mirror of
https://github.com/toeverything/AFFiNE.git
synced 2026-08-19 02:51:47 +08:00
fix(server): transcript key & bump models (#15485)
#### PR Dependency Tree * **PR #15485** 👈 This tree was auto-generated by [Charcoal](https://github.com/danerwilliams/charcoal) <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit * **New Features** * Upgraded built-in AI actions and chat model selection to Gemini 3.7 Flash. * Transcript attachments now retain storage keys and support secure presigned URLs during processing. * Completed transcript results preserve attachment metadata for reliable access. * **Bug Fixes** * Improved transcript handling across uploads, retries, and native processing. * Added safeguards for invalid or missing attachment URLs. * **Maintenance** * Added a migration to backfill storage keys for existing transcript records without altering their URLs. <!-- end of auto-generated comment: release notes by coderabbit.ai -->
This commit is contained in:
@@ -294,7 +294,7 @@ test('retryTask reuses failed task and queues a new action attempt', async t =>
|
||||
} as never,
|
||||
{} as never,
|
||||
{
|
||||
resolveTranscriptionModel: Sinon.stub().resolves('gemini-3.5-flash-lite'),
|
||||
resolveTranscriptionModel: Sinon.stub().resolves('gemini-3.7-flash'),
|
||||
} as never,
|
||||
{} as never,
|
||||
{} as never,
|
||||
@@ -303,7 +303,7 @@ test('retryTask reuses failed task and queues a new action attempt', async t =>
|
||||
|
||||
const result = await service.retryTask('user-1', 'workspace-1', 'task-1');
|
||||
|
||||
t.is(result.status, AiJobStatus.running);
|
||||
t.is(result?.status, AiJobStatus.running);
|
||||
t.like(queuedJobs[0] as Record<string, unknown>, {
|
||||
name: 'copilot.transcript.task.submit',
|
||||
});
|
||||
@@ -351,9 +351,7 @@ for (const status of ['ready', 'settled']) {
|
||||
} as never,
|
||||
{} as never,
|
||||
{
|
||||
resolveTranscriptionModel: Sinon.stub().resolves(
|
||||
'gemini-3.5-flash-lite'
|
||||
),
|
||||
resolveTranscriptionModel: Sinon.stub().resolves('gemini-3.7-flash'),
|
||||
} as never,
|
||||
{} as never,
|
||||
{} as never,
|
||||
@@ -403,7 +401,8 @@ test('transcriptTask runs native transcript recipe through action bridge when av
|
||||
],
|
||||
infos: [
|
||||
{
|
||||
url: 'data:image/png;base64,YXVkaW8=',
|
||||
key: 'blob-1-0',
|
||||
url: 'https://affine.fail/api/copilot/blob/user-1/workspace-1/blob-1-0',
|
||||
mimeType: 'audio/opus',
|
||||
index: 0,
|
||||
},
|
||||
@@ -428,7 +427,11 @@ test('transcriptTask runs native transcript recipe through action bridge when av
|
||||
},
|
||||
} as never,
|
||||
{} as never,
|
||||
{} as never,
|
||||
{
|
||||
presignGet: Sinon.stub().resolves(
|
||||
'https://canary.copilotcontent.affine.pro/blob-1-0?sig=test'
|
||||
),
|
||||
} as never,
|
||||
{} as never,
|
||||
createTranscriptPromptService() as never,
|
||||
createSuccessfulTranscriptBridge('run-bridge', bridgeInputs) as never
|
||||
@@ -447,6 +450,20 @@ test('transcriptTask runs native transcript recipe through action bridge when av
|
||||
slot: 'transcript.audio',
|
||||
builtInRouteId: 'Transcript audio structured',
|
||||
});
|
||||
t.deepEqual(
|
||||
(
|
||||
bridgeInputs[0] as {
|
||||
nativeInput: { input: { infos: unknown[] } };
|
||||
}
|
||||
