Normalize agent attachment output (#122)

This commit is contained in:
Eyejoker
2026-05-02 19:50:32 +09:00
committed by GitHub
parent 9e45534de0
commit d3c02265e5
16 changed files with 414 additions and 124 deletions

View File

@@ -1,15 +1,18 @@
export {
extractMarkdownImageAttachments,
extractImageTagPaths,
IMAGE_TAG_RE,
IPC_CLOSE_SENTINEL,
IPC_INPUT_SUBDIR,
IPC_POLL_MS,
normalizeAgentOutput,
normalizeEjclawStructuredOutput,
normalizePublicTextOutput,
OUTPUT_END_MARKER,
OUTPUT_START_MARKER,
writeProtocolOutput,
type NormalizedRunnerOutput,
type NormalizedAgentOutput,
type RunnerOutputAttachment,
type RunnerOutputPhase,
type RunnerOutputVerdict,

View File

@@ -1,64 +1,32 @@
import path from 'path';
import {
extractImageTagPaths,
normalizeEjclawStructuredOutput,
} from '../agent-protocol.js';
import { normalizeAgentOutput } from '../agent-protocol.js';
import type { OutboundAttachment } from '../types.js';
const IMAGE_EXTS = /\.(png|jpe?g|gif|webp|bmp)$/i;
const MD_LINK_RE = /\[[^\]]*\]\((\/[^)]+)\)/g;
const LOCAL_MARKDOWN_LINK_RE = /\[[^\]\n]*\]\((\/[^)\n]+)\)/g;
export interface PreparedDiscordOutbound {
text: string;
cleanText: string;
attachments: OutboundAttachment[];
attachmentSource: 'structured' | 'md-link' | 'image-tag' | 'none';
attachmentSource: 'structured' | 'md-link' | 'image-tag' | 'mixed' | 'none';
silent: boolean;
}
function extractMarkdownImageAttachments(text: string): {
cleanText: string;
attachments: OutboundAttachment[];
} {
const attachments: OutboundAttachment[] = [];
const seen = new Set<string>();
const cleanText = text.replace(MD_LINK_RE, (_full, rawPath: string) => {
function sanitizeLocalMarkdownLinks(text: string): string {
return text.replace(LOCAL_MARKDOWN_LINK_RE, (_full, rawPath: string) => {
const trimmed = rawPath.trim();
if (IMAGE_EXTS.test(trimmed)) {
if (!seen.has(trimmed)) {
attachments.push({
path: trimmed,
name: path.basename(trimmed),
});
seen.add(trimmed);
}
return '';
}
const basename = path.basename(trimmed.replace(/#.*$/, ''));
const lineMatch = trimmed.match(/#L(\d+)/);
return lineMatch ? `\`${basename}:${lineMatch[1]}\`` : `\`${basename}\``;
});
return { cleanText, attachments };
}
function imageTagPathsToAttachments(paths: string[]): OutboundAttachment[] {
return paths
.filter((filePath) => IMAGE_EXTS.test(filePath))
.map((filePath) => ({
path: filePath,
name: path.basename(filePath),
}));
}
export function prepareDiscordOutbound(
text: string,
optionAttachments: OutboundAttachment[] | undefined,
): PreparedDiscordOutbound {
const normalized = normalizeEjclawStructuredOutput(text);
const normalized = normalizeAgentOutput(text);
if (normalized.output?.visibility === 'silent') {
return {
text: '',
@@ -72,30 +40,26 @@ export function prepareDiscordOutbound(
const structuredOutput =
normalized.output?.visibility === 'public' ? normalized.output : null;
const outboundText = structuredOutput?.text ?? normalized.result ?? text;
const structuredAttachments =
const cleanText = sanitizeLocalMarkdownLinks(outboundText);
const attachments =
optionAttachments && optionAttachments.length > 0
? optionAttachments
: (structuredOutput?.attachments ?? []);
const hasStructuredAttachments = structuredAttachments.length > 0;
const markdownExtracted = extractMarkdownImageAttachments(outboundText);
const imageTagExtracted = extractImageTagPaths(markdownExtracted.cleanText);
const legacyImageTagAttachments = imageTagPathsToAttachments(
imageTagExtracted.imagePaths,
);
const hasAttachments = attachments.length > 0;
const attachmentSource =
optionAttachments && optionAttachments.length > 0
? 'structured'
: normalized.attachmentSource === 'legacy-ejclaw-json'
? 'structured'
: normalized.attachmentSource === 'markdown-image'
? 'md-link'
: (normalized.attachmentSource ?? 'none');
return {
text: outboundText,
cleanText: imageTagExtracted.cleanText,
attachments: hasStructuredAttachments
? structuredAttachments
: [...markdownExtracted.attachments, ...legacyImageTagAttachments],
attachmentSource: hasStructuredAttachments
? 'structured'
: markdownExtracted.attachments.length > 0
? 'md-link'
: legacyImageTagAttachments.length > 0
? 'image-tag'
: 'none',
cleanText,
attachments,
attachmentSource: hasAttachments ? attachmentSource : 'none',
silent: false,
};
}

