Handling different input types in FastAPI

Deploying AI into Production with FastAPI

Matt Eckerle

Software and Data Engineering Leader

Restaurant vs API

A restaurant being compared to inputs to an API

Validation flow

Incoming data via request
Input data validation happens using Pydantic
Process different types of data as per model requirements
Processed input sent to the model

A flowchart explaining request validation

Comment moderation system

class CommentMetrics(BaseModel):
    length: int           
    user_karma: int     
    report_count: int

class CommentText(BaseModel):
    content: str

An image of comment box icons

Endpoint for floating point numbers

app = FastAPI()

@app.post("/predict")

def predict_score(data: CommentMetrics):

    features = np.array([
        data.length,     
        data.user_karma,
        data.report_count
    ])

    model = CommentScorer()
    prediction = model.predict(features)

    return {"prediction": round(prediction, 2),
            "input": data.dict()}

Endpoint for textual input

@app.post("/analyze_text")

def analyze(comment: CommentText):

    forbidden = ["spam", "hate", "free"
                 "fake", "sign up"]

    text_lower = comment.lower()

    issues = [word for word in forbidden 
              if word in text_lower]

    return {
        "issues": issues,
        "needs_moderation": len(issues)
    }

Output for comment: Sign up for free

{
    "issues": ["free", "sign up"],
    "needs_moderation": 2
}

Let's practice!

Deploying AI into Production with FastAPI