feat(Ф2): llm_client.py — DI LLM (HttpxLLMClient + FakeLLMClient)

deploy/services/llm.py

@@ -1,34 +1,32 @@
-"""LLM service — call LLM API."""
+"""LLM service — call LLM API with optional DI."""
 import os, json
-import httpx
 
 LLM_URL = "https://api.aillm.ru/v1/chat/completions"
 LLM_KEY = os.environ.get("LLM_KEY") or os.environ.get("LLM_API_KEY", "")
 LLM_MODEL = "gpt-oss-120b"
 
+# Ленивый singleton — обратная совместимость
+_llm_client = None
+
+
+def _get_default_client():
+    global _llm_client
+    if _llm_client is None:
+        from .llm_client import HttpxLLMClient
+        _llm_client = HttpxLLMClient(url=LLM_URL, key=LLM_KEY, model=LLM_MODEL)
+    return _llm_client
+
+
+def call_llm(current_spec, doc_text, build_prompt_fn, llm_client=None):
+    """Call LLM. Returns (parsed_result, prompt_id).
+    llm_client: LLMClient (optional). Default — HttpxLLMClient (prod).
+    """
+    if llm_client is None:
+        llm_client = _get_default_client()
 
-def call_llm(current_spec, doc_text, build_prompt_fn):
-    """Call LLM. Returns (parsed_result, prompt_id)."""
     prompt, prompt_id = build_prompt_fn(current_spec, doc_text)
-    payload = {
-        "model": LLM_MODEL,
-        "messages": [{"role": "user", "content": prompt}],
-        "max_tokens": 8000,
-        "temperature": 0.1,
-    }
-    with httpx.Client(http2=True, timeout=120, verify=True) as client:
-        resp = client.post(
-            LLM_URL,
-            json=payload,
-            headers={
-                "Authorization": f"Bearer {LLM_KEY}",
-                "Content-Type": "application/json",
-            },
-        )
-        resp.raise_for_status()
-        data = resp.json()
+    raw_text = llm_client.complete(prompt)
 
-    raw_text = data.get("choices", [{}])[0].get("message", {}).get("content", "")
     json_text = raw_text
     if "```json" in json_text:
         json_text = json_text.split("```json")[1].split("```")[0]