244 lines
13 KiB
Python
244 lines
13 KiB
Python
import asyncio
|
|
import httpx
|
|
from openai import AsyncOpenAI
|
|
from fastapi import HTTPException
|
|
from sqlalchemy.orm import Session
|
|
from sqlalchemy.exc import IntegrityError
|
|
from .models import SystemSettings, AIModel, FinancialLog
|
|
from datetime import datetime, timezone
|
|
|
|
def determine_domain(model_id: str) -> str:
|
|
mid = model_id.lower()
|
|
if "vision" in mid or "vl" in mid or "omni" in mid: return "Vision & Texte"
|
|
if "coder" in mid or "code" in mid or "math" in mid: return "Code & Logique"
|
|
if "audio" in mid or "whisper" in mid: return "Audio"
|
|
return "Texte Polyvalent"
|
|
|
|
async def sync_finances(db: Session, settings: SystemSettings):
|
|
"""Interroge les fournisseurs pour récupérer le solde financier exact."""
|
|
async with httpx.AsyncClient(timeout=30.0) as client:
|
|
# OpenRouter
|
|
if settings.openrouter_management_key or settings.openrouter_api_key:
|
|
try:
|
|
key = settings.openrouter_management_key or settings.openrouter_api_key
|
|
cred_resp = await client.get("https://openrouter.ai/api/v1/credits", headers={"Authorization": f"Bearer {key}"})
|
|
balance = 0.0
|
|
usage = 0.0
|
|
if cred_resp.status_code == 200 and cred_resp.json().get("data"):
|
|
c_data = cred_resp.json().get("data", {})
|
|
balance = float(c_data.get("total_credits") or 0.0) - float(c_data.get("total_usage") or 0.0)
|
|
usage = float(c_data.get("total_usage") or 0.0)
|
|
else:
|
|
key_resp = await client.get("https://openrouter.ai/api/v1/auth/key", headers={"Authorization": f"Bearer {key}"})
|
|
if key_resp.status_code == 200:
|
|
data = key_resp.json().get("data", {})
|
|
limit = data.get("limit")
|
|
usage = float(data.get("usage") or 0.0)
|
|
balance = (float(limit) - usage) if limit is not None else -usage
|
|
update_finance_db(db, "OpenRouter", balance, usage)
|
|
except Exception as e: print(f"Erreur Finance OR: {e}")
|
|
|
|
# Groq
|
|
if settings.groq_api_key: update_finance_db(db, "Groq", 999.0, 0.0)
|
|
|
|
# DeepSeek
|
|
if settings.deepseek_api_key:
|
|
try:
|
|
resp = await client.get("https://api.deepseek.com/user/balance", headers={"Authorization": f"Bearer {settings.deepseek_api_key}"})
|
|
if resp.status_code == 200:
|
|
infos = resp.json().get("balance_infos", [{}])[0]
|
|
update_finance_db(db, "DeepSeek", float(infos.get("total_balance", 0)), 0.0)
|
|
except Exception: pass
|
|
|
|
# Mistral AI
|
|
if settings.mistral_api_key: update_finance_db(db, "Mistral", 0.0, 0.0)
|
|
|
|
# Gemini
|
|
if settings.gemini_api_key: update_finance_db(db, "Gemini", 0.0, 0.0)
|
|
|
|
try: db.commit()
|
|
except: db.rollback()
|
|
|
|
def update_finance_db(db, provider, balance, usage):
|
|
try:
|
|
log = db.query(FinancialLog).filter(FinancialLog.provider == provider).first()
|
|
if log:
|
|
log.balance = balance; log.total_usage = usage; log.checked_at = datetime.now(timezone.utc)
|
|
else:
|
|
db.add(FinancialLog(provider=provider, balance=balance, total_usage=usage))
|
|
except Exception as e: print(f"Finance DB Error: {e}")
|
|
|
|
async def sync_providers_models(db: Session, settings: SystemSettings, sync_type: str = "Automatique"):
|
|
added = 0
|
|
# On stocke les modèles en mémoire avant de les envoyer en base pour éviter les doublons
|
|
models_to_process = {}
|
|
|
|
async with httpx.AsyncClient(timeout=90.0) as client:
|
|
# 1. OpenRouter
|
|
if settings.openrouter_api_key:
|
|
try:
|
|
resp = await client.get("https://openrouter.ai/api/v1/models")
|
|
if resp.status_code == 200:
|
|
for item in resp.json().get("data", []):
|
|
try:
|
|
m_id = item.get("id")
|
|
if not m_id: continue
|
|
pricing = item.get("pricing") or {}
|
|
try: pp = float(pricing.get("prompt") or 0.0) * 1000000
|
|
except: pp = 0.0
|
|
try: pc = float(pricing.get("completion") or 0.0) * 1000000
|
|
except: pc = 0.0
|
|
is_free = (pp == 0.0 and pc == 0.0)
|
|
desc = item.get("description", "Modèle OpenRouter.")[:200] + "..."