).nativeInput.input.infos,
|
||||
[
|
||||
{
|
||||
url: 'https://canary.copilotcontent.affine.pro/blob-1-0?sig=test',
|
||||
mimeType: 'audio/opus',
|
||||
index: 0,
|
||||
},
|
||||
]
|
||||
);
|
||||
const messages = (
|
||||
bridgeInputs[0] as {
|
||||
step: {
|
||||
@@ -459,7 +476,10 @@ test('transcriptTask runs native transcript recipe through action bridge when av
|
||||
infos: [{ mimeType: 'audio/opus', index: 0 }],
|
||||
});
|
||||
t.deepEqual(messages.at(-1)?.attachments, [
|
||||
{ attachment: 'data:image/png;base64,YXVkaW8=', mimeType: 'audio/opus' },
|
||||
{
|
||||
attachment: 'https://canary.copilotcontent.affine.pro/blob-1-0?sig=test',
|
||||
mimeType: 'audio/opus',
|
||||
},
|
||||
]);
|
||||
t.like(complete.firstCall.args[1], {
|
||||
status: 'ready',
|
||||
@@ -471,6 +491,7 @@ test('transcriptTask runs native transcript recipe through action bridge when av
|
||||
complete.firstCall.args[1].protectedResult.normalizedTranscript,
|
||||
'00:00:05 A: Kickoff'
|
||||
);
|
||||
t.deepEqual(complete.firstCall.args[1].protectedResult.infos, payload.infos);
|
||||
});
|
||||
|
||||
test('transcriptTask fails task when native action bridge reports an error event', async t => {
|
||||
|
||||
@@ -5,6 +5,7 @@ import ava, { TestFn } from 'ava';
|
||||
import { createTestingModule, type TestingModule } from '../../__tests__/utils';
|
||||
import { Models } from '../../models';
|
||||
import { BackfillPermissionProjection1765500000000 } from '../migrations/1765500000000-backfill-permission-projection';
|
||||
import { BackfillTranscriptStorageKeys1786805802350 } from '../migrations/1786805802350-backfill-transcript-storage-keys';
|
||||
|
||||
interface Context {
|
||||
module: TestingModule;
|
||||
@@ -81,3 +82,44 @@ test('permission backfill repairs ownerless workspaces before runtime state proj
|
||||
{ role: 'owner' }
|
||||
);
|
||||
});
|
||||
|
||||
test('transcript backfill adds stable keys without removing compatibility URLs', async t => {
|
||||
const payload = {
|
||||
sourceAudio: { blobId: 'blob-1' },
|
||||
infos: [
|
||||
{
|
||||
url: 'https://affine.example/api/copilot/blob/user-1/workspace-1/blob-1',
|
||||
mimeType: 'audio/m4a',
|
||||
},
|
||||
{
|
||||
url: 'https://example.com/external.m4a',
|
||||
mimeType: 'audio/m4a',
|
||||
},
|
||||
],
|
||||
};
|
||||
await t.context.db.aiTranscriptTask.create({
|
||||
data: {
|
||||
userId: 'user-1',
|
||||
workspaceId: 'workspace-1',
|
||||
blobId: 'blob-1',
|
||||
status: 'failed',
|
||||
recipeId: 'transcript.audio',
|
||||
recipeVersion: 'v1',
|
||||
inputSnapshot: payload,
|
||||
protectedResult: payload,
|
||||
},
|
||||
});
|
||||
|
||||
await BackfillTranscriptStorageKeys1786805802350.up(t.context.db);
|
||||
await BackfillTranscriptStorageKeys1786805802350.up(t.context.db);
|
||||
|
||||
const task = await t.context.db.aiTranscriptTask.findFirstOrThrow({
|
||||
where: { blobId: 'blob-1' },
|
||||
});
|
||||
const expected = {
|
||||
...payload,
|
||||
infos: [{ ...payload.infos[0], key: 'blob-1' }, payload.infos[1]],
|
||||
};
|
||||
t.deepEqual(task.inputSnapshot, expected);
|
||||
t.deepEqual(task.protectedResult, expected);
|
||||
});
|
||||
|
||||
+96
@@ -0,0 +1,96 @@
|
||||
import { type Prisma, PrismaClient } from '@prisma/client';
|
||||
|
||||
