Spaces:
Running
Running
Commit ·
6ab2222
1
Parent(s): 2718be8
OR implementation 1
Browse files- __pycache__/app.cpython-313.pyc +0 -0
- app.py +301 -2
- requirements.txt +3 -1
- test_api.py +28 -0
__pycache__/app.cpython-313.pyc
ADDED
|
Binary file (43.7 kB). View file
|
|
|
app.py
CHANGED
|
@@ -12,6 +12,9 @@ import sqlite3
|
|
| 12 |
import re
|
| 13 |
from langdetect import detect, DetectorFactory
|
| 14 |
from urllib.parse import urlparse, urlunparse
|
|
|
|
|
|
|
|
|
|
| 15 |
|
| 16 |
# Make langdetect deterministic
|
| 17 |
DetectorFactory.seed = 0
|
|
@@ -32,6 +35,16 @@ model = None
|
|
| 32 |
translator_tokenizer = None
|
| 33 |
translator_model = None
|
| 34 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 35 |
|
| 36 |
# -----------------------
|
| 37 |
# DB + URL helpers
|
|
@@ -355,6 +368,117 @@ def predict_phishing(text: str):
|
|
| 355 |
logger.error(f"Error predicting phishing: {e}")
|
| 356 |
return "Error", 0.0
|
| 357 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 358 |
# -----------------------
|
| 359 |
# Unified pipeline
|
| 360 |
# -----------------------
|
|
@@ -487,7 +611,18 @@ def home():
|
|
| 487 |
|
| 488 |
@app.route("/health", methods=["GET"])
|
| 489 |
def health():
|
| 490 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 491 |
|
| 492 |
|
| 493 |
|
|
@@ -508,6 +643,158 @@ def analyze():
|
|
| 508 |
return jsonify({"error": "Internal server error"}), 500
|
| 509 |
|
| 510 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 511 |
# =============================
|
| 512 |
# /evaluate (GET form + POST CSV)
|
| 513 |
# =============================
|
|
@@ -698,4 +985,16 @@ def internal_error(error):
|
|
| 698 |
# -----------------------
|
| 699 |
# Startup
|
| 700 |
# -----------------------
|
| 701 |
-
load_model()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 12 |
import re
|
| 13 |
from langdetect import detect, DetectorFactory
|
| 14 |
from urllib.parse import urlparse, urlunparse
|
| 15 |
+
import time
|
| 16 |
+
from openrouter import OpenRouter
|
| 17 |
+
import requests
|
| 18 |
|
| 19 |
# Make langdetect deterministic
|
| 20 |
DetectorFactory.seed = 0
|
|
|
|
| 35 |
translator_tokenizer = None
|
| 36 |
translator_model = None
|
| 37 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
| 38 |
+
# OpenRouter client + model ids (env-configurable placeholders)
|
| 39 |
+
or_client = None
|
| 40 |
+
OR_MODEL_1 = os.environ.get("OR_MODEL_1", "or-model-1")
|
| 41 |
+
OR_MODEL_2 = os.environ.get("OR_MODEL_2", "or-model-2")
|
| 42 |
+
OR_TIMEOUT = int(os.environ.get("OR_TIMEOUT", "15"))
|
| 43 |
+
OR_CONSECUTIVE_FAILURES = 0
|
| 44 |
+
OR_CB_THRESHOLD = int(os.environ.get("OR_CB_THRESHOLD", "5"))
|
| 45 |
+
OR_CB_OPEN_SECONDS = int(os.environ.get("OR_CB_OPEN_SECONDS", "300"))
|
| 46 |
+
OR_LAST_ERROR = None
|
| 47 |
+
OR_CB_OPEN_UNTIL = 0
|
| 48 |
|
| 49 |
# -----------------------
|
| 50 |
# DB + URL helpers
|
|
|
|
| 368 |
logger.error(f"Error predicting phishing: {e}")
|
| 369 |
return "Error", 0.0
|
| 370 |
|
| 371 |
+
|
| 372 |
+
def generate_text(prompt: str, model_id: str, timeout: int = OR_TIMEOUT):
|
| 373 |
+
"""Send prompt to OpenRouter HTTP API and return dict {ok, text, error, status}.
