Add Deepgram and OpenRouter voice processing

.gitea/workflows/build-and-deploy.yml

@@ -7,7 +7,7 @@ on:
 
 jobs:
   build:
-    runs-on: build-host
+    runs-on: builder
     env:
       SERVICE_NAME: worker
       IMAGE_SHA: gitea.dsrptlab.com/mapflow/worker:${{ github.sha }}
@@ -25,12 +25,6 @@ jobs:
       - name: Build and push image
         run: |
           set -euo pipefail
-          builder="builder"
-          if ! docker buildx inspect "$builder" >/dev/null 2>&1; then
-            docker buildx create --name "$builder" --driver docker-container --buildkitd-config /etc/buildkit/buildkitd.toml
-          fi
-          docker buildx use "$builder"
-          docker buildx inspect --bootstrap
           docker buildx build --push --tag "$IMAGE_SHA" --tag "$IMAGE_LATEST" .
 
       - name: Skip stale deployment
@@ -58,8 +52,3 @@ jobs:
             -d "$payload")"
           cat "$response_file"
           [ "$status_code" = "200" ]
-
-      - name: Prune shared BuildKit cache
-        run: |
-          set -euo pipefail
-          docker buildx prune --builder builder --all --max-used-space 40gb -f

prisma/schema.prisma

@@ -17,14 +17,16 @@ enum VoiceExperienceStatus {
 }
 
 model Place {
-  id            String            @id @default(cuid())
+  id                String            @id @default(cuid())
-  googlePlaceId String            @unique
+  googlePlaceId     String            @unique
-  name          String
+  name              String
-  latitude      Float
+  latitude          Float
-  longitude     Float
+  longitude         Float
-  experiences   VoiceExperience[]
+  googlePrimaryType String?
-  createdAt     DateTime          @default(now())
+  googleTypes       String[]          @default([])
-  updatedAt     DateTime          @updatedAt
+  experiences       VoiceExperience[]
+  createdAt         DateTime          @default(now())
+  updatedAt         DateTime          @updatedAt
 }
 
 model User {
@@ -35,22 +37,26 @@ model User {
   lastName         String?
   photoUrl         String?
   languageCode     String?
+  isAdmin          Boolean           @default(false)
   voiceExperiences VoiceExperience[]
   createdAt        DateTime          @default(now())
   updatedAt        DateTime          @updatedAt
 }
 
 model VoiceExperience {
-  id              String                @id @default(cuid())
+  id                 String                @id @default(cuid())
-  placeId          String
+  placeId            String
-  place            Place                 @relation(fields: [placeId], references: [id])
+  place              Place                 @relation(fields: [placeId], references: [id])
-  userId           String?
+  userId             String?
-  user             User?                 @relation(fields: [userId], references: [id])
+  user               User?                 @relation(fields: [userId], references: [id])
-  durationSeconds  Int
+  durationSeconds    Int
-  audioObjectKey   String
+  audioObjectKey     String
-  status           VoiceExperienceStatus @default(UPLOADED)
+  audioContentBase64 String?
-  transcript       String?
+  audioMimeType      String?
-  analysis         Json?
+  audioAccessToken   String?               @unique
-  createdAt        DateTime              @default(now())
+  status             VoiceExperienceStatus @default(UPLOADED)
-  updatedAt        DateTime              @updatedAt
+  transcript         String?
+  analysis           Json?
+  createdAt          DateTime              @default(now())
+  updatedAt          DateTime              @updatedAt
 }

src/config.ts

@@ -8,4 +8,11 @@ export const config = {
   databaseUrl: process.env.DATABASE_URL ?? '',
   workerName: process.env.HATCHET_WORKER_NAME ?? 'mapflow-hatchet-worker',
   workerSlots: Number.parseInt(process.env.HATCHET_WORKER_SLOTS ?? '4', 10),
+  publicApiUrl: process.env.PUBLIC_API_URL ?? 'https://api.map.craftee.vn',
+  deepgramApiKey: process.env.DEEPGRAM_API_KEY ?? '',
+  deepgramModel: process.env.DEEPGRAM_MODEL ?? 'nova-3',
+  deepgramLanguage: process.env.DEEPGRAM_LANGUAGE ?? 'ru',
+  openRouterApiKey: process.env.OPENROUTER_API_KEY ?? '',
+  openRouterModel:
+    process.env.OPENROUTER_MODEL ?? 'minimax/minimax-m2.5:free',
 };