const BATCH_SIZE = 100;
|
||||
|
||||
function isRecord(value: unknown): value is Record<string, unknown> {
|
||||
return !!value && typeof value === 'object' && !Array.isArray(value);
|
||||
}
|
||||
|
||||
function backfillTranscriptStorageKeys(
|
||||
payload: Prisma.JsonValue | null,
|
||||
userId: string,
|
||||
workspaceId: string
|
||||
): Prisma.InputJsonValue | undefined {
|
||||
if (!isRecord(payload) || !Array.isArray(payload.infos)) {
|
||||
return;
|
||||
}
|
||||
|
||||
const prefix = `/api/copilot/blob/${encodeURIComponent(userId)}/${encodeURIComponent(workspaceId)}/`;
|
||||
let changed = false;
|
||||
const infos = payload.infos.map(info => {
|
||||
if (!isRecord(info) || info.key || typeof info.url !== 'string') {
|
||||
return info;
|
||||
}
|
||||
|
||||
try {
|
||||
const url = new URL(info.url);
|
||||
if (!url.pathname.startsWith(prefix)) {
|
||||
return info;
|
||||
}
|
||||
const key = decodeURIComponent(url.pathname.slice(prefix.length));
|
||||
if (!key) {
|
||||
return info;
|
||||
}
|
||||
changed = true;
|
||||
return { ...info, key };
|
||||
} catch {
|
||||
return info;
|
||||
}
|
||||
});
|
||||
|
||||
return changed ? ({ ...payload, infos } as Prisma.InputJsonValue) : undefined;
|
||||
}
|
||||
|
||||
export class BackfillTranscriptStorageKeys1786805802350 {
|
||||
static async up(db: PrismaClient) {
|
||||
let cursor: string | undefined;
|
||||
|
||||
while (true) {
|
||||
const tasks = await db.aiTranscriptTask.findMany({
|
||||
select: {
|
||||
id: true,
|
||||
userId: true,
|
||||
workspaceId: true,
|
||||
inputSnapshot: true,
|
||||
protectedResult: true,
|
||||
},
|
||||
orderBy: { id: 'asc' },
|
||||
take: BATCH_SIZE,
|
||||
...(cursor ? { cursor: { id: cursor }, skip: 1 } : {}),
|
||||
});
|
||||
if (!tasks.length) {
|
||||
return;
|
||||
}
|
||||
|
||||
const updates = tasks.flatMap(task => {
|
||||
const inputSnapshot = backfillTranscriptStorageKeys(
|
||||
task.inputSnapshot,
|
||||
task.userId,
|
||||
task.workspaceId
|
||||
);
|
||||
const protectedResult = backfillTranscriptStorageKeys(
|
||||
task.protectedResult,
|
||||
task.userId,
|
||||
task.workspaceId
|
||||
);
|
||||
if (!inputSnapshot && !protectedResult) {
|
||||
return [];
|
||||
}
|
||||
return db.aiTranscriptTask.update({
|
||||
where: { id: task.id },
|
||||
data: {
|
||||
...(inputSnapshot ? { inputSnapshot } : {}),
|
||||
...(protectedResult ? { protectedResult } : {}),
|
||||
},
|
||||
});
|
||||
});
|
||||
|
||||
if (updates.length) {
|
||||
await db.$transaction(updates);
|
||||
}
|
||||
cursor = tasks.at(-1)?.id;
|
||||
}
|
||||
}
|
||||
|
||||
static async down(_db: PrismaClient) {}
|
||||
}
|
||||
@@ -6,3 +6,4 @@ export * from './1751966744168-correct-session-update-time';
|
||||
export * from './1763800000000-rebuild-manticore-mixed-script-indexes';
|
||||
export * from './1765500000000-backfill-permission-projection';
|
||||
export * from './1765600000000-backfill-entitlement-projection';
|
||||
export * from './1786805802350-backfill-transcript-storage-keys';
|
||||
|
||||
@@ -59,14 +59,31 @@ export class CopilotStorage {
|
||||
): Promise<StorageRuntimeGetObjectResult> {
|
||||
const name = `${userId}/${workspaceId}/${key}`;
|
||||
if (signedUrl) {
|
||||
const presigned = await this.rt.presignGet('copilot', name);