|
| 374 |
+
Uses chat completions endpoint and returns the assistant message content.
|
| 375 |
+
Falls back gracefully on network or parsing errors.
|
| 376 |
+
"""
|
| 377 |
+
global OR_CONSECUTIVE_FAILURES, OR_LAST_ERROR, OR_CB_OPEN_UNTIL
|
| 378 |
+
api_key = os.environ.get("OPENROUTER_API_KEY")
|
| 379 |
+
# circuit breaker: short-circuit if open
|
| 380 |
+
if OR_CB_OPEN_UNTIL and time.time() < OR_CB_OPEN_UNTIL:
|
| 381 |
+
return {"ok": False, "text": None, "error": "circuit_open", "status": None}
|
| 382 |
+
if not api_key:
|
| 383 |
+
OR_CONSECUTIVE_FAILURES += 1
|
| 384 |
+
OR_LAST_ERROR = "Missing OPENROUTER_API_KEY"
|
| 385 |
+
return {"ok": False, "text": None, "error": "Missing OPENROUTER_API_KEY", "status": None}
|
| 386 |
+
url = "https://api.openrouter.ai/v1/chat/completions"
|
| 387 |
+
headers = {
|
| 388 |
+
"Authorization": f"Bearer {api_key}",
|
| 389 |
+
"Content-Type": "application/json"
|
| 390 |
+
}
|
| 391 |
+
body = {
|
| 392 |
+
"model": model_id,
|
| 393 |
+
"messages": [
|
| 394 |
+
{"role": "system", "content": "You are an automated phishing detector. Respond with only valid JSON."},
|
| 395 |
+
{"role": "user", "content": prompt}
|
| 396 |
+
],
|
| 397 |
+
"temperature": 0.0,
|
| 398 |
+
"max_tokens": 256
|
| 399 |
+
}
|
| 400 |
+
try:
|
| 401 |
+
resp = requests.post(url, headers=headers, json=body, timeout=timeout)
|
| 402 |
+
text = resp.text
|
| 403 |
+
if resp.status_code != 200:
|
| 404 |
+
logger.error(f"OpenRouter API error {resp.status_code}: {text}")
|
| 405 |
+
OR_CONSECUTIVE_FAILURES += 1
|
| 406 |
+
OR_LAST_ERROR = f"status {resp.status_code}"
|
| 407 |
+
if OR_CONSECUTIVE_FAILURES >= OR_CB_THRESHOLD:
|
| 408 |
+
OR_CB_OPEN_UNTIL = time.time() + OR_CB_OPEN_SECONDS
|
| 409 |
+
logger.warning("OpenRouter circuit opened due to repeated failures")
|
| 410 |
+
return {"ok": False, "text": text, "error": f"status {resp.status_code}", "status": resp.status_code}
|
| 411 |
+
# try to extract assistant content
|
| 412 |
+
j = resp.json()
|
| 413 |
+
content = None
|
| 414 |
+
if isinstance(j, dict) and "choices" in j and isinstance(j["choices"], list) and j["choices"]:
|
| 415 |
+
first = j["choices"][0]
|
| 416 |
+
# chat-style: message.content
|
| 417 |
+
if "message" in first and isinstance(first["message"], dict):
|
| 418 |
+
content = first["message"].get("content")
|
| 419 |
+
# or text field
|
| 420 |
+
if not content:
|
| 421 |
+
content = first.get("text")
|
| 422 |
+
# success -> reset failure counter
|
| 423 |
+
OR_CONSECUTIVE_FAILURES = 0
|
| 424 |
+
OR_LAST_ERROR = None
|
| 425 |
+
return {"ok": True, "text": content or text, "error": None, "status": resp.status_code}
|
| 426 |
+
except Exception as e:
|
| 427 |
+
logger.error(f"OpenRouter request error: {e}")
|
| 428 |
+
OR_CONSECUTIVE_FAILURES += 1
|
| 429 |
+
OR_LAST_ERROR = str(e)
|
| 430 |
+
if OR_CONSECUTIVE_FAILURES >= OR_CB_THRESHOLD:
|
| 431 |
+
OR_CB_OPEN_UNTIL = time.time() + OR_CB_OPEN_SECONDS
|
| 432 |
+
logger.warning("OpenRouter circuit opened due to repeated exceptions")
|
| 433 |
+
return {"ok": False, "text": None, "error": str(e), "status": None}
|
| 434 |
+
|
| 435 |
+
|
| 436 |
+
def predict_phishing_or(text: str, model_id: str):
|
| 437 |
+
"""Call OpenRouter to classify text. Return (label, confidence_percent, raw_response).
