Spaces:

Anabeldg
/

getaround-pricing-api

Sleeping

App Files Files Community

madamanastasia commited on Feb 18

Commit

aea70bb

1 Parent(s): e66b965

Fix predict matrix format + update docs

Browse files

Files changed (4) hide show

README.md +0 -29
docs.html +128 -85
main.py +65 -42
requirements.txt +1 -0

README.md DELETED Viewed

@@ -1,29 +0,0 @@
----
-title: Getaround Pricing API
-emoji: 🏢
-colorFrom: green
-colorTo: red
-sdk: docker
-app_port: 7860
-pinned: false
----
-## Production API
-Base URL:
-https://anabeldg-getaround-pricing-api.hf.space
-### Health check
-GET `/health`
-Example: https://anabeldg-getaround-pricing-api.hf.space/health
-### API documentation
-https://anabeldg-getaround-pricing-api.hf.space/docs
-### Example prediction request
-```bash
-curl -s -X POST "https://anabeldg-getaround-pricing-api.hf.space/predict" \
-  -H "Content-Type: application/json" \
-  -d '{"input":[{"model_key":"RENAULT","mileage":50000,"engine_power":110,"fuel":"diesel","paint_color":"black","car_type":"sedan","private_parking_available":true,"has_gps":true,"has_air_conditioning":true,"automatic_car":false,"has_getaround_connect":true,"has_speed_regulator":true,"winter_tires":false}]}' \
-| python -m json.tool

docs.html CHANGED Viewed

@@ -1,55 +1,114 @@
 <!doctype html>
 <html lang="en">
-  <head>
-    <meta charset="utf-8" />
-    <meta name="viewport" content="width=device-width,initial-scale=1" />
-    <title>Getaround — Pricing Optimization API</title>
-    <style>
-      body { font-family: -apple-system,BlinkMacSystemFont,Segoe UI,Roboto,Helvetica,Arial,sans-serif; line-height: 1.5; margin: 0; }
-      header { padding: 32px 20px; border-bottom: 1px solid #eee; }
-      main { max-width: 980px; margin: 0 auto; padding: 20px; }
-      h1 { margin: 0 0 8px 0; font-size: 28px; }
-      h2 { margin-top: 24px; font-size: 20px; }
-      p { margin: 8px 0; }
-      code, pre { font-family: ui-monospace,SFMono-Regular,Menlo,Monaco,Consolas,monospace; }
-      pre { background: #f7f7f8; padding: 14px; border-radius: 10px; overflow-x: auto; }
-      .endpoint { padding: 14px; border: 1px solid #eee; border-radius: 12px; margin: 14px 0; }
-      .method { display: inline-block; padding: 2px 8px; border-radius: 999px; font-size: 12px; font-weight: 700; }
-      .get { background: #e8f5e9; }
-      .post { background: #e3f2fd; }
-      .path { font-weight: 800; margin-left: 8px; }
-      .note { background: #fff8e1; border: 1px solid #ffe0b2; padding: 10px 12px; border-radius: 10px; margin-top: 10px; }
-    </style>
-  </head>
-  <body>
-    <header>
-      <h1>Getaround — Pricing Optimization API</h1>
-      <p>This API exposes a pricing model that predicts <code>rental_price_per_day</code> for a car based on its characteristics.</p>
-    </header>
-    <main>
-      <div class="endpoint">
-        <div><span class="method get">GET</span><span class="path">/health</span></div>
-        <p>Healthcheck endpoint.</p>
-        <pre><code>curl -s http://127.0.0.1:8000/health
-{"status":"ok"}</code></pre>
-      </div>
-      <div class="endpoint">
-        <div><span class="method post">POST</span><span class="path">/predict</span></div>
-        <p>Returns the predicted daily rental price for one or many cars.</p>
-        <p><strong>Accepted input format</strong> (JSON body):</p>
-        <ul>
-          <li><strong>List of dicts (recommended)</strong>: each dict contains the car fields.</li>
-        </ul>
-        <div class="note">
-          <strong>Note:</strong> If you want to support <code>list-of-lists</code> inputs, you must keep a strict feature order and validate the length.
-        </div>
-        <p><strong>Example (dict format):</strong></p>
-        <pre><code>{
   "input": [
     {
       "model_key": "Citroën",
@@ -69,43 +128,27 @@
   ]
 }</code></pre>
-        <p><strong>Example response:</strong></p>
-        <pre><code>{"prediction":[106.2]}</code></pre>
-        <p><strong>Example (curl):</strong></p>
-        <pre><code>curl -i -H "Content-Type: application/json" -X POST \
--d '{"input":[{"model_key":"Citroën","mileage":140411,"engine_power":100,"fuel":"diesel","paint_color":"black","car_type":"convertible","private_parking_available":true,"has_gps":true,"has_air_conditioning":false,"automatic_car":false,"has_getaround_connect":true,"has_speed_regulator":true,"winter_tires":true}]}' \
-http://127.0.0.1:8000/predict</code></pre>
-        <p><strong>Example (python):</strong></p>
-        <pre><code>import requests
-payload = {
-  "input": [{
-    "model_key": "Citroën",
-    "mileage": 140411,
-    "engine_power": 100,
-    "fuel": "diesel",
-    "paint_color": "black",
-    "car_type": "convertible",
-    "private_parking_available": True,
-    "has_gps": True,
-    "has_air_conditioning": False,
-    "automatic_car": False,
-    "has_getaround_connect": True,
-    "has_speed_regulator": True,
-    "winter_tires": True
-  }]
-}
-r = requests.post("http://127.0.0.1:8000/predict", json=payload)
-print(r.json())</code></pre>
-      </div>
-      <div class="endpoint">
-        <div><span class="method get">GET</span><span class="path">/docs</span></div>
-        <p>Returns this documentation page.</p>
-      </div>
-    </main>
-  </body>
 </html>