|
||||
if (presigned) {
|
||||
return { redirectUrl: presigned.url };
|
||||
}
|
||||
const redirectUrl = await this.presignGet(userId, workspaceId, key);
|
||||
if (redirectUrl) return { redirectUrl };
|
||||
}
|
||||
return this.rt.getObject('copilot', name);
|
||||
}
|
||||
|
||||
async presignGet(userId: string, workspaceId: string, key: string) {
|
||||
return (
|
||||
await this.rt.presignGet('copilot', `${userId}/${workspaceId}/${key}`)
|
||||
)?.url;
|
||||
}
|
||||
|
||||
keyFromUrl(userId: string, workspaceId: string, url: string) {
|
||||
try {
|
||||
const parsed = new URL(url);
|
||||
const prefix = `/api/copilot/blob/${encodeURIComponent(userId)}/${encodeURIComponent(workspaceId)}/`;
|
||||
if (parsed.pathname.startsWith(prefix)) {
|
||||
return decodeURIComponent(parsed.pathname.slice(prefix.length));
|
||||
}
|
||||
} catch {
|
||||
return;
|
||||
}
|
||||
return undefined;
|
||||
}
|
||||
|
||||
@CallMetric('ai', 'blob_delete')
|
||||
async delete(userId: string, workspaceId: string, key: string) {
|
||||
await this.rt.deleteObject('copilot', `${userId}/${workspaceId}/${key}`);
|
||||
|
||||
@@ -18,7 +18,8 @@ export const LegacyTranscriptionSchema = z.array(
|
||||
);
|
||||
|
||||
export const AudioBlobInfoSchema = z.object({
|
||||
url: z.string(),
|
||||
key: z.string().min(1).optional(),
|
||||
url: z.string().min(1),
|
||||
mimeType: z.string(),
|
||||
index: z.number().int().nullable().optional(),
|
||||
});
|
||||
|
||||
@@ -46,10 +46,6 @@ export class CopilotTranscriptionService {
|
||||
private readonly realtime: RealtimePublisher
|
||||
) {}
|
||||
|
||||
private parseTaskPayload(payload: unknown): TranscriptionPayloadV2 {
|
||||
return TranscriptPayloadSchema.parse(payload);
|
||||
}
|
||||
|
||||
private buildTaskPublicMeta(payload: TranscriptionPayloadV2) {
|
||||
return {
|
||||
sourceAudio: payload.sourceAudio,
|
||||
@@ -68,13 +64,10 @@ export class CopilotTranscriptionService {
|
||||
const infos: AudioBlobInfos = [];
|
||||
for (const [idx, blob] of blobs.entries()) {
|
||||
const buffer = await readStream(blob.createReadStream());
|
||||
const url = await this.storage.put(
|
||||
userId,
|
||||
workspaceId,
|
||||
`${blobId}-${idx}`,
|
||||
buffer
|
||||
);
|
||||
const key = `${blobId}-${idx}`;
|
||||
const url = await this.storage.put(userId, workspaceId, key, buffer);
|
||||
infos.push({
|
||||
key,
|
||||
url,
|
||||
mimeType: sniffMime(buffer, blob.mimetype) || blob.mimetype,
|
||||
index: idx,
|
||||
@@ -103,6 +96,47 @@ export class CopilotTranscriptionService {
|
||||
} satisfies TranscriptionPayloadV2;
|
||||
}
|
||||
|
||||
private async resolveAttachmentUrl(
|
||||
userId: string,
|
||||
workspaceId: string,
|
||||
info: AudioBlobInfos[number]
|
||||
) {
|
||||
if (info.url.startsWith('data:')) {
|
||||
return info.url;
|
||||
}
|
||||
|
||||
const key =
|
||||
info.key ?? this.storage.keyFromUrl(userId, workspaceId, info.url);
|
||||
if (!key) {
|
||||
throw new Error('Transcript attachment cannot be resolved');
|
||||
}
|
||||
|
||||
const signedUrl = await this.storage.presignGet(userId, workspaceId, key);
|
||||
if (!signedUrl) {
|
||||
throw new Error('Transcript attachment signing is not configured');
|
||||
}
|
||||
return signedUrl;
|
||||
}
|
||||
|
||||
