
streamingTools: [],
});
expect(steps).toEqual([
expect.objectContaining({
id: 'stream-thinking-0',
detail: 'Reviewing X.',
status: 'completed',
}),
expect.objectContaining({
id: 'stream-thinking-1',
detail: 'Comparing Y.',
status: 'completed',
}),
expect.objectContaining({
id: 'stream-thinking-2',
detail: 'Drafting answer.',
status: 'running',
}),
]);
});
it('keeps earlier reply segments in the graph when the last streaming segment is rendered separately', () => {
const steps = deriveTaskSteps({
messages: [],
streamingMessage: {
role: 'assistant',
content: [
{ type: 'text', text: 'Checked X.' },
{ type: 'text', text: 'Checked X. Checked Snowball.' },
{ type: 'text', text: 'Checked X. Checked Snowball. Here is the summary.' },
],
},
streamingTools: [],
omitLastStreamingMessageSegment: true,
});
expect(steps).toEqual([
expect.objectContaining({
id: 'stream-message-0',
detail: 'Checked X.',
status: 'completed',
}),
expect.objectContaining({
id: 'stream-message-1',
detail: 'Checked Snowball.',
status: 'completed',
}),
]);
});
it('folds earlier reply segments into the graph but leaves the final answer for the chat bubble', () => {
const steps = deriveTaskSteps({
messages: [
{
role: 'assistant',
id: 'assistant-reply',
content: [
{ type: 'text', text: 'Checked X.' },
{ type: 'text', text: 'Checked X. Checked Snowball.' },
{ type: 'text', text: 'Checked X. Checked Snowball. Here is the summary.' },
],
},
],
streamingMessage: null,
streamingTools: [],
});
expect(steps).toEqual([
expect.objectContaining({
id: 'history-message-assistant-reply-0',
detail: 'Checked X.',
status: 'completed',
}),
expect.objectContaining({
id: 'history-message-assistant-reply-1',
detail: 'Checked Snowball.',
status: 'completed',
}),
]);
});
it('strips folded process narration from the final reply text', () => {
expect(stripProcessMessagePrefix(
'Checked X. Checked Snowball. Here is the summary.',
['Checked X.', 'Checked Snowball.'],
)).toBe('Here is the summary.');
});
it('builds a branch for spawned subagents', () => {
const messages: RawMessage[] = [
{
role: 'assistant',
id: 'assistant-2',
content: [
{
type: 'tool_use',
id: 'spawn-1',
name: 'sessions_spawn',
input: { agentId: 'coder', task: 'inspect repo' },
},
{
type: 'tool_use',
id: 'yield-1',
name: 'sessions_yield',
input: { message: 'wait coder finishes' },
},
],
},
];
const steps = deriveTaskSteps({
messages,
streamingMessage: null,
streamingTools: [],
});
expect(steps).toEqual([
expect.objectContaining({
id: 'spawn-1',
label: 'sessions_spawn',
depth: 1,
}),
expect.objectContaining({
id: 'spawn-1:branch',
label: 'coder run',
depth: 2,
parentId: 'spawn-1',
}),
expect.objectContaining({
id: 'yield-1',
label: 'sessions_yield',
depth: 3,
parentId: 'spawn-1:branch',
}),
]);
});
it('parses internal subagent completion events from injected user messages', () => {
const info = parseSubagentCompletionInfo({
role: 'user',
content: [{
type: 'text',
text: `[Internal task completion event]
source: subagent
session_key: agent:coder:subagent:child-123
session_id: child-session-id
status: completed successfully`,
}],
} as RawMessage);
expect(info).toEqual({
sessionKey: 'agent:coder:subagent:child-123',
sessionId: 'child-session-id',
agentId: 'coder',
});
});
});