|
| 438 |
+
Falls back to HF `predict_phishing` on error or parse failure.
|
| 439 |
+
"""
|
| 440 |
+
# craft a strict prompt that asks for JSON
|
| 441 |
+
prompt = (
|
| 442 |
+
"Classify the following text as either 'Phishing' or 'Safe'.\n"
|
| 443 |
+
"Return ONLY a single JSON object with exactly these fields: {\"label\": \"Phishing\"|\"Safe\", \"confidence\": <0-100 float>}\n"
|
| 444 |
+
"Do not add any other text or explanation.\n\n"
|
| 445 |
+
f"TEXT:\n{text}"
|
| 446 |
+
)
|
| 447 |
+
resp = generate_text(prompt, model_id)
|
| 448 |
+
logger.info(f"OpenRouter raw response: {resp}")
|
| 449 |
+
if not resp.get("ok") or not resp.get("text"):
|
| 450 |
+
# fallback to HF
|
| 451 |
+
label, conf = predict_phishing(text)
|
| 452 |
+
return label, conf, {"fallback": True, "reason": resp.get("error"), "raw": resp.get("text")}
|
| 453 |
+
|
| 454 |
+
# try to parse JSON object from resp['text']
|
| 455 |
+
txt = resp.get("text")
|
| 456 |
+
# extract the first {...} block
|
| 457 |
+
import json, re
|
| 458 |
+
m = re.search(r"\{.*\}", txt, re.DOTALL)
|
| 459 |
+
if not m:
|
| 460 |
+
label, conf = predict_phishing(text)
|
| 461 |
+
return label, conf, {"fallback": True, "reason": "no_json", "raw": txt}
|
| 462 |
+
try:
|
| 463 |
+
obj = json.loads(m.group(0))
|
| 464 |
+
lab = obj.get("label")
|
| 465 |
+
conf = obj.get("confidence")
|
| 466 |
+
if isinstance(conf, str):
|
| 467 |
+
try:
|
| 468 |
+
conf = float(conf)
|
| 469 |
+
except Exception:
|
| 470 |
+
conf = 0.0
|
| 471 |
+
if lab and (lab.lower() in ("phishing", "safe")):
|
| 472 |
+
label = "Phishing" if lab.lower() == "phishing" else "Safe"
|
| 473 |
+
conf_val = round(float(conf), 1) if conf is not None else 0.0
|
| 474 |
+
return label, conf_val, {"fallback": False, "raw": txt}
|
| 475 |
+
else:
|
| 476 |
+
raise ValueError("Invalid label")
|
| 477 |
+
except Exception as e:
|
| 478 |
+
logger.error(f"Error parsing OpenRouter JSON: {e} -- raw: {txt}")
|
| 479 |
+
label, conf = predict_phishing(text)
|
| 480 |
+
return label, conf, {"fallback": True, "reason": str(e), "raw": txt}
|
| 481 |
+
|
| 482 |
# -----------------------
|
| 483 |
# Unified pipeline
|
| 484 |
# -----------------------
|
|
|
|
| 611 |
|
| 612 |
@app.route("/health", methods=["GET"])
|
| 613 |
def health():
|
| 614 |
+
or_configured = bool(os.environ.get("OPENROUTER_API_KEY"))
|
| 615 |
+
return jsonify({
|
| 616 |
+
"status": "healthy",
|
| 617 |
+
"model_loaded": model is not None,
|
| 618 |
+
"openrouter": {
|
| 619 |
+
"configured": or_configured,
|
| 620 |
+
"client_initialized": or_client is not None,
|
| 621 |
+
"consecutive_failures": OR_CONSECUTIVE_FAILURES,
|
| 622 |
+
"circuit_open": bool(OR_CB_OPEN_UNTIL and time.time() < OR_CB_OPEN_UNTIL),
|
| 623 |
+
"last_error": OR_LAST_ERROR
|
| 624 |
+
}
|
| 625 |
+
})
|
| 626 |
|
| 627 |
|
| 628 |
|
|
|
|
| 643 |
return jsonify({"error": "Internal server error"}), 500
|
| 644 |
|
| 645 |
|
| 646 |
+
def analyze_pipeline_with_model(message: str, model_id: str):
|
| 647 |
+
"""Same pipeline as analyze_pipeline but uses the specified OpenRouter model for classification.
