Merge pull request #15 from SunbirdAI/asr-stt-inference

Integrate ASR/STT inference

.flake8

@@ -0,0 +1,5 @@
+[flake8]
+    ignore = D203, E402, F403, F405, W503, W605
+    exclude = .git,env,__pycache__,docs/source/conf.py,old,build,dist, *migrations*,env,venv,alembic
+    max-complexity = 10
+    max-line-length = 119

.isort.cfg

@@ -0,0 +1,7 @@
+[settings]
+multi_line_output=3
+include_trailing_comma=True
+force_grid_wrap=0
+use_parentheses=True
+line_length=88
+skip=env,migrations,alembic,venv

Makefile

@@ -0,0 +1,12 @@
+.PHONY: lint-apply lint-check
+
+lint-check:
+	@echo "Checking for lint errors..."
+	flake8 .
+	black --check .
+	isort --check-only .
+
+lint-apply:
+	@echo "apply linting ..."
+	black .
+	isort .

app/api.py

@@ -1,33 +1,36 @@
-from fastapi import FastAPI, Request
+import os
 from functools import partial
-from app.routers.tasks import router as tasks_router
-from app.routers.auth import router as auth_router
-from app.routers.frontend import router as frontend_router
-from app.middleware.monitoring_middleware import log_request
-from app.docs import description, tags_metadata
-from fastapi.middleware.cors import CORSMiddleware
-from fastapi.staticfiles import StaticFiles
+from pathlib import Path
 
 import redis.asyncio as redis
-from fastapi_limiter import FastAPILimiter
 from dotenv import load_dotenv
-from pathlib import Path
-import os
+from fastapi import FastAPI
+from fastapi.middleware.cors import CORSMiddleware
+from fastapi.staticfiles import StaticFiles
+from fastapi_limiter import FastAPILimiter
+
+from app.docs import description, tags_metadata
+from app.middleware.monitoring_middleware import log_request
+from app.routers.auth import router as auth_router
+from app.routers.frontend import router as frontend_router
+from app.routers.tasks import router as tasks_router
 
 load_dotenv()
 
 
 app = FastAPI(
-    title="Sunbird AI API",
-    description=description,
-    openapi_tags=tags_metadata
+    title="Sunbird AI API", description=description, openapi_tags=tags_metadata
 )
 
+
 @app.on_event("startup")
 async def startup():
-    redis_instance = redis.from_url(os.getenv('REDIS_URL'), encoding="utf-8", decode_responses=True)
+    redis_instance = redis.from_url(
+        os.getenv("REDIS_URL"), encoding="utf-8", decode_responses=True
+    )
     await FastAPILimiter.init(redis_instance)
 
+
 static_files_directory = Path(__file__).parent.absolute() / "static"
 app.mount("/static", StaticFiles(directory=static_files_directory), name="static")
 
@@ -41,7 +44,7 @@ async def startup():
     allow_origins=origins,
     allow_credentials=True,
     allow_methods=["*"],
-    allow_headers=["*"]
+    allow_headers=["*"],
 )
 
 

app/crud/monitoring.py

@@ -1,27 +1,35 @@
 from contextlib import contextmanager
 
 from sqlalchemy.orm import Session
+
+from app.database.db import SessionLocal
 from app.models import monitoring as models
 from app.schemas import monitoring as schemas
-from app.database.db import SessionLocal
+
 
 @contextmanager
 def auto_session():
     sess = SessionLocal()
     try:
         yield sess
         sess.commit()
-    except:
+    except Exception:
         sess.rollback()
     finally:
         sess.close()
 
 
 def create_endpoint_log(log: schemas.EndpointLog):
     with auto_session() as sess:
-        db_log = models.EndpointLog(username=log.username, endpoint=log.endpoint, time_taken=log.time_taken)
+        db_log = models.EndpointLog(
+            username=log.username, endpoint=log.endpoint, time_taken=log.time_taken
+        )
         sess.add(db_log)
 