src/generated/prisma/index-browser.js

@@ -126,6 +126,8 @@ exports.Prisma.PlaceScalarFieldEnum = {
   name: 'name',
   latitude: 'latitude',
   longitude: 'longitude',
+  googlePrimaryType: 'googlePrimaryType',
+  googleTypes: 'googleTypes',
   createdAt: 'createdAt',
   updatedAt: 'updatedAt'
 };
@@ -138,6 +140,7 @@ exports.Prisma.UserScalarFieldEnum = {
   lastName: 'lastName',
   photoUrl: 'photoUrl',
   languageCode: 'languageCode',
+  isAdmin: 'isAdmin',
   createdAt: 'createdAt',
   updatedAt: 'updatedAt'
 };
@@ -148,6 +151,9 @@ exports.Prisma.VoiceExperienceScalarFieldEnum = {
   userId: 'userId',
   durationSeconds: 'durationSeconds',
   audioObjectKey: 'audioObjectKey',
+  audioContentBase64: 'audioContentBase64',
+  audioMimeType: 'audioMimeType',
+  audioAccessToken: 'audioAccessToken',
   status: 'status',
   transcript: 'transcript',
   analysis: 'analysis',

src/generated/prisma/package.json

@@ -1,5 +1,5 @@
 {
-  "name": "prisma-client-359590a92aedb32557346cd24ba7dd77e7dd4a806bf36f6641715b1b22de87e3",
+  "name": "prisma-client-404088dc715856cb4ed2e7f2d4b8eb35c87427ce32c0e67bc04d6ff67961fc2b",
   "main": "index.js",
   "types": "index.d.ts",
   "browser": "default.js",

src/generated/prisma/schema.prisma

@@ -17,14 +17,16 @@ enum VoiceExperienceStatus {
 }
 
 model Place {
-  id            String            @id @default(cuid())
+  id                String            @id @default(cuid())
-  googlePlaceId String            @unique
+  googlePlaceId     String            @unique
-  name          String
+  name              String
-  latitude      Float
+  latitude          Float
-  longitude     Float
+  longitude         Float
-  experiences   VoiceExperience[]
+  googlePrimaryType String?
-  createdAt     DateTime          @default(now())
+  googleTypes       String[]          @default([])
-  updatedAt     DateTime          @updatedAt
+  experiences       VoiceExperience[]
+  createdAt         DateTime          @default(now())
+  updatedAt         DateTime          @updatedAt
 }
 
 model User {
@@ -35,22 +37,26 @@ model User {
   lastName         String?
   photoUrl         String?
   languageCode     String?
+  isAdmin          Boolean           @default(false)
   voiceExperiences VoiceExperience[]
   createdAt        DateTime          @default(now())
   updatedAt        DateTime          @updatedAt
 }
 
 model VoiceExperience {
-  id              String                @id @default(cuid())
+  id                 String                @id @default(cuid())
-  placeId         String
+  placeId            String
-  place           Place                 @relation(fields: [placeId], references: [id])
+  place              Place                 @relation(fields: [placeId], references: [id])
-  userId          String?
+  userId             String?
-  user            User?                 @relation(fields: [userId], references: [id])
+  user               User?                 @relation(fields: [userId], references: [id])
-  durationSeconds Int
+  durationSeconds    Int
-  audioObjectKey  String
+  audioObjectKey     String
-  status          VoiceExperienceStatus @default(UPLOADED)
+  audioContentBase64 String?
-  transcript      String?
+  audioMimeType      String?
-  analysis        Json?
+  audioAccessToken   String?               @unique
-  createdAt       DateTime              @default(now())
+  status             VoiceExperienceStatus @default(UPLOADED)
-  updatedAt       DateTime              @updatedAt
+  transcript         String?
+  analysis           Json?
+  createdAt          DateTime              @default(now())
+  updatedAt          DateTime              @updatedAt
 }

src/jobs/analyze-voice-experience.ts

@@ -1,6 +1,27 @@
-import { buildPlaceAnalysis } from '../ontology/place-ontology.js';
+import { config } from '../config.js';
+import { placeOntology } from '../ontology/place-ontology.js';
 import { prisma } from '../prisma.js';
 