|
| 648 |
+
Falls back to HF classifier when OR fails.
|
| 649 |
+
"""
|
| 650 |
+
try:
|
| 651 |
+
text = (message or "").strip()
|
| 652 |
+
if not text:
|
| 653 |
+
return {"error": "empty", "blacklist": False}
|
| 654 |
+
|
| 655 |
+
# extract both full URLs and domains once
|
| 656 |
+
full_urls, domains = extract_urls_and_domains(text)
|
| 657 |
+
|
| 658 |
+
# --- load whitelist set (lowercased) ---
|
| 659 |
+
try:
|
| 660 |
+
conn = get_db_connection()
|
| 661 |
+
cursor = conn.cursor()
|
| 662 |
+
cursor.execute("SELECT LOWER(domain) FROM whitelist")
|
| 663 |
+
rows = cursor.fetchall()
|
| 664 |
+
whitelist_set = {r[0] for r in rows if r and r[0]}
|
| 665 |
+
conn.close()
|
| 666 |
+
except Exception as e:
|
| 667 |
+
logger.warning(f"Whitelist load failed: {e}")
|
| 668 |
+
whitelist_set = set()
|
| 669 |
+
|
| 670 |
+
# 1) domain whitelist check (allow subdomains)
|
| 671 |
+
if domains and whitelist_set:
|
| 672 |
+
for domain in domains:
|
| 673 |
+
d = (domain or "").lower().strip().strip('/')
|
| 674 |
+
if not d:
|
| 675 |
+
continue
|
| 676 |
+
# direct or subdomain match
|
| 677 |
+
matched_whitelist = False
|
| 678 |
+
if d in whitelist_set:
|
| 679 |
+
matched_whitelist = True
|
| 680 |
+
else:
|
| 681 |
+
for wl in whitelist_set:
|
| 682 |
+
if d == wl or d.endswith("." + wl):
|
| 683 |
+
matched_whitelist = True
|
| 684 |
+
break
|
| 685 |
+
|
| 686 |
+
if matched_whitelist:
