integration openrouter
This commit is contained in:
1 parent
0ac8a14ea7
commit
2a98c37ca1
3 files changed
+113
-129
No files matched your search
+89
-115
@@ -1,20 +1,17 @@
|
||||
"""
|
||||
FILE: app/services/llm_service.py
|
||||
DESCRIPTION: Hybrid-Client für Ollama & Google Gemini.
|
||||
Verwaltet Prompts, Background-Last (Semaphore) und Cloud-Routing.
|
||||
VERSION: 3.1.0 (WP-20 Full Integration: Provider-Aware Prompting)
|
||||
STATUS: Active
|
||||
DEPENDENCIES: httpx, yaml, asyncio, google-generativeai, app.config
|
||||
EXTERNAL_CONFIG: config/prompts.yaml
|
||||
DESCRIPTION: Hybrid-Client für Ollama, Google GenAI und OpenRouter.
|
||||
Verwaltet provider-spezifische Prompts und Background-Last.
|
||||
VERSION: 3.3.0 (Full SDK Integration)
|
||||
"""
|
||||
|
||||
import httpx
|
||||
import yaml
|
||||
import logging
|
||||
import os
|
||||
import asyncio
|
||||
import json
|
||||
import google.generativeai as genai
|
||||
from google import genai
|
||||
from google.genai import types
|
||||
from openai import AsyncOpenAI # Für OpenRouter
|
||||
from pathlib import Path
|
||||
from typing import Optional, Dict, Any, Literal
|
||||
from app.config import get_settings
|
||||
@@ -22,122 +19,117 @@ from app.config import get_settings
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
class LLMService:
|
||||
# GLOBALER SEMAPHOR für Hintergrund-Last Steuerung (WP-06 / WP-20)
|
||||
_background_semaphore = None
|
||||
|
||||
def __init__(self):
|
||||
self.settings = get_settings()
|
||||
self.prompts = self._load_prompts()
|
||||
|
||||
# Initialisiere Semaphore einmalig auf Klassen-Ebene
|
||||
# WP-06: Semaphore-Initialisierung
|
||||
if LLMService._background_semaphore is None:
|
||||
limit = getattr(self.settings, "BACKGROUND_LIMIT", 2)
|
||||
logger.info(f"🚦 LLMService: Initializing Background Semaphore with limit: {limit}")
|
||||
limit = self.settings.BACKGROUND_LIMIT
|
||||
logger.info(f"🚦 LLMService: Background Semaphore initialized with limit: {limit}")
|
||||
LLMService._background_semaphore = asyncio.Semaphore(limit)
|
||||
|
||||
# Ollama Setup
|
||||
self.timeout = httpx.Timeout(self.settings.LLM_TIMEOUT, connect=10.0)
|
||||
# 1. Lokaler Ollama Client
|
||||
self.ollama_client = httpx.AsyncClient(
|
||||
base_url=self.settings.OLLAMA_URL,
|
||||
timeout=self.timeout
|
||||
timeout=httpx.Timeout(self.settings.LLM_TIMEOUT)
|
||||
)
|
||||
|
||||
# Gemini Setup [WP-20]
|
||||
if hasattr(self.settings, "GOOGLE_API_KEY") and self.settings.GOOGLE_API_KEY:
|
||||
genai.configure(api_key=self.settings.GOOGLE_API_KEY)
|
||||
model_name = getattr(self.settings, "GEMINI_MODEL", "gemini-1.5-flash")
|
||||
self.gemini_model = genai.GenerativeModel(model_name)
|
||||
logger.info(f"✨ LLMService: Gemini Cloud Mode active ({model_name})")
|
||||
else:
|
||||
self.gemini_model = None
|
||||
logger.warning("⚠️ LLMService: No GOOGLE_API_KEY found. Gemini mode disabled.")
|
||||
# 2. Google GenAI Client (Modern SDK)
|
||||
self.google_client = None
|
||||
if self.settings.GOOGLE_API_KEY:
|
||||
self.google_client = genai.Client(api_key=self.settings.GOOGLE_API_KEY)
|
||||
logger.info("✨ LLMService: Google GenAI (Gemini) active.")
|
||||
|
||||
# 3. OpenRouter Client
|
||||
self.openrouter_client = None
|
||||
if self.settings.OPENROUTER_API_KEY:
|
||||
self.openrouter_client = AsyncOpenAI(
|
||||
base_url="https://openrouter.ai/api/v1",
|
||||
api_key=self.settings.OPENROUTER_API_KEY
|
||||
)
|
||||
logger.info("🛰️ LLMService: OpenRouter Integration active.")
|
||||
|
||||
def _load_prompts(self) -> dict:
|
||||
"""Lädt die Prompt-Konfiguration aus der YAML-Datei."""
|
||||
path = Path(self.settings.PROMPTS_PATH)
|
||||
if not path.exists(): return {}
|
||||
try:
|
||||
with open(path, "r", encoding="utf-8") as f: return yaml.safe_load(f)
|
||||
with open(path, "r", encoding="utf-8") as f: return yaml.safe_load(f) or {}
|
||||
except Exception as e:
|
||||
logger.error(f"Failed to load prompts: {e}")
|
||||
return {}
|
||||
|
||||
def get_prompt(self, key: str, provider: str = None) -> str:
|
||||
"""
|
||||
Wählt das Template basierend auf dem Provider aus (WP-20).
