WP15 #9
|
|
@ -1,6 +1,6 @@
|
|||
"""
|
||||
app/services/llm_service.py — LLM Client (Ollama)
|
||||
Version: 0.5.2 (Fix: Removed strict limits, increased Context)
|
||||
app/services/llm_service.py — LLM Client
|
||||
Version: 2.7.0 (Clean Architecture: Explicit Priority Queues)
|
||||
"""
|
||||
|
||||
import httpx
|
||||
|
|
@ -9,13 +9,12 @@ import logging
|
|||
import os
|
||||
import asyncio
|
||||
from pathlib import Path
|
||||
from typing import Optional, Dict, Any
|
||||
from typing import Optional, Dict, Any, Literal
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
class Settings:
|
||||
OLLAMA_URL = os.getenv("MINDNET_OLLAMA_URL", "http://127.0.0.1:11434")
|
||||
# Timeout für die Generierung (lang)
|
||||
LLM_TIMEOUT = float(os.getenv("MINDNET_LLM_TIMEOUT", 300.0))
|
||||
LLM_MODEL = os.getenv("MINDNET_LLM_MODEL", "phi3:mini")
|
||||
PROMPTS_PATH = os.getenv("MINDNET_PROMPTS_PATH", "./config/prompts.yaml")
|
||||
|
|
@ -24,16 +23,13 @@ def get_settings():
|
|||
return Settings()
|
||||
|
||||
class LLMService:
|
||||
# GLOBALER SEMAPHOR (Drosselung für Hintergrund-Prozesse)
|
||||
_background_semaphore = asyncio.Semaphore(2)
|
||||
|
||||
def __init__(self):
|
||||
self.settings = get_settings()
|
||||
self.prompts = self._load_prompts()
|
||||
|
||||
# FIX 1: Keine künstlichen Limits mehr. httpx defaults (100) sind besser.
|
||||
# Wir wollen nicht, dass der Chat wartet, nur weil im Hintergrund Embeddings laufen.
|
||||
|
||||
# Timeout-Konfiguration:
|
||||
# connect=10.0: Wenn Ollama nicht da ist, failen wir schnell.
|
||||
# read=LLM_TIMEOUT: Wenn Ollama denkt, geben wir ihm Zeit.
|
||||
self.timeout = httpx.Timeout(self.settings.LLM_TIMEOUT, connect=10.0)
|
||||
|
||||
self.client = httpx.AsyncClient(
|
||||
|
|
@ -43,11 +39,9 @@ class LLMService:
|
|||
|
||||
def _load_prompts(self) -> dict:
|
||||
path = Path(self.settings.PROMPTS_PATH)
|
||||
if not path.exists():
|
||||
return {}
|
||||
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)
|
||||
except Exception as e:
|
||||
logger.error(f"Failed to load prompts: {e}")
|
||||
return {}
|
||||
|
|
@ -58,19 +52,31 @@ class LLMService:
|
|||
system: str = None,
|
||||
force_json: bool = False,
|
||||
max_retries: int = 0,
|
||||
base_delay: float = 2.0
|
||||
base_delay: float = 2.0,
|
||||
priority: Literal["realtime", "background"] = "realtime" # <--- NEU & EXPLIZIT
|
||||
) -> str:
|
||||
"""
|
||||
Führt einen LLM Call aus.
|
||||
Führt einen LLM Call aus.
|
||||
priority="realtime": Chat (Sofort, keine Bremse).
|
||||
priority="background": Import/Analyse (Gedrosselt durch Semaphore).
|
||||
"""
|
||||
|
||||
# Entscheidung basierend auf explizitem Parameter, nicht Format!
|
||||
use_semaphore = (priority == "background")
|
||||
|
||||
if use_semaphore:
|
||||
async with LLMService._background_semaphore:
|
||||
return await self._execute_request(prompt, system, force_json, max_retries, base_delay)
|
||||
else:
|
||||
return await self._execute_request(prompt, system, force_json, max_retries, base_delay)
|
||||
|
||||
async def _execute_request(self, prompt, system, force_json, max_retries, base_delay):
|
||||
payload: Dict[str, Any] = {
|
||||
"model": self.settings.LLM_MODEL,
|
||||
"prompt": prompt,
|
||||
"stream": False,
|
||||
"options": {
|
||||
"temperature": 0.1 if force_json else 0.7,
|
||||
# FIX 2: Kontext auf 8192 erhöht.
|
||||
# Wichtig für komplexe Schemas und JSON-Stabilität.
|
||||
"num_ctx": 8192
|
||||
}
|
||||
}
|
||||
|
|
@ -97,7 +103,6 @@ class LLMService:
|
|||
attempt += 1
|
||||
if attempt > max_retries:
|
||||
logger.error(f"LLM Final Error (Versuch {attempt}): {e}")
|
||||
# Wir werfen den Fehler weiter, damit der Router nicht "Interner Fehler" als Typ interpretiert
|
||||
raise e
|
||||
|
||||
wait_time = base_delay * (2 ** (attempt - 1))
|
||||
|
|
@ -106,8 +111,7 @@ class LLMService:
|
|||
|
||||
async def generate_rag_response(self, query: str, context_str: str) -> str:
|
||||
"""
|
||||
WICHTIG FÜR CHAT:
|
||||
Kein JSON, keine Retries (User-Latency).
|
||||
Chat-Wrapper: Immer Realtime.