 <!doctype html>
 <html lang="en">
+<head>
+  <meta charset="utf-8" />
+  <meta name="viewport" content="width=device-width,initial-scale=1" />
+  <title>Getaround — Pricing Optimization API</title>
+  <style>
+    body { font-family: Arial, sans-serif; margin: 40px; max-width: 900px; }
+    h1, h2, h3 { color: #222; }
+    code { background: #f4f4f4; padding: 2px 4px; border-radius: 4px; }
+    pre { background: #f4f4f4; padding: 12px; border-radius: 6px; overflow-x: auto; }
+    .note { background: #eef6ff; padding: 12px; border-left: 4px solid #3b82f6; margin: 20px 0; }
+  </style>
+</head>
+<body>
+<h1>Getaround — Pricing Optimization API</h1>
+<p>
+This API exposes a machine learning model that predicts
+<strong>rental_price_per_day</strong> for a car based on its characteristics.
+</p>
+<hr>
+<h2>Health Check</h2>
+<p><strong>GET /health</strong></p>
+<pre><code>curl -s http://127.0.0.1:8000/health</code></pre>
+<p>Response:</p>
+<pre><code>{"status":"ok"}</code></pre>
+<hr>
+<h2>Prediction Endpoint</h2>
+<p><strong>POST /predict</strong></p>
+<p>
+Returns predicted daily rental prices for one or multiple cars.
+</p>
+<h3>Accepted Input Format (JSON body)</h3>
+<ul>
+  <li>
+    <strong>Matrix format (required by evaluation)</strong>:
+    <code>{"input": [[...], [...]]}</code>
+  </li>
+  <li>
+    <strong>Dict format (optional / backward compatible)</strong>:
+    <code>{"input": [{"feature": value, ...}]}</code>
+  </li>
+</ul>
+<div class="note">
+  <strong>Matrix format requirements:</strong><br><br>
+  Each row must contain <strong>exactly 13 values</strong> in the following strict order:
+  <br><br>
+  <code>
+    model_key, mileage, engine_power, fuel, paint_color, car_type,
+    private_parking_available, has_gps, has_air_conditioning, automatic_car,
+    has_getaround_connect, has_speed_regulator, winter_tires
+  </code>
+  <br><br>
+  Types:
+  <ul>
+    <li><strong>model_key, fuel, paint_color, car_type</strong> → strings</li>
+    <li><strong>mileage, engine_power</strong> → numeric values</li>
+    <li>All remaining features → 0 or 1</li>
+  </ul>
+</div>
+<hr>
+<h3>Example — Matrix Format (Required)</h3>
+<pre><code>{
+  "input": [
+    ["Citroën", 140411, 100, "diesel", "black", "convertible", 1, 1, 0, 0, 1, 1, 1]
+  ]
+}</code></pre>
+<h4>curl</h4>
+<pre><code>curl -X POST "http://127.0.0.1:8000/predict" \
+-H "Content-Type: application/json" \
+-d '{"input":[["Citroën",140411,100,"diesel","black","convertible",1,1,0,0,1,1,1]]}'</code></pre>
+<h4>Python</h4>
+<pre><code>import requests
+payload = {
+  "input": [
+    ["Citroën", 140411, 100, "diesel", "black", "convertible", 1, 1, 0, 0, 1, 1, 1]
+  ]
+}
+r = requests.post("http://127.0.0.1:8000/predict", json=payload)
+print(r.json())</code></pre>
+<hr>
+<h3>Optional — Dict Format (Backward Compatible)</h3>
+<pre><code>{
   "input": [
     {
       "model_key": "Citroën",
   ]
 }</code></pre>
+<hr>
+<h3>Response Format</h3>
+<pre><code>{
+  "prediction": [97.15]
+}</code></pre>
+<p>
+If multiple rows are provided, the response will contain one prediction per row.
+</p>
+<hr>
+<h2>Notes</h2>
+<ul>
+  <li>Unknown categorical values are handled automatically by the model.</li>
+  <li>Invalid matrix shape (wrong number of values) will return a 422 error.</li>
+  <li>The API is designed for batch predictions.</li>
+</ul>
+</body>
 </html>

