SysOps: approved_backup — 2026-08-16 11:01 UTC

cockpit/app/services/llm_router.py

@@ -255,6 +255,9 @@ async def chat_messages(
     provider_id: int | None = None,
     model: str | None = None,
     timeout: float = 120.0,
+    user: str | None = None,
+    max_tokens: int | None = None,
+    temperature: float | None = None,
 ) -> tuple[str, dict[str, Any]]:
     """Returns (reply_text, meta dict with provider info)."""
     prov = resolve_provider(provider_id)
@@ -309,9 +312,14 @@ async def chat_messages(
     payload: dict[str, Any] = {
         "model": use_model,
         "messages": messages,
-        "temperature": float(extra.get("temperature", 0.4)),
-        "max_tokens": int(extra.get("max_tokens", 2048)),
+        "temperature": float(temperature if temperature is not None else extra.get("temperature", 0.4)),
+        "max_tokens": int(max_tokens if max_tokens is not None else extra.get("max_tokens", 2048)),
     }
+    # Hermes: `user` = Telegram chat_id → koppelt API-chat aan bestaande memory/sessie
+    if user:
+        payload["user"] = str(user)
+    elif ptype == "hermes":
+        payload["user"] = str(getattr(settings, "HERMES_CEO_CHAT_ID", None) or "8859782446")
 
     async with httpx.AsyncClient(timeout=timeout) as client:
         resp = await client.post(url, headers=headers, json=payload)
@@ -332,9 +340,99 @@ async def chat_messages(
             "provider_type": ptype,
             "provider_label": prov.get("label"),
             "model": use_model,
+            "hermes_user": payload.get("user"),
         }
 
 
+
+async def stream_chat_messages(
+    messages: list[dict[str, str]],
+    *,
+    provider_id: int | None = None,
+    model: str | None = None,
+    timeout: float = 90.0,
+    user: str | None = None,
+    max_tokens: int | None = None,
+    temperature: float | None = None,
+):
+    """Yield text deltas from Hermes/OpenAI-compatible streaming chat."""
+    prov = resolve_provider(provider_id)
+    ptype = (prov.get("provider_type") or "ollama").lower()
+    use_model = model or prov.get("model") or settings.OLLAMA_MODEL
+    if ptype != "hermes":
+        # non-stream fallback: one chunk
+        reply, _meta = await chat_messages(
+            messages,
+            provider_id=provider_id,
+            model=model,
+            timeout=timeout,
+            user=user,
+            max_tokens=max_tokens,
+            temperature=temperature,
+        )
+        if reply:
+            yield reply
+        return
+
+    api_key = (prov.get("api_key") or "").strip()
+    if not api_key:
+        raise RuntimeError("Geen Hermes API key")
+    base = (prov.get("api_base_url") or "").strip().rstrip("/")
+    if not base:
+        raise RuntimeError("Geen Hermes API URL")
+    extra = prov.get("extra_config") or {}
+    if isinstance(extra, str):
+        try:
+            extra = json.loads(extra)
+        except Exception:
+            extra = {}
+    url = f"{base}/chat/completions"
+    headers = {
+        "Authorization": f"Bearer {api_key}",
+        "Content-Type": "application/json",
+        "Accept": "text/event-stream",
+    }
+    payload: dict[str, Any] = {
+        "model": use_model,
+        "messages": messages,
+        "temperature": float(temperature if temperature is not None else extra.get("temperature", 0.35)),
+        "max_tokens": int(max_tokens if max_tokens is not None else extra.get("max_tokens", 512)),
+        "stream": True,
+    }
+    if user:
+        payload["user"] = str(user)
+    else:
+        payload["user"] = str(getattr(settings, "HERMES_CEO_CHAT_ID", None) or "8859782446")
+
+    async with httpx.AsyncClient(timeout=timeout) as client:
+        async with client.stream("POST", url, headers=headers, json=payload) as resp:
+            if resp.status_code >= 400:
+                body = (await resp.aread()).decode("utf-8", errors="replace")[:500]
+                raise RuntimeError(f"Hermes stream fout ({resp.status_code}): {body}")
+            async for line in resp.aiter_lines():
+                if not line:
+                    continue
+                if line.startswith(":"):
+                    continue
+                if not line.startswith("data:"):
+                    continue
+                data = line[5:].strip()
+                if not data or data == "[DONE]":
+                    if data == "[DONE]":
+                        break
+                    continue
+                try:
+                    obj = json.loads(data)
+                except Exception:
+                    continue
+                choices = obj.get("choices") or []
+                if not choices:
+                    continue
+                delta = (choices[0].get("delta") or {}).get("content")
+                if delta:
+                    yield delta
+
+
 async def generate(
     prompt: str,
     system: str | None = None,