+type OpenRouterResponse = {
+  choices?: Array<{
+    message?: {
+      content?: string;
+    };
+  }>;
+};
+
+type PlaceAnalysis = {
+  placeName: string;
+  tags: string[];
+  signals: Array<{
+    axis: string;
+    leaf: string;
+    evidence: string;
+    confidence: number;
+  }>;
+  summary: string;
+};
+
 export async function analyzeVoiceExperience(experienceId: string) {
   const experience = await prisma.voiceExperience.findUnique({
     where: { id: experienceId },
@@ -20,7 +41,7 @@ export async function analyzeVoiceExperience(experienceId: string) {
     data: { status: 'ANALYZING' },
   });
 
-  const analysis = buildPlaceAnalysis({
+  const analysis = await buildPlaceAnalysis({
     placeName: experience.place.name,
     transcript: experience.transcript,
   });
@@ -35,3 +56,98 @@ export async function analyzeVoiceExperience(experienceId: string) {
 
   return { tags: analysis.tags };
 }
+
+async function buildPlaceAnalysis(input: {
+  placeName: string;
+  transcript: string;
+}): Promise<PlaceAnalysis> {
+  if (!config.openRouterApiKey) {
+    throw new Error('OPENROUTER_API_KEY is required for voice analysis.');
+  }
+
+  const ontology = placeOntology.map((axis) => ({
+    axis: axis.id,
+    options: axis.leaves.map((leaf) => ({
+      tag: `${axis.id}:${leaf.id}`,
+      keywords: leaf.keywords,
+    })),
+  }));
+  const response = await fetch('https://openrouter.ai/api/v1/chat/completions', {
+    method: 'POST',
+    headers: {
+      authorization: `Bearer ${config.openRouterApiKey}`,
+      'content-type': 'application/json',
+      'http-referer': 'https://map.craftee.vn',
+      'x-title': 'MapFlow',
+    },
+    body: JSON.stringify({
+      model: config.openRouterModel,
+      temperature: 0.1,
+      max_tokens: 900,
+      messages: [
+        {
+          role: 'system',
+          content:
+            'You classify voice reviews about places. Return only valid JSON. Use only allowed ontology tags.',
+        },
+        {
+          role: 'user',
+          content: JSON.stringify({
+            task:
+              'Analyze the transcript and select place-experience tags that are explicitly supported by the text. Do not invent facts.',
+            outputShape: {
+              placeName: 'string',
+              tags: ['axis:leaf'],
+              signals: [
+                {
+                  axis: 'string',
+                  leaf: 'string',
+                  evidence: 'short quote or paraphrase from transcript',
+                  confidence: 'number from 0 to 1',
+                },
+              ],
+              summary: 'one short sentence',
+            },
+            placeName: input.placeName,
+            ontology,
+            transcript: input.transcript,
+          }),
+        },
+      ],
+    }),
+  });
+
+  if (!response.ok) {
+    throw new Error(`OpenRouter analysis failed with ${response.status}.`);
+  }
+
+  const payload = (await response.json()) as OpenRouterResponse;
+  const content = payload.choices?.[0]?.message?.content;
+  if (!content) {
+    throw new Error('OpenRouter returned an empty analysis.');
+  }
+
+  const analysis = JSON.parse(content) as PlaceAnalysis;
+  assertValidAnalysis(analysis);
+  return analysis;
+}
+
+function assertValidAnalysis(analysis: PlaceAnalysis) {
+  const allowedTags = new Set(
+    placeOntology.flatMap((axis) =>
+      axis.leaves.map((leaf) => `${axis.id}:${leaf.id}`),
+    ),
+  );
+
+  if (!Array.isArray(analysis.tags)) {
+    throw new Error('OpenRouter analysis tags must be an array.');
+  }
+  for (const tag of analysis.tags) {
+    if (!allowedTags.has(tag)) {
+      throw new Error(`OpenRouter returned unsupported tag: ${tag}.`);
+    }
+  }
+  if (!Array.isArray(analysis.signals)) {
+    throw new Error('OpenRouter analysis signals must be an array.');
+  }
+}

src/jobs/transcribe-voice-experience.ts

@@ -1,5 +1,19 @@
+import { config } from '../config.js';
 import { prisma } from '../prisma.js';
 
