Interview Transcriber

.github/workflows/docker-image-deploy.yml

@@ -0,0 +1,25 @@
+
+name: Docker Image CD
+
+on:
+  push:
+    branches: [ main ]
+  pull_request:
+    branches: [ main ]
+
+jobs:
+  Build-and-Push-Image:
+    runs-on: ubuntu-latest
+    name: Docker Build, Tag, Push
+
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v1
+      - name: Build container image
+        uses: docker/build-push-action@v1
+        with:
+          username: ${{github.actor}}
+          password: ${{secrets.GITHUB_TOKEN}}
+          registry: docker.pkg.github.com
+          repository: umass-rescue/596-s22-backend/backend
+          tags: latest

.github/workflows/docker-image.yml

@@ -0,0 +1,18 @@
+name: Docker Image CI
+
+on:
+  push:
+    branches: [ main ]
+  pull_request:
+    branches: [ main ]
+
+jobs:
+
+  build:
+
+    runs-on: ubuntu-latest
+
+    steps:
+    - uses: actions/checkout@v2
+    - name: Build the Docker image
+      run: docker build . --file Dockerfile --tag my-image-name:$(date +%s)

Dockerfile

@@ -4,6 +4,15 @@ WORKDIR /rescue
 
 COPY requirements.txt requirements.txt
 
+RUN apt-get update && apt-get install -f -y postgresql-client
+
 RUN pip install --no-cache-dir --upgrade -r /rescue/requirements.txt
 
-COPY ./app /rescue/app
+COPY alembic/ ./alembic/
+COPY alembic.ini .
+COPY startup.sh .
+
+COPY ./app /rescue/app
+
+# Run alembic configuration
+CMD ["./startup.sh"]

