Create rfp_api_app.py

rfp_api_app.py

@@ -0,0 +1,164 @@
+# rfp_api_app.py
+# -*- coding: utf-8 -*-
+from __future__ import annotations
+from typing import Dict, Any
+import os, json, uuid, threading, time, traceback
+from pathlib import Path
+import logging
+import requests
+
+from fastapi import FastAPI, HTTPException, Query
+from fastapi.responses import JSONResponse, FileResponse
+from fastapi.middleware.cors import CORSMiddleware
+
+# === Imports depuis ton repo cloné RFPmaster ===
+from rfp_parser.prompting import build_chat_payload
+from rfp_parser.exports_xls import build_xls_from_doc
+
+# --------- Config ---------
+DEEPINFRA_API_KEY = os.environ.get("DEEPINFRA_API_KEY", "")
+MODEL_NAME        = os.environ.get("RFP_MODEL", "meta-llama/Meta-Llama-3.1-70B-Instruct")
+DEEPINFRA_URL     = os.environ.get("DEEPINFRA_URL", "https://api.deepinfra.com/v1/openai/chat/completions")
+RFP_DEBUG         = str(os.environ.get("RFP_DEBUG", "0")).lower() in {"1", "true", "yes"}
+BASE_TMP          = Path("/tmp/rfp_jobs"); BASE_TMP.mkdir(parents=True, exist_ok=True)
+
+logger = logging.getLogger("RFP_API")
+if not logger.handlers:
+    h = logging.StreamHandler()
+    h.setFormatter(logging.Formatter("[API] %(levelname)s: %(message)s"))
+    logger.addHandler(h)
+logger.setLevel(logging.DEBUG if RFP_DEBUG else logging.INFO)
+
+# --------- Jobs en mémoire ---------
+JOBS: Dict[str, Dict[str, Any]] = {}
+JOBS_LOCK = threading.Lock()
+
+def new_job(text: str) -> str:
+    job_id = uuid.uuid4().hex[:12]
+    with JOBS_LOCK:
+        JOBS[job_id] = {
+            "status": "queued",
+            "error": None,
+            "xlsx_path": None,
+            "xlsx_url": None,
+            "started_at": time.time(),
+            "done_at": None,
+            "meta": {"model": MODEL_NAME, "length": len(text or "")},
+        }
+    return job_id
+
+def set_job_status(job_id: str, **updates):
+    with JOBS_LOCK:
+        if job_id in JOBS:
+            JOBS[job_id].update(**updates)
+
+# --------- Cœur pipeline ---------
+def parse_with_deepinfra(text: str) -> Dict[str, Any]:
+    if not DEEPINFRA_API_KEY:
+        raise RuntimeError("DEEPINFRA_API_KEY manquant (Settings → Secrets du Space).")
+    payload = build_chat_payload(text, model=MODEL_NAME)
+    headers = {"Authorization": f"Bearer {DEEPINFRA_API_KEY}", "Content-Type": "application/json"}
+    logger.info("DeepInfra call model=%s max_tokens=%s", payload.get("model"), payload.get("max_tokens"))
+    r = requests.post(DEEPINFRA_URL, headers=headers, json=payload, timeout=120)
+    if r.status_code // 100 != 2:
+        raise RuntimeError(f"DeepInfra HTTP {r.status_code}: {r.text}")
+    data = r.json()
+    try:
+        content = data["choices"][0]["message"]["content"]
+    except Exception:
+        raise RuntimeError(f"Réponse inattendue DeepInfra: {json.dumps(data)[:400]}")
+    try:
+        doc = json.loads(content)
+    except Exception as e:
+        logger.warning("json.loads(content) a échoué; strip fallback. Err=%s", e)
+        doc = json.loads(content.strip().strip('`').strip())
+    if not isinstance(doc, dict):
+        raise RuntimeError("Le contenu renvoyé n'est pas un objet JSON.")
+    return doc
+
+def build_xlsx(doc: Dict[str, Any], job_dir: Path) -> str:
+    job_dir.mkdir(parents=True, exist_ok=True)
+    out_path = str(job_dir / "feuille_de_charge.xlsx")
+    baseline = (doc.get("assumptions") or {}).get("baseline_uop_kg") or 100.0
+    try:
+        baseline = float(baseline)
+    except Exception:
+        baseline = 100.0
+    build_xls_from_doc(doc, out_path, baseline_kg=baseline)
+    return out_path
+
+def run_job(job_id: str, text: str) -> None:
+    set_job_status(job_id, status="running")
+    job_dir = BASE_TMP / job_id
+    try:
+        doc = parse_with_deepinfra(text)
+        xlsx_path = build_xlsx(doc, job_dir)
+        # Ici, comme on n’est PAS monté sous /api : le chemin est direct
+        xlsx_url = f"/results/{job_id}/feuille_de_charge.xlsx"
+        set_job_status(
+            job_id,
+            status="done",
+            xlsx_path=xlsx_path,
+            xlsx_url=xlsx_url,
+            done_at=time.time(),
+            meta={**JOBS[job_id]["meta"], "assumptions": doc.get("assumptions")},
+        )
+        logger.info("Job %s terminé -> %s", job_id, xlsx_path)
+    except Exception as e:
+        logger.error("Job %s échoué: %s\n%s", job_id, e, traceback.format_exc())
+        set_job_status(job_id, status="error", error=str(e), done_at=time.time())
+
+# --------- FastAPI ---------
+app = FastAPI(title="RFP_MASTER API", version="1.0.0")
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],   # serre si tu veux limiter à ton Space Gradio
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+@app.get("/health")
+def health():
+    return {"ok": True, "ts": time.time(), "model": MODEL_NAME}
+
+@app.post("/submit")
+def submit(payload: Dict[str, Any]):
+    text = (payload or {}).get("text", "")
+    if not isinstance(text, str) or not text.strip():
+        raise HTTPException(400, "Champ 'text' manquant ou vide.")
+    job_id = new_job(text)
+    t = threading.Thread(target=run_job, args=(job_id, text), daemon=True)
+    t.start()
+    return JSONResponse({"job_id": job_id, "status": "queued"})
+
+@app.get("/status")
+def status(job_id: str = Query(..., description="Identifiant renvoyé par /submit")):
+    with JOBS_LOCK:
+        info = JOBS.get(job_id)
+    if not info:
+        raise HTTPException(404, f"job_id inconnu: {job_id}")
+    return JSONResponse({
+        "job_id": job_id,
+        "status": info.get("status"),
+        "xlsx_url": info.get("xlsx_url"),
+        "error": info.get("error"),
+        "meta": info.get("meta"),
+    })
+
+@app.get("/results/{job_id}/feuille_de_charge.xlsx")
+def download(job_id: str):
+    with JOBS_LOCK:
+        info = JOBS.get(job_id)
+    if not info:
+        raise HTTPException(404, f"job_id inconnu: {job_id}")
+    if info.get("status") != "done":
+        raise HTTPException(409, f"job {job_id} non prêt (status={info.get('status')})")
+    xlsx_path = info.get("xlsx_path")
+    if not xlsx_path or not Path(xlsx_path).exists():
+        raise HTTPException(404, "Fichier indisponible.")
+    return FileResponse(
+        xlsx_path,
+        media_type="application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
+        filename="feuille_de_charge.xlsx",
+    )