do caching ourself on /model

llm_server/routes/v1/info.py

@@ -1,6 +1,6 @@
 import time
 
-from flask import jsonify
+from flask import jsonify, request
 
 from . import bp
 from ..helpers.http import cache_control
@@ -20,18 +20,27 @@ from ..cache import cache
 
 
 @bp.route('/model', methods=['GET'])
-@cache.cached(timeout=60, query_string=True)
 def get_model():
+    # We will manage caching ourself since we don't want to cache
+    # when the backend is down. Also, Cloudflare won't cache 500 errors.
+    cache_key = 'model_cache::' + request.url
+    cached_response = cache.get(cache_key)
+
+    if cached_response:
+        return cached_response
+
     model, error = get_running_model()
     if not model:
-        return jsonify({
-            'result': None,
+        response = jsonify({
             'code': 502,
             'error': 'failed to reach backend',
             'type': error.__class__.__name__
-        }), 200  # return 200 so Cloudflare caches the response
+        }), 500  # return 500 so Cloudflare doesn't intercept us
     else:
-        return jsonify({
+        response = jsonify({
             'result': model,
             'timestamp': int(time.time())
         }), 200
+        cache.set(cache_key, response, timeout=60)
+
+    return response