Upload 2 files

app.py

@@ -0,0 +1,86 @@
+# -*- coding: utf-8 -*-
+import os, json, pickle
+from typing import List, Dict, Any
+import numpy as np
+from fastapi import FastAPI, HTTPException
+from pydantic import BaseModel
+from sentence_transformers import SentenceTransformer
+from pydantic import BaseModel, ConfigDict
+
+# ---- load artifacts once ----
+DATA_DIR = os.path.join(os.path.dirname(__file__), "data")
+PRED_PATH = os.path.join(DATA_DIR, "predictor.pkl")
+if not os.path.exists(PRED_PATH):
+    raise FileNotFoundError("Put your predictor.pkl under data/")
+
+with open(PRED_PATH, "rb") as f:
+    lw = pickle.load(f)
+
+model_name: str = lw["model_name"]
+sbert = SentenceTransformer(model_name)
+
+centroides = {int(k): np.array(v, dtype=np.float32) for k, v in lw["centroides"].items()}
+# normalize centroid vectors (to match normalized embeddings)
+for k in list(centroides.keys()):
+    c = centroides[k]
+    n = float(np.linalg.norm(c) + 1e-12)
+    centroides[k] = c / n
+
+meta: Dict[int, Dict[str, Any]] = lw["meta"]
+cids = sorted(centroides.keys())
+
+
+
+class MailItem(BaseModel):
+    subject: str
+    body: str
+    # allow unknown fields
+    model_config = ConfigDict(extra="allow")
+
+class PredictRequest(BaseModel):
+    data: List[MailItem]
+
+class PredictResponseItem(BaseModel):
+    json: Dict[str, Any]
+
+class PredictResponse(BaseModel):
+    results: List[PredictResponseItem]
+
+# ---- core ----
+def _encode(texts: List[str]) -> np.ndarray:
+    # normalized embeddings
+    emb = sbert.encode(texts, convert_to_numpy=True, normalize_embeddings=True)
+    return emb.astype(np.float32)
+
+def _assign(vec: np.ndarray) -> int:
+    dists = [np.linalg.norm(vec - centroides[c]) for c in cids]
+    return cids[int(np.argmin(dists))]
+
+def _predict(records: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
+    texts = [f"{r.get('subject','')} — {r.get('body','')}" for r in records]
+    emb = _encode(texts)
+    out = []
+    for r, v in zip(records, emb):
+        cid = _assign(v)
+        j = dict(r)
+        j["cluster"] = cid
+        j["cluster_nombre"] = meta.get(cid, {}).get("nombre", f"cluster_{cid}")
+        j["cluster_desc"] = meta.get(cid, {}).get("descripcion", "")
+        out.append({"json": j})
+    return out
+
+# ---- FastAPI app ----
+app = FastAPI(title="Mail Cluster Inference", version="1.0.0")
+
+@app.get("/healthz")
+def healthz():
+    return {"ok": True, "clusters": len(cids), "model": model_name}
+
+@app.post("/predict", response_model=PredictResponse)
+def predict(req: PredictRequest):
+    try:
+        records = [m.dict() for m in req.data]
+        results = _predict(records)
+        return {"results": results}
+    except Exception as e:
+        raise HTTPException(status_code=400, detail=f"prediction error: {e}")

requirements.txt

@@ -0,0 +1,6 @@
+fastapi>=0.115
+uvicorn>=0.30
+pydantic>=2.7
+numpy>=2.0.0
+sentence-transformers>=3.0.1
+torch>=2.3