|
||||
Unterstützt sowohl flache Strings als auch Dictionary-basierte Provider-Zweige.
|
||||
"""
|
||||
active_provider = provider or getattr(self.settings, "MINDNET_LLM_PROVIDER", "ollama")
|
||||
"""Hole provider-spezifisches Template mit Fallback-Kaskade."""
|
||||
active_provider = provider or self.settings.MINDNET_LLM_PROVIDER
|
||||
data = self.prompts.get(key, "")
|
||||
|
||||
if isinstance(data, dict):
|
||||
# Versuche den Provider-Key, Fallback auf 'ollama'
|
||||
return data.get(active_provider, data.get("ollama", ""))
|
||||
return str(data)
|
||||
|
||||
async def generate_raw_response(
|
||||
self,
|
||||
prompt: str,
|
||||
system: str = None,
|
||||
force_json: bool = False,
|
||||
max_retries: int = 2,
|
||||
base_delay: float = 2.0,
|
||||
self, prompt: str, system: str = None, force_json: bool = False,
|
||||
max_retries: int = 2, base_delay: float = 2.0,
|
||||
priority: Literal["realtime", "background"] = "realtime",
|
||||
provider: Optional[str] = None
|
||||
provider: Optional[str] = None,
|
||||
model_override: Optional[str] = None
|
||||
) -> str:
|
||||
"""
|
||||
Führt einen LLM Call aus mit Priority-Handling und Provider-Wahl.
|
||||
"""
|
||||
# Bestimme Provider: Parameter-Override > Config-Default
|
||||
target_provider = provider or getattr(self.settings, "MINDNET_LLM_PROVIDER", "ollama")
|
||||
"""Einstiegspunkt mit Priority-Handling."""
|
||||
target_provider = provider or self.settings.MINDNET_LLM_PROVIDER
|
||||
|
||||
use_semaphore = (priority == "background")
|
||||
|
||||
if use_semaphore and LLMService._background_semaphore:
|
||||
if priority == "background":
|
||||
async with LLMService._background_semaphore:
|
||||
return await self._dispatch_request(target_provider, prompt, system, force_json, max_retries, base_delay)
|
||||
else:
|
||||
return await self._dispatch_request(target_provider, prompt, system, force_json, max_retries, base_delay)
|
||||
return await self._dispatch(target_provider, prompt, system, force_json, max_retries, base_delay, model_override)
|
||||
return await self._dispatch(target_provider, prompt, system, force_json, max_retries, base_delay, model_override)
|
||||
|
||||
async def _dispatch_request(self, provider, prompt, system, force_json, max_retries, base_delay):
|
||||
"""Routet die Anfrage an den gewählten Provider mit Fallback-Logik."""
|
||||
async def _dispatch(self, provider, prompt, system, force_json, max_retries, base_delay, model_override):
|
||||
try:
|
||||
if provider == "gemini" and self.gemini_model:
|
||||
return await self._execute_gemini(prompt, system, force_json)
|
||||
else:
|
||||
return await self._execute_ollama(prompt, system, force_json, max_retries, base_delay)
|
||||
if provider == "openrouter" and self.openrouter_client:
|
||||
return await self._execute_openrouter(prompt, system, force_json, model_override)
|
||||
if provider == "gemini" and self.google_client:
|
||||
return await self._execute_google(prompt, system, force_json, model_override)
|
||||
return await self._execute_ollama(prompt, system, force_json, max_retries, base_delay)
|
||||
except Exception as e:
|
||||
# Automatischer Fallback auf Ollama bei Cloud-Fehlern (WP-20)
|
||||
if provider == "gemini" and getattr(self.settings, "LLM_FALLBACK_ENABLED", True):
|
||||
logger.warning(f"🔄 Gemini failed: {e}. Falling back to Ollama.")
|
||||
if self.settings.LLM_FALLBACK_ENABLED and provider != "ollama":
|
||||
logger.warning(f"🔄 Provider {provider} failed: {e}. Falling back to Ollama.")
|
||||
return await self._execute_ollama(prompt, system, force_json, max_retries, base_delay)
|
||||
raise e
|
||||
|
||||
async def _execute_gemini(self, prompt, system, force_json) -> str:
|
||||
"""Asynchroner Google Gemini Call (WP-20)."""
|
||||
full_prompt = f"System: {system}\n\nUser: {prompt}" if system else prompt
|
||||
|
||||
# Gemini JSON Mode Support
|
||||
gen_config = {}
|
||||
if force_json:
|
||||
gen_config["response_mime_type"] = "application/json"
|
||||
|
||||
response = await self.gemini_model.generate_content_async(
|
||||
full_prompt,
|
||||
generation_config=gen_config
|
||||
async def _execute_google(self, prompt, system, force_json, model_override):
|
||||
"""Native Google SDK Integration."""