main.py CHANGED Viewed

@@ -1,17 +1,19 @@
 from fastapi import FastAPI
 from fastapi.responses import HTMLResponse, JSONResponse
-from pydantic import BaseModel
-from typing import List, Any, Dict, Union
-import pandas as pd
-import joblib
-from pathlib import Path
 app = FastAPI(
     title="Getaround API",
     version="1.0.0",
     docs_url=None,
-    redoc_url=None
 )
 @app.get("/", response_class=HTMLResponse)
@@ -25,56 +27,77 @@ def home():
 APP_DIR = Path(__file__).resolve().parent
 MODEL_PATH = APP_DIR / "pricing_model.joblib"
-FEATURE_ORDER = [
-    "model_key",
-    "mileage",
-    "engine_power",
-    "fuel",
-    "paint_color",
-    "car_type",
-    "private_parking_available",
-    "has_gps",
-    "has_air_conditioning",
-    "automatic_car",
-    "has_getaround_connect",
-    "has_speed_regulator",
-    "winter_tires",
-]
 model = joblib.load(MODEL_PATH)
-class PredictIn(BaseModel):
-    input: List[Any]  # accepts list[dict] OR list[list] (see docs)
 @app.get("/health")
 def health():
     return {"status": "ok"}
-@app.post("/predict")
-def predict(payload: PredictIn):
     rows = payload.input
-    # Accept either:
-    # 1) list[dict] with keys matching FEATURE_ORDER
-    # 2) list[list] with values in FEATURE_ORDER order
     if len(rows) == 0:
-        return JSONResponse({"prediction": []})
     first = rows[0]
-    if isinstance(first, dict):
-        X = pd.DataFrame(rows)
-        # Ensure column order & missing columns are handled
-        for c in FEATURE_ORDER:
-            if c not in X.columns:
-                X[c] = None
-        X = X[FEATURE_ORDER]
-    else:
-        # list-like
-        X = pd.DataFrame(rows, columns=FEATURE_ORDER)
     preds = model.predict(X)
-    return {"prediction": [float(p) for p in preds]}
 BASE_DIR = Path(__file__).resolve().parent
@@ -82,4 +105,4 @@ BASE_DIR = Path(__file__).resolve().parent
 @app.get("/docs", response_class=HTMLResponse)
 def docs():
     html = (BASE_DIR / "docs.html").read_text(encoding="utf-8")
-    return HTMLResponse(content=html)

