feat: save account model and thinking settings for Hermes

control/model_catalog.py

@@ -0,0 +1,63 @@
+"""Trusted upstream discovery. Never return provider URLs, keys or raw metadata."""
+from __future__ import annotations
+
+import asyncio
+import os
+import time
+
+import httpx
+
+
+class CatalogUnavailable(Exception):
+    pass
+
+
+class ModelCatalog:
+    def __init__(self, *, base_url=None, api_key=None, client=None):
+        self.base_url = (base_url or os.getenv('HERMES_UPSTREAM_URL', 'http://192.168.86.5:2099/v1')).rstrip('/')
+        self.api_key = api_key if api_key is not None else os.getenv('HERMES_UPSTREAM_API_KEY', '')
+        self.client = client or httpx.AsyncClient(timeout=10, trust_env=False, follow_redirects=False)
+        self._owns_client = client is None
+        self._cache = None
+        self._expires = 0
+        self._lock = asyncio.Lock()
+
+    async def aclose(self):
+        if self._owns_client:
+            await self.client.aclose()
+
+    async def models(self):
+        async with self._lock:
+            if self._cache is not None and time.monotonic() < self._expires:
+                return self._cache
+            try:
+                if not self.api_key:
+                    raise CatalogUnavailable()
+                async with self.client.stream('GET', self.base_url + '/models', headers={'Authorization': 'Bearer ' + self.api_key}) as response:
+                    response.raise_for_status()
+                    body = bytearray()
+                    async for chunk in response.aiter_bytes():
+                        body.extend(chunk)
+                        if len(body) > 2 * 1024 * 1024:
+                            raise CatalogUnavailable()
+                import json
+                data = json.loads(body).get('data')
+                if not isinstance(data, list):
+                    raise CatalogUnavailable()
+                models = []
+                seen = set()
+                for item in data[:1000]:
+                    if not isinstance(item, dict):
+                        continue
+                    model = item.get('id')
+                    if not isinstance(model, str) or not 1 <= len(model) <= 200 or not all(33 <= ord(c) <= 126 for c in model) or model in seen:
+                        continue
+                    seen.add(model)
+                    models.append({'id': model, 'name': model, 'thinking_levels': ['none', 'minimal', 'low', 'medium', 'high', 'xhigh']})
+                if not models:
+                    raise CatalogUnavailable()
+                self._cache = models
+                self._expires = time.monotonic() + 60
+                return models
+            except (httpx.HTTPError, ValueError, TypeError, AttributeError, CatalogUnavailable):
+                raise CatalogUnavailable() from None