 
 def get_logs_by_username(db: Session, username: str):
-    return db.query(models.EndpointLog).filter(models.EndpointLog.username == username).all()
+    return (
+        db.query(models.EndpointLog)
+        .filter(models.EndpointLog.username == username)
+        .all()
+    )

app/crud/users.py

@@ -1,10 +1,16 @@
 from sqlalchemy.orm import Session
-from app.schemas import users as schema
+
 from app.models import users as models
+from app.schemas import users as schema
 
 
 def create_user(db: Session, user: schema.UserInDB) -> schema.User:
-    db_user = models.User(email=user.email, username=user.username, organization=user.organization, hashed_password=user.hashed_password)
+    db_user = models.User(
+        email=user.email,
+        username=user.username,
+        organization=user.organization,
+        hashed_password=user.hashed_password,
+    )
     db.add(db_user)
     db.commit()
     db.refresh(db_user)

app/database/db.py

@@ -1,13 +1,12 @@
 import os
 
-from sqlalchemy.orm import declarative_base
-from sqlalchemy import create_engine
-from sqlalchemy.orm import sessionmaker
 from dotenv import load_dotenv
+from sqlalchemy import create_engine
+from sqlalchemy.orm import declarative_base, sessionmaker
 
 load_dotenv()
 
-engine = create_engine(os.getenv('DATABASE_URL'), pool_size=20, max_overflow=0)
+engine = create_engine(os.getenv("DATABASE_URL"), pool_size=20, max_overflow=0)
 SessionLocal = sessionmaker(autocommit=False, autoflush=False, bind=engine)
 
 Base = declarative_base()

app/deps.py

@@ -1,5 +1,6 @@
 from app.database.db import SessionLocal
 
+
 def get_db():
     db = SessionLocal()
     try:

app/docs.py

@@ -1,5 +1,5 @@
 description = """
-Welcome to the Sunbird AI API documentation. The Sunbird AI API provides you access to Sunbird's language models. The currently supported models are:
+Welcome to the Sunbird AI API documentation. The Sunbird AI API provides you access to Sunbird's language models. The currently supported models are:  # noqa E501
 - **Translation (English to Multiple)**: translate from English to Acholi, Ateso, Luganda, Lugbara and Runyankole.
 - **Translation (Multiple to English)**: translate from the 5 local language above to English.
 - **Speech To Text (Luganda)**: Convert Luganda speech audio to text.
@@ -25,12 +25,9 @@
 """
 
 tags_metadata = [
-    {
-        "name": "AI Tasks",
-        "description": "Operations for AI inference."
-    },
+    {"name": "AI Tasks", "description": "Operations for AI inference."},
     {
         "name": "Authentication Endpoints",
-        "description": "Operations for Authentication, including Sign up and Login"
-    }
+        "description": "Operations for Authentication, including Sign up and Login",
+    },
 ]

app/inference_services/base.py

@@ -1,24 +1,17 @@
 import json
+import os
+
 import requests
 from dotenv import load_dotenv
-import os
 
 load_dotenv()
 
-url = f"https://europe-west1-aiplatform.googleapis.com/v1/projects/{os.getenv('PROJECT_ID')}/locations/europe-west1/endpoints/{os.getenv('ENDPOINT_ID')}:rawPredict"
+url = f"https://europe-west1-aiplatform.googleapis.com/v1/projects/{os.getenv('PROJECT_ID')}/locations/europe-west1/endpoints/{os.getenv('ENDPOINT_ID')}:rawPredict"  # noqa E501
 
 
 def inference_request(payload):
-    token = os.popen('gcloud auth print-access-token').read().strip()
-    headers = {
-        "Authorization": f"Bearer {token}",
-        "Content-Type": "application/json"
-    }
-    response = requests.request(
-        "POST",
-        url,
-        headers=headers,
-        data=json.dumps(payload)
-    )
+    token = os.popen("gcloud auth print-access-token").read().strip()
+    headers = {"Authorization": f"Bearer {token}", "Content-Type": "application/json"}
+    response = requests.request("POST", url, headers=headers, data=json.dumps(payload))
 
     return response

app/inference_services/stt_inference.py

@@ -1,5 +1,6 @@
 import base64
 from io import BytesIO
+
 from app.inference_services.base import inference_request
 