README.md

@@ -21,10 +21,20 @@ Backend Repo for 596RL Spring 2022
 - SQLAlchemy  (Locally for database manipulation)
 ```
 
+### 🤝 How to Run the Package in Another Repo
+To learn how to run the backend in YOUR repo, follow this [guide](https://www.notion.so/Setting-Up-Backend-Docker-Container-dd2ce1e805e84b44a245991c96d46591) here.
+
+
 ### 👀 Configuring Secrets
 To successfully connect to the database, we use a `.env` file, which you will need to generate using:
 `touch .env` in the backend directory. This is an untracked file in our repo. Here you will need to paste in the secrets of the backend. DM Sam DuBois and he will send you the file.
 
+```
+# PostgreSQL Container Secrets
+POSTGRES_USER=
+POSTGRES_PASSWORD=
+POSTGRES_DB=
+```
 
 ### 🚀 How to Run the Container
 
@@ -34,8 +44,8 @@ Make sure the Docker and the postgresSQL library are working.
 
 From there, all you need to do to run the container is:
 ```
-docker-compose up -d --build
-docker-compose up
+docker compose build
+docker compose up
 ```
 
 ### 🧪 Useful Tools for Debugging and Testing

alembic.ini

@@ -52,7 +52,7 @@ version_path_separator = os  # Use os.pathsep. Default configuration used for ne
 # are written from script.py.mako
 # output_encoding = utf-8
 
-sqlalchemy.url = postgresql://username:password@localhost:5432/default_database
+sqlalchemy.url = postgresql://username:password@db:5432/default_database
 
 
 [post_write_hooks]

app/crud.py

@@ -119,4 +119,78 @@ def get_footage_for_plate_id(license_plate_number: str, db: Session, skip: int =
         if plate.footage_id not in data:
             data[plate.footage_id] = db.query(models.LicenseFootage).filter(models.LicenseFootage.id == plate.footage_id).offset(skip).limit(limit).all()[0]
 
-    return list(data.values())
+    return list(data.values())
+
+# # Create Interview object with audio file name
+# ##THIS IS ASSUMING AUDIO FILES HAS ALREADY BEEN SAVED TO DATABASE
+# def create_interview(audio_filename: str, db: Session):
+
+#     # Insert code to retrieve audio file from db
+
+#     ####################
+#     ## YOUR CODE HERE ##
+#     ####################
+
+#     # return type - audio file of interview
+#     # NEED TO FIGURE OUT IF AUDIO FILE OR PATH OF FILE SHOULD BE RETURNED
+
+#     #following this post: https://stackoverflow.com/questions/64558200/python-requests-in-docker-compose-containers
+#     path = "the path to the server" + "/sendTranscription"
+#     response = session.post(path, json={"audio_filename": audio_filename})
+
+#     #get the full text transcription from the response
+#     reponse_json = json.loads(response.text)
+#     full_text = reponse_json['transcription']
+
+#     # Add Interview object
+#     db_message = models.Interview(filename=filename, full_name = interview.full_name,
+#                                   created_at = interview.created_at,
+#                                   address = interview.address,
+#                                   case = interview.case, full_text = full_text)
+#     db.add(db_message)
+#     db.commit()
+#     db.refresh(db_message)
+
+#     return db_message
+
+# # Create Interview object with audio file name
+# ##THIS IS ASSUMING AUDIO FILES HAS ALREADY BEEN SAVED TO DATABASE
+# def analyze_interview(case: int, audio_filename: str, db: Session):
+
+#     #get full text for this interview
+#     full_text = db.query(models.Interview).filter(models.Interview.filename == audio_filename).offset(skip).limit(limit).all()
+
+#     #get list of questions for this interview
+#     questions = db.query(models.Question).filter(models.Question.case == case).offset(skip).limit(limit).all()
+#     add_questions = db.query(models.Additional_Question).filter(models.Additional_Question.interview_id == question_answer_pair.interview_id).offset(skip).limit(limit).all()
+#     questions = questions + add_questions
+
+#     #following this post: https://stackoverflow.com/questions/64558200/python-requests-in-docker-compose-containers
+#     path = "the path to the server" + "/analyzeText"
+#     response = session.post(path, json={"full_text": full_text, "questions": questions})
+
+#     #get the json from the response
+#     reponse_json = json.loads(response.text)
+#     all_pairs = reponse_json['transcription']
+
+#     # Add each pair and NER
+#     # CAN I ADD MULTIPLE OBJECTS TO THE DB IN ONE COMMIT?
+#     for pair in all_pairs:
+#         db_pair_message = models.Question_Answer_Pair(interview_id=interview.id, question=pair['question'], answer=pair['answer'])
+#         db.add(db_pair_message)
+#         db.flush()  #allows for the primary key of the pair to be generated
+#         for ner in pair['ner']:
+#             db_ner_message = models.Answer_NER(question_answer_pair_id=db_pair_message.id, label=ner['label'], answer=pair['answer'])
+#             db.add(db_ner_message)
+#         db.commit()
+#         db.refresh(db_pair_message)
+
+#     return all_pairs
+
+# # Get all question and answer pairs for an interview
+# def get_question_answer_pairs(interview_id: int, db: Session, skip: int = 0, limit: int = 100):
+#     return db.query(models.Question_Answer_Pair).filter(models.Question_Answer_Pair.interview_id == interview_id).offset(skip).limit(limit).all()
+
+# # Get all ner objects for an answer
+# def get_ners_for_answer(question_answer_pair_id: int, db: Session, skip: int = 0, limit: int = 100):
+#     return db.query(models.Answer_NER).filter(models.Answer_NER.question_answer_pair_id == question_answer_pair_id).offset(skip).limit(limit).all()

app/main.py

@@ -88,7 +88,31 @@ def get_plates_for_footage_id(footage_id: int, skip: int = 0, limit: int= 100, d
     plates = crud.get_license_plates_for_filename(footage_id=footage_id, skip=skip, limit=limit, db=db)
     return plates
 
+# Route - GET - get all footage for a specific plate number
 @app.get("/licenses/plates/{license_plate_number}", response_model=List[schemas.LicenseFootage])
 def get_footage_for_plate_id(license_plate_number: str, skip: int = 0, limit: int = 100, db: Session = Depends(get_db)):
     footage = crud.get_footage_for_plate_id(license_plate_number=license_plate_number, skip=skip, limit=limit, db=db)
-    return footage
+    return footage
+
+# # Route - POST - Create a interview object and add transcribed full text
+# @app.post("/transciber/{audio_filename}", response_model=schemas.LicenseFootage)
