fixed proxy

web/backend/app/services/llm.py

@@ -37,6 +37,101 @@ TABLE_TYPE_LABELS = {
     "simple_list": "Бланк заказа",
 }
 
+
+def _openrouter_proxy_url() -> str | None:
+    settings = get_settings()
+    proxy = (settings.openrouter_proxy or "").strip() or None
+    if proxy and "://" not in proxy:
+        proxy = f"socks5://{proxy}"
+    return proxy
+
+
+def openrouter_headers() -> dict[str, str]:
+    settings = get_settings()
+    return {
+        "Authorization": f"Bearer {settings.openrouter_api_key}",
+        "Content-Type": "application/json",
+        "HTTP-Referer": "https://gostgenerator.local",
+        "X-Title": "GostGenerator Web",
+    }
+
+
+def openrouter_chat_url() -> str:
+    return get_settings().openrouter_base_url.rstrip("/") + "/chat/completions"
+
+
+async def openrouter_chat(
+    messages: list[dict[str, str]],
+    *,
+    temperature: float = 0.2,
+    max_tokens: int = 2000,
+) -> str:
+    """Shared OpenRouter chat/completions call (same key, proxy, model as ESKD chat)."""
+    settings = get_settings()
+    if not settings.openrouter_api_key:
+        raise RuntimeError("OpenRouter API key не настроен (OPENROUTER_API_KEY).")
+
+    payload = {
+        "model": settings.openrouter_model,
+        "messages": messages,
+        "temperature": temperature,
+        "max_tokens": max_tokens,
+    }
+    client_kwargs: dict[str, Any] = {"timeout": 120.0}
+    proxy = _openrouter_proxy_url()
+    if proxy:
+        client_kwargs["proxy"] = proxy
+
+    async with httpx.AsyncClient(**client_kwargs) as client:
+        resp = await client.post(
+            openrouter_chat_url(),
+            headers=openrouter_headers(),
+            json=payload,
+        )
+        if resp.status_code >= 400:
+            detail = resp.text[:500]
+            raise RuntimeError(f"OpenRouter HTTP {resp.status_code}: {detail}")
+        data = resp.json()
+
+    return data["choices"][0]["message"]["content"] or ""
+
+
+def openrouter_chat_sync(
+    messages: list[dict[str, str]],
+    *,
+    temperature: float = 0.2,
+    max_tokens: int = 2000,
+) -> str:
+    """Sync variant for non-async endpoints (same transport as ESKD chat)."""
+    settings = get_settings()
+    if not settings.openrouter_api_key:
+        raise RuntimeError("OpenRouter API key не настроен (OPENROUTER_API_KEY).")
+
+    payload = {
+        "model": settings.openrouter_model,
+        "messages": messages,
+        "temperature": temperature,
+        "max_tokens": max_tokens,
+    }
+    client_kwargs: dict[str, Any] = {"timeout": 120.0}
+    proxy = _openrouter_proxy_url()
+    if proxy:
+        client_kwargs["proxy"] = proxy
+
+    with httpx.Client(**client_kwargs) as client:
+        resp = client.post(
+            openrouter_chat_url(),
+            headers=openrouter_headers(),
+            json=payload,
+        )
+        if resp.status_code >= 400:
+            detail = resp.text[:500]
+            raise RuntimeError(f"OpenRouter HTTP {resp.status_code}: {detail}")
+        data = resp.json()
+
+    return data["choices"][0]["message"]["content"] or ""
+
+
 def estimate_tokens(text: str) -> int:
     """Rough token estimate (works for mixed RU/EN)."""
     if not text:
@@ -159,36 +254,8 @@ async def chat_edit_table(
         hist.append({"role": "user", "content": message})
 
     messages = build_llm_messages(table_type, rows, hist)
-    context = build_context_usage(messages)
+    content = await openrouter_chat(messages, temperature=0.2, max_tokens=2000)
 
-    payload = {
-        "model": settings.openrouter_model,
-        "messages": messages,
-        "temperature": 0.2,
-        "max_tokens": 2000,
-    }
-    headers = {
-        "Authorization": f"Bearer {settings.openrouter_api_key}",
-        "Content-Type": "application/json",
-        "HTTP-Referer": "https://gostgenerator.local",
-        "X-Title": "GostGenerator Web",
-    }
-    url = settings.openrouter_base_url.rstrip("/") + "/chat/completions"
-
-    proxy = (settings.openrouter_proxy or "").strip() or None
-    if proxy and "://" not in proxy:
-        proxy = f"socks5://{proxy}"
-
-    client_kwargs: dict[str, Any] = {"timeout": 90.0}
-    if proxy:
-        client_kwargs["proxy"] = proxy
-
-    async with httpx.AsyncClient(**client_kwargs) as client:
-        resp = await client.post(url, headers=headers, json=payload)
-        resp.raise_for_status()
-        data = resp.json()
-
-    content = data["choices"][0]["message"]["content"]
     parsed = _extract_json(content)
     edits = parsed.get("edits") or []
     norm_edits = []
@@ -204,22 +271,7 @@ async def chat_edit_table(
             }
         )
 
-    # include assistant reply in usage estimate
     context = build_context_usage(messages + [{"role": "assistant", "content": content}])
-    usage = data.get("usage") or {}
-    if usage.get("prompt_tokens"):
-        prompt_t = int(usage["prompt_tokens"])
-        completion_t = int(usage.get("completion_tokens") or 0)
-        total_t = int(usage.get("total_tokens") or (prompt_t + completion_t))
-        limit = settings.llm_context_limit
-        context = {
-            "used": total_t,
-            "limit": limit,
-            "percent": min(100.0, round(100.0 * total_t / limit, 1)) if limit else 0.0,
-            "prompt_tokens": prompt_t,
-            "completion_tokens": completion_t,
-        }
-
     return {"reply": parsed.get("reply") or "", "edits": norm_edits, "context": context}