private async materializePayload(
|
||||
userId: string,
|
||||
workspaceId: string,
|
||||
payload: TranscriptionPayloadV2
|
||||
) {
|
||||
return {
|
||||
...payload,
|
||||
infos: payload.infos
|
||||
? await Promise.all(
|
||||
payload.infos.map(async info => ({
|
||||
url: await this.resolveAttachmentUrl(userId, workspaceId, info),
|
||||
mimeType: info.mimeType,
|
||||
index: info.index,
|
||||
}))
|
||||
)
|
||||
: payload.infos,
|
||||
} satisfies TranscriptionPayloadV2;
|
||||
}
|
||||
|
||||
private async buildTranscriptActionMessages(payload: TranscriptionPayloadV2) {
|
||||
const prompt = await this.prompts.get(TRANSCRIPT_PROMPT_REF);
|
||||
if (!prompt) {
|
||||
@@ -118,13 +152,12 @@ export class CopilotTranscriptionService {
|
||||
index: info.index ?? null,
|
||||
})) ?? null,
|
||||
};
|
||||
const attachments =
|
||||
payload.infos?.map(info => ({
|
||||
role: 'user' as const,
|
||||
content: `Audio attachment ${info.index ?? 0}`,
|
||||
attachments: [{ attachment: info.url, mimeType: info.mimeType }],
|
||||
params: { mimetype: info.mimeType },
|
||||
})) ?? [];
|
||||
const attachments = (payload.infos ?? []).map(info => ({
|
||||
role: 'user' as const,
|
||||
content: `Audio attachment ${info.index ?? 0}`,
|
||||
attachments: [{ attachment: info.url, mimeType: info.mimeType }],
|
||||
params: { mimetype: info.mimeType },
|
||||
}));
|
||||
return [
|
||||
...this.prompts.finish(prompt, {
|
||||
content: JSON.stringify(metadata),
|
||||
@@ -205,7 +238,7 @@ export class CopilotTranscriptionService {
|
||||
);
|
||||
}
|
||||
|
||||
const payload = this.parseTaskPayload(task.protectedResult);
|
||||
const payload = TranscriptPayloadSchema.parse(task.protectedResult);
|
||||
await this.runtime.assertRoute(
|
||||
'transcript.audio',
|
||||
{},
|
||||
@@ -287,14 +320,19 @@ export class CopilotTranscriptionService {
|
||||
let bridgeFailed = false;
|
||||
let bridgeError = 'transcript native recipe failed';
|
||||
let finalResult: unknown = null;
|
||||
const messages = await this.buildTranscriptActionMessages(payload);
|
||||
const runtimePayload = await this.materializePayload(
|
||||
task.userId,
|
||||
task.workspaceId,
|
||||
payload
|
||||
);
|
||||
const messages = await this.buildTranscriptActionMessages(runtimePayload);
|
||||
for await (const event of this.actionBridge.runStream({
|
||||
userId: task.userId,
|
||||
workspaceId: task.workspaceId,
|
||||
actionId: TRANSCRIPT_ACTION_ID,
|
||||
actionVersion: TRANSCRIPT_ACTION_VERSION,
|
||||
retryOf: retryOf ?? null,
|
||||
inputSnapshot: payload,
|
||||
inputSnapshot: runtimePayload,
|
||||
onRunCreated: async ({ runId }) => {
|
||||
await this.models.copilotTranscriptTask.markRunning(taskId, runId);
|
||||
this.publishTaskChanged(
|
||||
@@ -329,7 +367,10 @@ export class CopilotTranscriptionService {
|
||||
if (bridgeFailed) {
|
||||
throw new Error(bridgeError);
|
||||
}
|
||||
const parsedResult = TranscriptPayloadSchema.parse(finalResult);
|
||||
const parsedResult = {
|
||||
...TranscriptPayloadSchema.parse(finalResult),
|
||||
infos: payload.infos,
|
||||
} satisfies TranscriptionPayloadV2;
|
||||
await this.models.copilotTranscriptTask.complete(taskId, {
|
||||
status: 'ready',
|
||||
actionRunId,
|
||||
|
||||
Reference in New Issue
Block a user