+type DeepgramResponse = {
+  results?: {
+    channels?: Array<{
+      alternatives?: Array<{
+        transcript?: string;
+        paragraphs?: {
+          transcript?: string;
+        };
+      }>;
+    }>;
+  };
+};
+
 export async function transcribeVoiceExperience(experienceId: string) {
   const experience = await prisma.voiceExperience.findUnique({
     where: { id: experienceId },
@@ -14,11 +28,13 @@ export async function transcribeVoiceExperience(experienceId: string) {
     data: { status: 'TRANSCRIBING' },
   });
 
-  const transcript = await transcribeAudioObject(experience.audioObjectKey);
+  const transcript = await transcribeAudioObject(experience.id);
 
   await prisma.voiceExperience.update({
     where: { id: experienceId },
     data: {
+      audioContentBase64: null,
+      audioAccessToken: null,
       transcript,
       status: 'TRANSCRIBED',
     },
@@ -27,7 +43,68 @@ export async function transcribeVoiceExperience(experienceId: string) {
   return { transcript };
 }
 
-async function transcribeAudioObject(audioObjectKey: string): Promise<string> {
+async function transcribeAudioObject(experienceId: string): Promise<string> {
-  // TODO: replace this adapter with the production speech-to-text provider.
+  if (!config.deepgramApiKey) {
-  return `Transcription placeholder for ${audioObjectKey}`;
+    throw new Error('DEEPGRAM_API_KEY is required for transcription.');
+  }
+
+  const experience = await prisma.voiceExperience.findUnique({
+    where: { id: experienceId },
+    select: {
+      id: true,
+      audioObjectKey: true,
+      audioAccessToken: true,
+      audioContentBase64: true,
+      audioMimeType: true,
+    },
+  });
+
+  if (!experience) {
+    throw new Error(`Voice experience ${experienceId} was not found.`);
+  }
+  if (!experience.audioAccessToken) {
+    throw new Error(
+      `Voice audio ${experience.audioObjectKey} has no access token.`,
+    );
+  }
+  if (!experience.audioContentBase64 || !experience.audioMimeType) {
+    throw new Error(
+      `Voice audio ${experience.audioObjectKey} has no stored content.`,
+    );
+  }
+
+  const params = new URLSearchParams({
+    model: config.deepgramModel,
+    language: config.deepgramLanguage,
+    smart_format: 'true',
+    punctuate: 'true',
+  });
+  const audioUrl = `${config.publicApiUrl.replace(/\/$/, '')}/audio/voice-experiences/${experience.id}?token=${encodeURIComponent(experience.audioAccessToken)}`;
+  const response = await fetch(
+    `https://api.deepgram.com/v1/listen?${params.toString()}`,
+    {
+      method: 'POST',
+      headers: {
+        authorization: `Token ${config.deepgramApiKey}`,
+        'content-type': 'application/json',
+      },
+      body: JSON.stringify({ url: audioUrl }),
+    },
+  );
+
+  if (!response.ok) {
+    throw new Error(`Deepgram transcription failed with ${response.status}.`);
+  }
+
+  const payload = (await response.json()) as DeepgramResponse;
+  const alternative = payload.results?.channels?.[0]?.alternatives?.[0];
+  const transcript =
+    alternative?.paragraphs?.transcript?.trim() ??
+    alternative?.transcript?.trim() ??
+    '';
+  if (!transcript) {
+    throw new Error('Deepgram returned an empty transcript.');
+  }
+
+  return transcript;
 }