File size: 2,657 Bytes
d379bc3
 
424a139
d379bc3
d5c5b11
d379bc3
 
7705b09
 
 
 
d379bc3
 
 
 
 
 
 
 
c2e0e06
 
424a139
 
 
 
 
d5c5b11
424a139
d5c5b11
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
424a139
d5c5b11
 
424a139
 
d379bc3
 
 
c2e0e06
d379bc3
c2e0e06
 
 
 
 
 
424a139
d379bc3
424a139
d379bc3
 
 
424a139
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
from fastapi import FastAPI
from controllers import chat_controller, test_controller, admin_controller, health_controller
from core import service_config, session_store, llm_models
from llm_model import LLMModel
from intent_utils import background_training
from log import log
import os
import warnings

# FutureWarning'leri sustur
warnings.simplefilter(action='ignore', category=FutureWarning)

app = FastAPI()

app.include_router(health_controller.router)
app.include_router(chat_controller.router)
app.include_router(test_controller.router)
app.include_router(admin_controller.router)

BASE_PROJECTS_DIR = "/data/projects"

def load_project(project_name, config, project_path):
    llm_config = config.get_project_llm_config(project_name)
    model_instance = LLMModel()
    model_instance.setup(config, llm_config, project_path)

    # Intent modeli path
    intent_model_path = os.path.join(project_path, "intent", "trained_model")

    # Eğer intent modeli klasörü yoksa → eğitim başlat
    if not os.path.exists(intent_model_path) or not os.path.isfile(os.path.join(intent_model_path, "config.json")):
        log(f"🛠 Intent modeli bulunamadı, eğitim başlatılıyor: {intent_model_path}")
        intents = config.get_project_intents(project_name)
        os.makedirs(intent_model_path, exist_ok=True)
        background_training(
            project_name,
            intents,
            llm_config["intent_model_id"],
            intent_model_path,
            llm_config["train_confidence_treshold"]
        )

    # Eğitim sonrası intent modelini yükle
    if os.path.exists(intent_model_path) and os.path.isfile(os.path.join(intent_model_path, "config.json")):
        model_instance.load_intent_model(intent_model_path)
    else:
        log(f"⚠️ Intent modeli yüklenemedi: {intent_model_path}, yükleme atlandı.")

    return model_instance

log("🌐 Servis başlatılıyor...")
service_config.load(is_reload=False)

for project_name in service_config.projects:
    project_path = os.path.join(BASE_PROJECTS_DIR, project_name)
    os.makedirs(project_path, exist_ok=True)
    os.makedirs(os.path.join(project_path, "llm", "base_model"), exist_ok=True)
    os.makedirs(os.path.join(project_path, "llm", "fine_tune"), exist_ok=True)
    os.makedirs(os.path.join(project_path, "intent", "trained_model"), exist_ok=True)

    model_instance = load_project(project_name, service_config, project_path)
    llm_models[project_name] = model_instance
    log(f"✅ '{project_name}' için tüm modeller yüklenip belleğe alındı.")

if __name__ == "__main__":
    import uvicorn
    uvicorn.run(app, host="0.0.0.0", port=7860)