|
| 687 |
+
# BEFORE returning Safe, check whether any full URL in the message
|
| 688 |
+
# is exactly present in the blacklist (normalized). If so, treat as Phishing.
|
| 689 |
+
if full_urls:
|
| 690 |
+
try:
|
| 691 |
+
bconn = get_blacklist_connection()
|
| 692 |
+
bcur = bconn.cursor()
|
| 693 |
+
for u in full_urls:
|
| 694 |
+
norm_u = normalize_full_url(u).lower()
|
| 695 |
+
bcur.execute("SELECT 1 FROM blacklist WHERE LOWER(url) = ?", (norm_u,))
|
| 696 |
+
if bcur.fetchone():
|
| 697 |
+
bconn.close()
|
| 698 |
+
return {
|
| 699 |
+
"result": "Phishing",
|
| 700 |
+
"confidence": "100.0%",
|
| 701 |
+
"message": text,
|
| 702 |
+
"blacklist": True,
|
| 703 |
+
"whitelist": False,
|
| 704 |
+
"detected_lang": None,
|
| 705 |
+
"translated_text": None,
|
| 706 |
+
"model_used": model_id
|
| 707 |
+
}
|
| 708 |
+
bconn.close()
|
| 709 |
+
except Exception as e:
|
| 710 |
+
logger.warning(f"Blacklist-full-url check failed: {e}")
|
| 711 |
+
# If no full-url blacklist hit, return Safe because domain is whitelisted
|
| 712 |
+
return {
|
| 713 |
+
"result": "Safe",
|
| 714 |
+
"confidence": "100.0%",
|
| 715 |
+
"message": text,
|
| 716 |
+
"blacklist": False,
|
| 717 |
+
"whitelist": True,
|
| 718 |
+
"detected_lang": None,
|
| 719 |
+
"translated_text": None,
|
| 720 |
+
"model_used": model_id
|
| 721 |
+
}
|
| 722 |
+
|
| 723 |
+
# 2) not whitelisted by domain (or whitelist didn't apply) -> regular blacklist check
|
| 724 |
+
is_black, _reason = is_blacklisted(text)
|
| 725 |
+
if is_black:
|
| 726 |
+
return {
|
| 727 |
+
"result": "Phishing",
|
| 728 |
+
"confidence": "100.0%",
|
| 729 |
+
"message": text,
|
| 730 |
+
"blacklist": True,
|
| 731 |
+
"whitelist": False,
|
| 732 |
+
"detected_lang": None,
|
| 733 |
+
"translated_text": None,
|
| 734 |
+
"model_used": model_id
|
| 735 |
+
}
|
| 736 |
+
|
| 737 |
+
# 3) language detection + optional translate
|
| 738 |
+
lang = detect_language(text)
|
| 739 |
+
translated_text = None
|
| 740 |
+
analysis_text = text
|
| 741 |
+
if lang in ("tl", "fil"):
|
| 742 |
+
translated_text = translate_tl_to_en(text)
|
| 743 |
+
analysis_text = translated_text or text
|
| 744 |
+
|
| 745 |
+
# 4) classification via OpenRouter model (with HF fallback inside)
|
| 746 |
+
label, conf, raw = predict_phishing_or(analysis_text, model_id)
|
| 747 |
+
return {
|
| 748 |
+
"result": label,
|
| 749 |
+
"confidence": f"{conf}%",
|
| 750 |
+
"message": text,
|
| 751 |
+
"blacklist": False,
|
| 752 |
+
"whitelist": False,
|
| 753 |
+
"detected_lang": lang,
|
| 754 |
+
"translated_text": translated_text,
|
| 755 |
+
"model_used": model_id,
|
| 756 |
+
"or_raw": raw
|
| 757 |
+
}
|
| 758 |
+
|
| 759 |
+
except Exception as e:
|
| 760 |
+
logger.error(f"Pipeline error: {e}")
|
| 761 |
+
return {"result": "Error", "confidence": "0%", "message": message, "blacklist": False, "whitelist": False}
|
| 762 |
+
|
| 763 |
+
|
| 764 |
+
@app.route("/analyze/or1", methods=["POST"])
|
| 765 |
+
def analyze_or1():
|
| 766 |
+
try:
|
| 767 |
+
data = request.get_json()
|
| 768 |
+
if not data or "message" not in data:
|
| 769 |
+
return jsonify({"error": "Missing 'message' field"}), 400
|
| 770 |
+
message = data["message"]
|
| 771 |
+
if not message or not message.strip():
|
| 772 |
+
return jsonify({"error": "Message cannot be empty"}), 400
|
| 773 |
+
|
| 774 |
+