|
|
models_to_process[m_id] = {"provider": "openrouter", "name": item.get("name", "Inconnu"), "desc": desc, "domain": determine_domain(m_id), "is_free": is_free, "ctx": item.get("context_length", 0), "pp": pp, "pc": pc}
|
|
except: pass
|
|
except Exception as e: print(f"Erreur OR Models: {e}")
|
|
|
|
# 2. Groq
|
|
if settings.groq_api_key:
|
|
try:
|
|
resp = await client.get("https://api.groq.com/openai/v1/models", headers={"Authorization": f"Bearer {settings.groq_api_key}"})
|
|
if resp.status_code == 200:
|
|
for item in resp.json().get("data", []):
|
|
try:
|
|
m_id = item["id"]
|
|
models_to_process[m_id] = {"provider": "groq", "name": m_id.capitalize(), "desc": "Modèle rapide LPU Groq.", "domain": determine_domain(m_id), "is_free": True, "ctx": 8192, "pp": 0.0, "pc": 0.0}
|
|
except: pass
|
|
except Exception: pass
|
|
|
|
# 3. DeepSeek
|
|
if settings.deepseek_api_key:
|
|
try:
|
|
resp = await client.get("https://api.deepseek.com/models", headers={"Authorization": f"Bearer {settings.deepseek_api_key}"})
|
|
if resp.status_code == 200:
|
|
for item in resp.json().get("data", []):
|
|
try:
|
|
m_id = item["id"]
|
|
models_to_process[m_id] = {"provider": "deepseek", "name": m_id.capitalize(), "desc": "Modèle officiel DeepSeek.", "domain": determine_domain(m_id), "is_free": False, "ctx": 64000, "pp": 0.14, "pc": 0.28}
|
|
except: pass
|
|
except Exception: pass
|
|
|
|
# 4. Mistral
|
|
if settings.mistral_api_key:
|
|
try:
|
|
resp = await client.get("https://api.mistral.ai/v1/models", headers={"Authorization": f"Bearer {settings.mistral_api_key}"})
|
|
if resp.status_code == 200:
|
|
for item in resp.json().get("data", []):
|
|
try:
|
|
m_id = item["id"]
|
|
models_to_process[m_id] = {"provider": "mistral", "name": m_id.capitalize(), "desc": "Modèle officiel Mistral AI.", "domain": determine_domain(m_id), "is_free": False, "ctx": 32000, "pp": 0.2, "pc": 0.6}
|
|
except: pass
|
|
except Exception: pass
|
|
|
|
# 5. Gemini (Google)
|
|
if settings.gemini_api_key:
|
|
try:
|
|
resp = await client.get(f"https://generativelanguage.googleapis.com/v1beta/models?key={settings.gemini_api_key}")
|
|
if resp.status_code == 200:
|
|
for item in resp.json().get("models", []):
|
|
try:
|
|
m_id = item["name"].replace("models/", "")
|
|
desc = item.get("description", "Modèle Google Gemini.")[:200] + "..."
|
|
models_to_process[m_id] = {"provider": "gemini", "name": item.get("displayName", m_id), "desc": desc, "domain": determine_domain(m_id), "is_free": True, "ctx": item.get("inputTokenLimit", 32000), "pp": 0.0, "pc": 0.0}
|
|
except: pass
|
|
except Exception: pass
|
|
|
|
# 6. Cloudflare
|
|
if settings.cloudflare_account_id and settings.cloudflare_api_token:
|
|
try:
|
|
url = f"https://api.cloudflare.com/client/v4/accounts/{settings.cloudflare_account_id}/ai/models/search"
|
|
resp = await client.get(url, headers={"Authorization": f"Bearer {settings.cloudflare_api_token}"})
|
|
if resp.status_code == 200:
|
|
for item in resp.json().get("result", []):
|
|
try:
|
|
m_id = item.get("name")
|
|
desc = item.get("description", "Modèle Cloudflare Workers AI.")[:200] + "..."
|
|
models_to_process[m_id] = {"provider": "cloudflare", "name": m_id.split("/")[-1], "desc": desc, "domain": determine_domain(m_id), "is_free": True, "ctx": 4096, "pp": 0.0, "pc": 0.0}
|
|
except: pass
|
|
except Exception: pass
|
|
|
|
# --- Phase d'enregistrement sécurisée ---
|
|
for m_id, data in models_to_process.items():
|
|
try:
|
|
existing = db.query(AIModel).filter(AIModel.model_id == m_id).first()
|
|
if existing:
|
|
existing.pricing_prompt = data["pp"]; existing.pricing_completion = data["pc"]; existing.is_free = data["is_free"]; existing.last_updated = datetime.now(timezone.utc)
|
|
else:
|
|
db.add(AIModel(provider=data["provider"], model_id=m_id, name=data["name"], description_fr=data["desc"], domain=data["domain"], is_free=data["is_free"], context_length=data["ctx"], pricing_prompt=data["pp"], pricing_completion=data["pc"]))
|
|
added += 1
|
|
# Commit très fréquent pour éviter les gros blocs qui plantent
|
|
if added % 50 == 0:
|
|
db.commit()
|
|
except IntegrityError:
|
|
db.rollback() # Si conflit, on annule cette insertion et on continue
|
|
except Exception:
|
|
db.rollback()
|
|
|
|
try:
|
|
settings.last_sync_date = datetime.now(timezone.utc)
|
|
settings.last_sync_type = sync_type
|
|
db.commit()
|
|
except Exception as e:
|
|
db.rollback()
|
|
print(f"Erreur DB Commit Sync: {e}")
|
|
|
|
await sync_finances(db, settings)
|
|
return {"status": "success", "models_processed": added}
|
|
|
|
def get_client_for_model(db: Session, model_id: str, settings: SystemSettings):
|
|
"""Récupère dynamiquement le bon client OpenAI en fonction du fournisseur du modèle."""