|
||||
model = model_override or self.settings.GEMINI_MODEL
|
||||
config = types.GenerateContentConfig(
|
||||
system_instruction=system,
|
||||
response_mime_type="application/json" if force_json else "text/plain"
|
||||
)
|
||||
# Synchroner SDK-Call in Thread auslagern
|
||||
response = await asyncio.to_thread(
|
||||
self.google_client.models.generate_content,
|
||||
model=model, contents=prompt, config=config
|
||||
)
|
||||
return response.text.strip()
|
||||
|
||||
async def _execute_ollama(self, prompt, system, force_json, max_retries, base_delay) -> str:
|
||||
"""Ollama Call mit exponentieller Backoff-Retry-Logik."""
|
||||
payload: Dict[str, Any] = {
|
||||
"model": self.settings.LLM_MODEL,
|
||||
"prompt": prompt,
|
||||
"stream": False,
|
||||
"options": {
|
||||
"temperature": 0.1 if force_json else 0.7,
|
||||
"num_ctx": 8192
|
||||
}
|
||||
async def _execute_openrouter(self, prompt, system, force_json, model_override):
|
||||
"""OpenRouter (OpenAI-kompatibel)."""
|
||||
model = model_override or self.settings.OPENROUTER_MODEL
|
||||
messages = []
|
||||
if system: messages.append({"role": "system", "content": system})
|
||||
messages.append({"role": "user", "content": prompt})
|
||||
|
||||
response = await self.openrouter_client.chat.completions.create(
|
||||
model=model,
|
||||
messages=messages,
|
||||
response_format={"type": "json_object"} if force_json else None
|
||||
)
|
||||
return response.choices[0].message.content.strip()
|
||||
|
||||
async def _execute_ollama(self, prompt, system, force_json, max_retries, base_delay):
|
||||
"""Ollama mit exponentiellem Backoff."""
|
||||
payload = {
|
||||
"model": self.settings.LLM_MODEL, "prompt": prompt, "stream": False,
|
||||
"options": {"temperature": 0.1 if force_json else 0.7, "num_ctx": 8192}
|
||||
}
|
||||
if force_json: payload["format"] = "json"
|
||||
if system: payload["system"] = system
|
||||
@@ -145,41 +137,23 @@ class LLMService:
|
||||
attempt = 0
|
||||
while True:
|
||||
try:
|
||||
response = await self.ollama_client.post("/api/generate", json=payload)
|
||||
if response.status_code == 200:
|
||||
return response.json().get("response", "").strip()
|
||||
response.raise_for_status()
|
||||
res = await self.ollama_client.post("/api/generate", json=payload)
|
||||
res.raise_for_status()
|
||||
return res.json().get("response", "").strip()
|
||||
except Exception as e:
|
||||
attempt += 1
|
||||
if attempt > max_retries:
|
||||
logger.error(f"Ollama Error after {attempt} retries: {e}")
|
||||
raise e
|
||||
# Exponentieller Backoff: base_delay * (2 ^ (attempt - 1))
|
||||
wait_time = base_delay * (2 ** (attempt - 1))
|
||||
logger.warning(f"⚠️ Ollama attempt {attempt} failed. Retrying in {wait_time}s...")
|
||||
await asyncio.sleep(wait_time)
|
||||
if attempt > max_retries: raise e
|
||||
wait = base_delay * (2 ** (attempt - 1))
|
||||
logger.warning(f"⚠️ Ollama retry {attempt} in {wait}s...")
|
||||
await asyncio.sleep(wait)
|
||||
|
||||
async def generate_rag_response(self, query: str, context_str: str) -> str:
|
||||
"""Standard RAG Chat-Interface mit Provider-spezifischen Templates."""
|
||||
provider = getattr(self.settings, "MINDNET_LLM_PROVIDER", "ollama")
|
||||
|
||||
# Holen der Templates über die neue get_prompt Methode
|
||||
system_prompt = self.get_prompt("system_prompt", provider)
|
||||
rag_template = self.get_prompt("rag_template", provider)
|
||||
|
||||
# Fallback für RAG Template Struktur
|
||||
if not rag_template:
|
||||
rag_template = "{context_str}\n\n{query}"
|
||||
|
||||
final_prompt = rag_template.format(context_str=context_str, query=query)
|
||||
|
||||
return await self.generate_raw_response(
|
||||
final_prompt,
|
||||
system=system_prompt,
|
||||
priority="realtime"
|
||||
)
|
||||
"""Vollständiger RAG-Wrapper."""
|
||||
provider = self.settings.MINDNET_LLM_PROVIDER
|
||||
system = self.get_prompt("system_prompt", provider)
|
||||
template = self.get_prompt("rag_template", provider)
|
||||
final_prompt = template.format(context_str=context_str, query=query)
|
||||
return await self.generate_raw_response(final_prompt, system=system, priority="realtime")
|
||||
|
||||
async def close(self):
|
||||
"""Schließt alle offenen HTTP-Verbindungen."""
|
||||
if self.ollama_client:
|
||||
await self.ollama_client.aclose()
|
||||
await self.ollama_client.aclose()
|
||||
Reference in new issue
Block a user