result = analyze_pipeline_with_model(message, OR_MODEL_1)
|
| 775 |
+
return jsonify(result), 200
|
| 776 |
+
except Exception as e:
|
| 777 |
+
logger.error(f"Error in analyze/or1 endpoint: {e}")
|
| 778 |
+
return jsonify({"error": "Internal server error"}), 500
|
| 779 |
+
|
| 780 |
+
|
| 781 |
+
@app.route("/analyze/or2", methods=["POST"])
|
| 782 |
+
def analyze_or2():
|
| 783 |
+
try:
|
| 784 |
+
data = request.get_json()
|
| 785 |
+
if not data or "message" not in data:
|
| 786 |
+
return jsonify({"error": "Missing 'message' field"}), 400
|
| 787 |
+
message = data["message"]
|
| 788 |
+
if not message or not message.strip():
|
| 789 |
+
return jsonify({"error": "Message cannot be empty"}), 400
|
| 790 |
+
|
| 791 |
+
result = analyze_pipeline_with_model(message, OR_MODEL_2)
|
| 792 |
+
return jsonify(result), 200
|
| 793 |
+
except Exception as e:
|
| 794 |
+
logger.error(f"Error in analyze/or2 endpoint: {e}")
|
| 795 |
+
return jsonify({"error": "Internal server error"}), 500
|
| 796 |
+
|
| 797 |
+
|
| 798 |
# =============================
|
| 799 |
# /evaluate (GET form + POST CSV)
|
| 800 |
# =============================
|
|
|
|
| 985 |
# -----------------------
|
| 986 |
# Startup
|
| 987 |
# -----------------------
|
| 988 |
+
load_model()
|
| 989 |
+
try:
|
| 990 |
+
api_key = os.environ.get("OPENROUTER_API_KEY")
|
| 991 |
+
if api_key:
|
| 992 |
+
try:
|
| 993 |
+
or_client = OpenRouter(api_key=api_key)
|
| 994 |
+
logger.info("OpenRouter client initialized.")
|
| 995 |
+
except Exception as e:
|
| 996 |
+
logger.warning(f"Failed to initialize OpenRouter client object: {e}")
|
| 997 |
+
else:
|
| 998 |
+
logger.warning("OPENROUTER_API_KEY not set; OpenRouter endpoints will still attempt HTTP calls but client object is not created.")
|
| 999 |
+
except Exception as e:
|
| 1000 |
+
logger.warning(f"OpenRouter init warning: {e}")
|
requirements.txt
CHANGED
|
@@ -10,4 +10,6 @@ google-auth-oauthlib
|
|
| 10 |
google-auth-httplib2
|
| 11 |
langdetect
|
| 12 |
sentencepiece
|
| 13 |
-
sacremoses
|
|
|
|
|
|
|
|
|
| 10 |
google-auth-httplib2
|
| 11 |
langdetect
|
| 12 |
sentencepiece
|
| 13 |
+
sacremoses
|
| 14 |
+
openrouter
|
| 15 |
+
requests
|
test_api.py
ADDED
|
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os
|
| 2 |
+
import requests
|
| 3 |
+
import sys
|
| 4 |
+
import json
|
| 5 |
+
|
| 6 |
+
API_BASE = os.getenv("API_BASE", "http://localhost:8080")
|
| 7 |
+
MESSAGE = os.getenv("TEST_MESSAGE", "Please verify your account at http://example.com/login")
|
| 8 |
+
TIMEOUT = int(os.getenv("TEST_TIMEOUT", "10"))
|
| 9 |
+
|
| 10 |
+
endpoints = ["/analyze", "/analyze/or1", "/analyze/or2"]
|
| 11 |
+
|
| 12 |
+
print(f"Testing API at {API_BASE} with message: {MESSAGE}\n")
|
| 13 |
+
for ep in endpoints:
|
| 14 |
+
url = API_BASE.rstrip("/") + ep
|
| 15 |
+
payload = {"message": MESSAGE}
|
| 16 |
+
try:
|
| 17 |
+
resp = requests.post(url, json=payload, timeout=TIMEOUT)
|
| 18 |
+
try:
|
| 19 |
+
body = resp.json()
|
| 20 |
+
except Exception:
|
| 21 |
+
body = resp.text
|
| 22 |
+
print(f"Endpoint: {ep} -> Status: {resp.status_code}")
|
| 23 |
+
print(json.dumps(body, indent=2, ensure_ascii=False))
|
| 24 |
+
except requests.exceptions.RequestException as e:
|
| 25 |
+
print(f"Endpoint: {ep} -> Request failed: {e}")
|
| 26 |
+
print("---")
|
| 27 |
+
|
| 28 |
+
print("Done.")
|