+import json
+from pathlib import Path
+from typing import Any, Dict, List, Union
+import joblib
+import pandas as pd
 from fastapi import FastAPI
 from fastapi.responses import HTMLResponse, JSONResponse
+from pydantic import BaseModel, Field, ValidationError
 app = FastAPI(
     title="Getaround API",
     version="1.0.0",
     docs_url=None,
+    redoc_url=None,
 )
 @app.get("/", response_class=HTMLResponse)
 APP_DIR = Path(__file__).resolve().parent
 MODEL_PATH = APP_DIR / "pricing_model.joblib"
+FEATURE_ORDER_PATH = APP_DIR / "feature_order.json"
 model = joblib.load(MODEL_PATH)
+FEATURE_ORDER: List[str] = json.loads(FEATURE_ORDER_PATH.read_text(encoding="utf-8"))
+class PredictMatrix(BaseModel):
+    input: List[List[Any]] = Field(
+        ...,
+        description="List of rows; each row is a list of feature values in the fixed FEATURE_ORDER.",
+    )
+class PredictDicts(BaseModel):
+    input: List[Dict[str, Any]] = Field(
+        ...,
+        description="List of objects with feature names as keys (backward compatibility).",
+    )
+PredictPayload = Union[PredictMatrix, PredictDicts]
 @app.get("/health")
 def health():
     return {"status": "ok"}
+def _payload_to_df(payload: PredictPayload) -> pd.DataFrame:
     rows = payload.input
     if len(rows) == 0:
+        return pd.DataFrame(columns=FEATURE_ORDER)
     first = rows[0]
+    if isinstance(first, list):
+        return pd.DataFrame(rows, columns=FEATURE_ORDER)
+    X = pd.DataFrame(rows)
+    missing = [c for c in FEATURE_ORDER if c not in X.columns]
+    if missing:
+        raise ValueError(f"Missing columns: {missing}")
+    return X[FEATURE_ORDER]
+@app.post("/predict")
+def predict(payload: ...):
+    N = len(FEATURE_ORDER)
+    rows = payload.input
+    if rows and isinstance(rows[0], list):
+        bad = [i for i, row in enumerate(rows) if len(row) != N]
+        if bad:
+            got = [len(rows[i]) for i in bad[:5]]
+            raise HTTPException(
+                status_code=422,
+                detail=(
+                    f"Invalid input shape. Each row must have exactly {N} values "
+                    f"(FEATURE_ORDER). Bad rows indices: {bad[:10]}. "
+                    f"Example bad lengths: {got}. "
+                    f"Expected order: {FEATURE_ORDER}"
+                ),
+            )
+    X = payload_to_df(payload)
     preds = model.predict(X)
+    return {"prediction": preds.tolist()}
 BASE_DIR = Path(__file__).resolve().parent
 @app.get("/docs", response_class=HTMLResponse)
 def docs():
     html = (BASE_DIR / "docs.html").read_text(encoding="utf-8")
+    return HTMLResponse(content=html)

requirements.txt CHANGED Viewed

@@ -6,3 +6,4 @@ scikit-learn==1.4.2
 joblib==1.4.2
 streamlit==1.36.0
 openpyxl==3.1.5

 joblib==1.4.2
 streamlit==1.36.0
 openpyxl==3.1.5
+mlflow==2.7.0