|
||||
"""
|
||||
system_prompt = self.prompts.get("system_prompt", "")
|
||||
rag_template = self.prompts.get("rag_template", "{context_str}\n\n{query}")
|
||||
|
|
@ -117,7 +121,9 @@ class LLMService:
|
|||
return await self.generate_raw_response(
|
||||
final_prompt,
|
||||
system=system_prompt,
|
||||
max_retries=0
|
||||
max_retries=0,
|
||||
force_json=False,
|
||||
priority="realtime" # <--- Standard
|
||||
)
|
||||
|
||||
async def close(self):
|
||||
|
|
|
|||
|
|
@ -1,6 +1,6 @@
|
|||
"""
|
||||
app/services/semantic_analyzer.py — Edge Validation & Filtering
|
||||
Version: 1.4 (Merged: Retry Strategy + Extended Observability)
|
||||
Version: 2.0 (Update: Background Priority for Batch Jobs)
|
||||
"""
|
||||
|
||||
import json
|
||||
|
|
@ -24,6 +24,7 @@ class SemanticAnalyzer:
|
|||
|
||||
Features:
|
||||
- Retry Strategy: Wartet bei Überlastung (max_retries=5).
|
||||
- Priority Queue: Läuft als "background" Task, um den Chat nicht zu blockieren.
|
||||
- Observability: Loggt Input-Größe, Raw-Response und Parsing-Details.
|
||||
"""
|
||||
if not all_edges:
|
||||
|
|
@ -44,27 +45,27 @@ class SemanticAnalyzer:
|
|||
# 2. Kandidaten-Liste formatieren
|
||||
edges_str = "\n".join([f"- {e}" for e in all_edges])
|
||||
|
||||
# LOG: Request Info (Wichtig für Debugging)
|
||||
# LOG: Request Info
|
||||
logger.debug(f"🔍 [SemanticAnalyzer] Request: {len(chunk_text)} chars Text, {len(all_edges)} Candidates.")
|
||||
|
||||
# 3. Prompt füllen
|
||||
final_prompt = prompt_template.format(
|
||||
chunk_text=chunk_text[:3500], # Etwas mehr Kontext als früher (3000 -> 3500)
|
||||
chunk_text=chunk_text[:3500],
|
||||
edge_list=edges_str
|
||||
)
|
||||
|
||||
try:
|
||||
# 4. LLM Call mit JSON Erzwingung UND Retry-Logik (Merged V1.3)
|
||||
# max_retries=5 bedeutet: 5s -> 10s -> 20s -> 40s -> 80s Pause.
|
||||
# 4. LLM Call mit Traffic Control (NEU: priority="background")
|
||||
# Wir nutzen die "Slow Lane", damit der User im Chat nicht warten muss.
|
||||
response_json = await self.llm.generate_raw_response(
|
||||
prompt=final_prompt,
|
||||
force_json=True,
|
||||
max_retries=5,
|
||||
base_delay=5.0
|
||||
base_delay=5.0,
|
||||
priority="background" # <--- WICHTIG: Drosselung aktivieren
|
||||
)
|
||||
|
||||
# LOG: Raw Response Preview (Wichtig um zu sehen, was das LLM liefert)
|
||||
# Zeigt nur die ersten 200 Zeichen, um Log nicht zu fluten
|
||||
# LOG: Raw Response Preview
|
||||
logger.debug(f"📥 [SemanticAnalyzer] Raw Response (Preview): {response_json[:200]}...")
|
||||
|
||||
# 5. Parsing & Cleaning
|
||||
|
|
@ -77,10 +78,9 @@ class SemanticAnalyzer:
|
|||
try:
|
||||
data = json.loads(clean_json)
|
||||
except json.JSONDecodeError as json_err:
|
||||
# LOG: Detaillierter Fehlerbericht für den User
|
||||
logger.error(f"❌ [SemanticAnalyzer] JSON Decode Error.")
|
||||
logger.error(f" Grund: {json_err}")
|
||||
logger.error(f" Empfangener String: {clean_json[:500]}") # Zeige max 500 chars des Fehlers
|
||||
logger.error(f" Empfangener String: {clean_json[:500]}")
|
||||
logger.info(" -> Workaround: Fallback auf 'Alle Kanten' (durch Chunker).")
|
||||
return []
|
||||
|
||||
|
|
@ -92,7 +92,7 @@ class SemanticAnalyzer:
|
|||
valid_edges = [str(e) for e in data if isinstance(e, str) and ":" in e]
|
||||
|
||||
elif isinstance(data, dict):
|
||||
# Abweichende Formate behandeln (Extended Logging V1.2)
|
||||
# Abweichende Formate behandeln
|
||||
logger.info(f"ℹ️ [SemanticAnalyzer] LLM lieferte Dict statt Liste. Versuche Reparatur. Keys: {list(data.keys())}")
|
||||
|
||||
for key, val in data.items():
|
||||
|
|
@ -100,7 +100,7 @@ class SemanticAnalyzer:
|
|||
if key.lower() in ["edges", "results", "kanten", "matches"] and isinstance(val, list):
|
||||
valid_edges.extend([str(e) for e in val if isinstance(e, str) and ":" in e])
|
||||
|
||||
# Fall B: {"kind": "target"} (Das beobachtete Format im Log)
|
||||
# Fall B: {"kind": "target"}
|
||||
elif isinstance(val, str):
|
||||
valid_edges.append(f"{key}:{val}")
|
||||
|
||||
|
|
@ -115,7 +115,6 @@ class SemanticAnalyzer:
|
|||
|
||||
# LOG: Ergebnis
|
||||
if final_result:
|
||||
# Nur Info, wenn wirklich was gefunden wurde, sonst spammt es bei leeren Chunks
|
||||
logger.info(f"✅ [SemanticAnalyzer] Success. {len(final_result)} Kanten zugewiesen.")
|
||||
else:
|
||||
logger.debug(" [SemanticAnalyzer] Keine spezifischen Kanten erkannt (Empty Result).")
|
||||
|
|
|
|||
Loading…
Reference in New Issue
Block a user