+# def create_interview(license_footage: schemas.CreateLicenseFootageObj, db:Session = Depends(get_db)):
+#     return crud.create_interview(audio_filename=audio_filename, db=db)
+
+# # Route - POST - Create question_answer_pairs and ners from analyzed text
+# #HOW ARE WE GOING TO DEAL WITH CASE NUMBERS? INCLUDE IN POST REQUEST?
+# @app.post("/transciber/{audio_filename}/analyze", response_model=schemas.LicenseFootage)
+# def create_interview(license_footage: schemas.CreateLicenseFootageObj, db:Session = Depends(get_db)):
+#     return crud.analyze_interview(case=case, audio_filename=audio_filename, db=db)
+
+# # Route - GET - get all question answer pairs for an interview_id
+# @app.get("/transcriber/{interview_id}/question_answer_pairs", response_model=List[schemas.Question_Answer_Pair])
+# def get_question_answer_pair_for_interview_id(interview_id: int, skip: int = 0, limit: int= 100, db: Session = Depends(get_db)):
+#     pairs = crud.get_question_answer_pairs(interview_id=interview_id, skip=skip, limit=limit, db=db)
+#     return pairs
+
+# # Route - GET - get all ners for an question_answer_pair_id
+# @app.get("/transcriber/{question_answer_pair_id}/ners", response_model=List[schemas.Question_Answer_Pair])
+# def get_ner_for_question_answer_pair_id(question_answer_pair_id: int, skip: int = 0, limit: int= 100, db: Session = Depends(get_db)):
+#     ners = crud.get_ners_for_answer(question_answer_pair_id=question_answer_pair_id, skip=skip, limit=limit, db=db)
+#     return ners

app/models.py

@@ -53,4 +53,53 @@ class RecognizedPlate(Base):
     time = Column(TIMESTAMP(timezone=False))
     footage_id = Column(Integer, ForeignKey("license_footage.id"))
 
-    footage = relationship("LicenseFootage", back_populates="recognized_plates")
+    footage = relationship("LicenseFootage", back_populates="recognized_plates")
+
+class Interview(Base):
+    __tablename__ = "interview"
+
+    id = Column(Integer, primary_key=True, index=True)
+    filename = Column(String, index=True)
+    full_name = Column(String, index=True)
+    created_at = Column(TIMESTAMP(timezone=False))
+    address = Column(String, index=True)
+    case = Column(Integer, ForeignKey("license_footage.id"))
+    full_text = Column(String, index=True)
+
+    interview = relationship("Interview", back_populates="interview")
+
+class Question_Answer_Pair(Base):
+    __tablename__ = "question_answer_pair"
+
+    id = Column(Integer, primary_key=True, index=True)
+    interview_id = Column(Integer, ForeignKey("interview.id")) ##Please check if this is correct
+    question = Column(String, index=True)
+    answer = Column(String, index=True)
+
+    question_answer_pair = relationship("Question_Answer_Pair", back_populates="question_answer_pair")
+
+class Question(Base):
+    __tablename__ = "question"
+
+    case = Column(Integer, index=True)
+    question = Column(String, index=True)
+
+question = relationship("Question", back_populates="question")
+
+class Additional_Question(Base):
+    __tablename__ = "additional_question"
+
+    interview_id = Column(Integer, ForeignKey("interview.id")) ##Please check if this is correct
+    question = Column(String, index=True)
+
+    additional_question = relationship("Additional_Question", back_populates="additional_question")
+
+class Answer_NER(Base):
+    __tablename__ = "answer_ner"
+
+    question_answer_pair_id = Column(Integer, ForeignKey("question_answer_pair.id")) ##Please check if this is correct
+    ner_label = Column(String, index=True)
+    start_index = Column(Integer, index=True)
+    end_index = Column(Integer, index=True)
+
+    answer_ner = relationship("Answer_NER", back_populates="answer_ner")

app/schemas.py

@@ -85,5 +85,78 @@ class LicenseFootage(LicenseFootageBase):
     date_uploaded = datetime.now()
     recognized_plates: List[RecognizedPlate] = []
 
+    class Config:
+        orm_mode = True
+
+
+############ Transcribed Interview Data #############
+
+class InterviewCreate(BaseModel):
+    filename: str
+    full_name: str
+    created_at: datetime = datetime
+    address: str
+    case: int
+
+class Interview(BaseModel):
+    id: int
+    filename: str
+    full_name: str
+    created_at: datetime = datetime
+    address: str
+    case: int
+    full_text: str
+
+    class Config:
+        orm_mode = True
+
+class Question_Answer_PairCreate(BaseModel):
+    interview_id: int ##I don't know how to link this to another table
+    question: str
+    answer: str
+
+class Question_Answer_Pair(BaseModel):
+    id: int
+    interview_id: int ##I don't know how to link this to another table
+    question: str
+    answer: str
+
+    class Config:
+        orm_mode = True
+
+class QuestionCreate(BaseModel):
+    case: int
+    question: str
+
+class Question(BaseModel):
+    case: int
+    question: str
+
+    class Config:
+        orm_mode = True
+
+class Additional_QuestionCreate(BaseModel):
+    interview_id: int ##I don't know how to link this to another table
+    question: str
+
+class Additional_Question(BaseModel):
+    interview_id: int ##I don't know how to link this to another table
+    question: str
+
+    class Config:
+        orm_mode = True
+
+class Answer_NERCreate(BaseModel):
+    question_answer_pair_id: int ##I don't know how to link this to another table
+    ner_label: str
+    start_index: int
+    end_index: int
+
+class Answer_NER(BaseModel):
+    question_answer_pair_id: int ##I don't know how to link this to another table
+    ner_label: str
+    start_index: int
+    end_index: int
+
     class Config:
         orm_mode = True

docker-compose.yml

@@ -2,24 +2,25 @@ version: '3.8'
 
 # Setting up the PostgreSQL DB Container
 services:
-  db:
-    image: 'postgres:latest'
-    restart: always
-    env_file: # The location we use to share all of our secrets
-      - .env
-    volumes:
-      - ./db-data/:/var/lib/postgresql/data/
-    ports:
-      - 5432:5432
-
   server:
     build:
       context: ./
       dockerfile: Dockerfile
+    depends_on:
+      - db
     volumes:
       - ./be-data/:/backend/
-    command: uvicorn app.main:app --host 0.0.0.0 --port 8000
     env_file:
       - .env
     ports:
       - 8000:8000
+
+  db:
+    image: 'postgres:latest'
+    restart: always
+    env_file: # The location we use to share all of our secrets
+      - .env
+    volumes:
+      - ./db-data/:/var/lib/postgresql/data/
+    ports:
+      - 5432:5432