|
|
model_db = db.query(AIModel).filter(AIModel.model_id == model_id).first()
|
|
provider = model_db.provider if model_db else "openrouter"
|
|
|
|
if provider == "gemini" and settings.gemini_api_key:
|
|
return AsyncOpenAI(base_url="https://generativelanguage.googleapis.com/v1beta/openai/", api_key=settings.gemini_api_key), "gemini"
|
|
elif provider == "groq" and settings.groq_api_key:
|
|
return AsyncOpenAI(base_url="https://api.groq.com/openai/v1", api_key=settings.groq_api_key), "groq"
|
|
elif provider == "deepseek" and settings.deepseek_api_key:
|
|
return AsyncOpenAI(base_url="https://api.deepseek.com/v1", api_key=settings.deepseek_api_key), "deepseek"
|
|
elif provider == "mistral" and settings.mistral_api_key:
|
|
return AsyncOpenAI(base_url="https://api.mistral.ai/v1", api_key=settings.mistral_api_key), "mistral"
|
|
elif provider == "cloudflare" and settings.cloudflare_account_id and settings.cloudflare_api_token:
|
|
return AsyncOpenAI(base_url=f"https://api.cloudflare.com/client/v4/accounts/{settings.cloudflare_account_id}/ai/v1", api_key=settings.cloudflare_api_token), "cloudflare"
|
|
|
|
return AsyncOpenAI(base_url="https://openrouter.ai/api/v1", api_key=settings.openrouter_api_key), "openrouter"
|
|
|
|
async def ask_agent(client, model_id, messages, provider="openrouter"):
|
|
kwargs = {"model": model_id, "messages": messages}
|
|
if provider == "openrouter":
|
|
kwargs["extra_headers"] = {"HTTP-Referer": "https://aethas38.duckdns.org", "X-Title": "AETHAS38 Orchestrator"}
|
|
resp = await client.chat.completions.create(**kwargs)
|
|
return resp.choices[0].message.content
|
|
|
|
async def run_orchestrator(db: Session, history: list, settings: SystemSettings, config: dict) -> str:
|
|
workers = config.get("workers", ["gemini-3.5-flash-lite"])
|
|
user_prompt = history[-1].content
|
|
formatted_history = [{"role": msg.role, "content": msg.content} for msg in history[:-1]]
|
|
|
|
final_response = ""
|
|
try:
|
|
if len(workers) == 1:
|
|
w_mod = workers[0]
|
|
client, provider = get_client_for_model(db, w_mod, settings)
|
|
final_response = await ask_agent(client, w_mod, formatted_history + [{"role": "user", "content": user_prompt}], provider)
|
|
else:
|
|
p_mod = config.get("prompter", "gemini-3.5-flash-lite")
|
|
p_client, p_prov = get_client_for_model(db, p_mod, settings)
|
|
optimized = await ask_agent(p_client, p_mod, [{"role": "system", "content": "Optimise cette requête."}, {"role": "user", "content": user_prompt}], p_prov)
|
|
|
|
w_tasks = []
|
|
for w in workers:
|
|
w_client, w_prov = get_client_for_model(db, w, settings)
|
|
w_tasks.append(ask_agent(w_client, w, formatted_history + [{"role": "user", "content": optimized}], w_prov))
|
|
|
|
responses = await asyncio.gather(*w_tasks, return_exceptions=True)
|
|
|
|
c_mod = config.get("concatenator", "gemini-3.5-flash-lite")
|
|
c_client, c_prov = get_client_for_model(db, c_mod, settings)
|
|
synth = f"Requête: {user_prompt}\n\n" + "\n".join([f"--- EXPERT {i+1} ---\n{r}" for i, r in enumerate(responses)]) + "\n\nFais une synthèse finale."
|
|
final_response = await ask_agent(c_client, c_mod, [{"role": "user", "content": synth}], c_prov)
|
|
except Exception as e:
|
|
final_response = f"L'IA a rencontré une erreur critique: {str(e)}"
|
|
|
|
await sync_finances(db, settings)
|
|
return final_response |