 
@@ -11,21 +12,15 @@ def create_payload(audio_file):
         audio_bytes = BytesIO(contents)
     encoded_audio = base64.b64encode(audio_bytes.read())
 
-    utf_audio = encoded_audio.decode('utf-8')
+    utf_audio = encoded_audio.decode("utf-8")
 
-    payload = {
-        "instances": [
-            {
-                "audio": utf_audio,
-                "task": "asr"
-            }
-        ]
-    }
+    payload = {"instances": [{"audio": utf_audio, "task": "asr"}]}
 
     return payload
 
+
 def transcribe(audio_file):
     # TODO: Handle error cases
     payload = create_payload(audio_file)
     response = inference_request(payload).json()
-    return response['transcripts'][0]
+    return response["transcripts"][0]

app/inference_services/translate_inference.py

@@ -3,7 +3,11 @@
 
 
 def get_task(target_language):
-    return 'translate_from_english' if target_language != 'English' else 'translate_to_english'
+    return (
+        "translate_from_english"
+        if target_language != "English"
+        else "translate_to_english"
+    )
 
 
 def create_payload(text, source_language=None, target_language=None):
@@ -14,45 +18,48 @@ def create_payload(text, source_language=None, target_language=None):
                 "sentence": text,
                 "task": task,
                 "target_language": target_language,
-                "source_language": source_language
+                "source_language": source_language,
             }
         ]
     }
 
     return payload
 
+
 def create_batch_payload(request: TranslationBatchRequest):
     payload = {
         "instances": [
             {
                 "sentence": request.text,
                 "task": get_task(request.target_language),
                 "target_language": request.target_language.value,
-                "source_language": request.source_language.value
+                "source_language": request.source_language.value,
             }
             for request in request.requests
         ]
     }
 
     return payload
 
+
 def translate(text, source_language=None, target_language=None):
     payload = create_payload(text, source_language, target_language)
     response = inference_request(payload).json()
     # TODO: Handle error cases i.e if there's an error from the inference server.
-    if target_language == 'English':
+    if target_language == "English":
         response = response["to_english_translations"][0]
     else:
         response = response["from_english_translations"][0]
     return response
 
+
 def translate_batch(request: TranslationBatchRequest):
     payload = create_batch_payload(request)
     response = inference_request(payload).json()
     response_list = []
-    if 'to_english_translations' in response:
-        response_list = response['to_english_translations']
-    if 'from_english_translations' in response:
-        response_list.extend(response['from_english_translations'])
+    if "to_english_translations" in response:
+        response_list = response["to_english_translations"]
+    if "from_english_translations" in response:
+        response_list.extend(response["from_english_translations"])
 
     return response_list

app/inference_services/tts_inference.py

@@ -1,25 +1,18 @@
-from app.inference_services.base import inference_request
-from app.schemas.tasks import TTSRequest
-
-from google.cloud import storage
-
 import base64
-import uuid
 import os
+import uuid
+
 from dotenv import load_dotenv
+from google.cloud import storage
+
+from app.inference_services.base import inference_request
+from app.schemas.tasks import TTSRequest
 
 load_dotenv()
 
 
 def create_payload(text):
-    payload = {
-        "instances": [
-            {
-                "sentence": text,
-                "task": "tts"
-            }
-        ]
-    }
+    payload = {"instances": [{"sentence": text, "task": "tts"}]}
 
     return payload
 
@@ -42,7 +35,7 @@ def tts(request: TTSRequest):
 
         os.environ["GOOGLE_APPLICATION_CREDENTIALS"] = "service_account_key.json"
         bucket_name = os.getenv("GOOGLE_CLOUD_BUCKET_NAME")
-        bucket_file = f"{str(uuid.uuid4())}.wav" # using a uuid for the audio file name
+        bucket_file = f"{str(uuid.uuid4())}.wav"  # using a uuid for the audio file name
         client = storage.Client()
         bucket = client.bucket(bucket_name)
         blob = bucket.blob(bucket_file)