View File

@@ -225,4 +225,57 @@ ${JSON.stringify({
'"ejclaw"',
);
});
it('normalizes markdown images and sends them as files', async () => {
const channel = new DiscordChannel('test-token', createTestOpts());
await channel.connect();
const filePath = path.join(
os.tmpdir(),
`ejclaw-discord-markdown-image-${Date.now()}.png`,
);
fs.writeFileSync(filePath, ONE_PIXEL_PNG);
tempFiles.push(filePath);
const mockChannel = {
send: vi.fn().mockResolvedValue({ id: 'discord-message-3' }),
sendTyping: vi.fn(),
};
clientRef.current.channels.fetch.mockResolvedValue(mockChannel);
await channel.sendMessage(
'dc:1234567890123456',
`스크린샷입니다.\n![screenshot](${filePath})`,
);
expect(mockChannel.send).toHaveBeenCalledWith({
content: '스크린샷입니다.',
files: [
{
attachment: fs.realpathSync(filePath),
name: path.basename(filePath),
},
],
flags: 1 << 2,
});
});
it('keeps non-image local markdown links readable without uploading files', async () => {
const channel = new DiscordChannel('test-token', createTestOpts());
await channel.connect();
const mockChannel = {
send: vi.fn().mockResolvedValue({ id: 'discord-message-4' }),
sendTyping: vi.fn(),
};
clientRef.current.channels.fetch.mockResolvedValue(mockChannel);
await channel.sendMessage(
'dc:1234567890123456',
'수정 위치: [source](/tmp/project/src/index.ts#L42)',
);
expect(mockChannel.send).toHaveBeenCalledWith({
content: '수정 위치: `index.ts:42`',
files: undefined,
flags: 1 << 2,
});
});
});

View File

@@ -2,7 +2,7 @@ import type {
IpcMessageForwardResult,
IpcMessagePayload,
} from './ipc-types.js';
import { normalizeEjclawStructuredOutput } from './agent-protocol.js';
import { normalizeAgentOutput } from './agent-protocol.js';
import type { RegisteredGroup } from './types.js';
export async function forwardAuthorizedIpcMessage(
@@ -81,7 +81,7 @@ export async function forwardAuthorizedIpcMessage(
};
}
const normalized = normalizeEjclawStructuredOutput(msg.text);
const normalized = normalizeAgentOutput(msg.text);
if (normalized.output?.visibility === 'silent') {
return {
outcome: 'sent',

View File

@@ -152,4 +152,39 @@ describe('web dashboard attachment data', () => {
],
});
});
it('turns markdown image output into dashboard attachments', () => {
const message: NewMessage = {
id: 'msg-markdown-image',
chat_jid: 'dc:ops',
sender: 'bot-1',
sender_name: 'owner',
content:
'라벨 좌측 클리핑 회귀 수정했습니다.\n![screenshot](/tmp/bar-chart-label-fit-playwright.png)',
timestamp: '2026-04-26T05:31:00.000Z',
is_from_me: true,
is_bot_message: true,
message_source_kind: 'bot',
};
const activity = buildWebDashboardRoomActivity({
serviceId: 'codex-main',
entry: roomEntry,
pairedTask: null,
turns: [],
attempts: [],
outputs: [],
messages: [message],
});
expect(activity.messages[0]).toMatchObject({
content: '라벨 좌측 클리핑 회귀 수정했습니다.',
attachments: [
{
path: '/tmp/bar-chart-label-fit-playwright.png',
name: 'bar-chart-label-fit-playwright.png',
},
],
});
});
});

View File

@@ -1,6 +1,6 @@
import {
extractImageTagPaths,
normalizeEjclawStructuredOutput,
normalizeAgentOutput,
} from './agent-protocol.js';
import { isWatchCiTask } from './task-watch-status.js';
import type {
@@ -228,17 +228,18 @@ function normalizeStructuredVisibleContent(value: string): {
text: string | null;
attachments: OutboundAttachment[];
} {
if (!hasStructuredOutputHint(value)) return splitLegacyImageTags(value);
const candidates = [value, decodeCommonHtmlEntities(value)];
const candidates = hasStructuredOutputHint(value)
? [value, decodeCommonHtmlEntities(value)]
: [value];
for (const candidate of candidates) {
const normalized = normalizeEjclawStructuredOutput(candidate);
const normalized = normalizeAgentOutput(candidate);
if (normalized.output?.visibility === 'silent') {
return { text: null, attachments: [] };
}
if (
normalized.output?.visibility === 'public' &&
normalized.output.text !== candidate
(normalized.output.text !== candidate ||
(normalized.output.attachments?.length ?? 0) > 0)
) {
return splitLegacyImageTags(
normalized.output.text,