Re-added rate limiting, fixed a bunch of issues

api/config/config.yml

@@ -0,0 +1,59 @@
+max-credits: 100001
+max-credits-owner: 694201337
+start-credits: 1000
+
+costs:
+  other: 10
+
+  chat-models:
+    gpt-3: 10
+    gpt-4: 30
+    gpt-4-32k: 100
+
+## Roles Explanation
+
+# Bonuses: They are a multiplier for costs
+#          They work like: final_cost = cost * bonus
+# Rate limits: Limit the requests of the user
+#              The rate limit is by how many seconds until a new request can be done.
+
+## TODO: Setup proper rate limit settings for each role
+##       Current settings are:
+##       **NOT MEANT FOR PRODUCTION. DO NOT USE WITH THESE SETTINGS.**
+
+roles:
+  owner:
+    bonus: 0.1
+    rate_limit: 
+      other: 60
+      gpt-3: 60
+      gpt-4: 35
+      gpt-4-32k: 5
+  admin:
+    bonus: 0.3
+    rate_limit: 
+      other: 60
+      gpt-3: 60
+      gpt-4: 30
+      gpt-4-32k: 4
+  helper:
+    bonus: 0.4
+    rate_limit: 
+      other: 60
+      gpt-3: 60
+      gpt-4: 25
+      gpt-4-32k: 3
+  booster:
+    bonus: 0.5
+    rate_limit: 
+      other: 60 
+      gpt-3: 60
+      gpt-4: 20
+      gpt-4-32k: 2
+  default:
+    bonus: 0
+    rate_limit: 
+      other: 60 
+      gpt-3: 60
+      gpt-4: 15
+      gpt-4-32k: 1

api/config/credits.yml

@@ -1,19 +0,0 @@
-max-credits: 100001
-max-credits-owner: 694201337
-start-credits: 1000
-
-costs:
-  other: 10
-
-  chat-models:
-    gpt-3: 10
-    gpt-4: 30
-    gpt-4-32k: 100
-
-# bonuses are multiplier for costs:
-# final_cost = cost * bonus
-bonuses:
-  owner: 0.1
-  admin: 0.3
-  helper: 0.4
-  booster: 0.5

api/db/stats.py

@@ -26,7 +26,7 @@ class StatsManager:
         self.conn = AsyncIOMotorClient(os.getenv('MONGO_URI'))
 
     async def _get_collection(self, collection_name: str):
-        return self.conn[os.getenv('MONGO_NAME', 'nova-test')][collection_name]
+        return self.conn['nova-core'][collection_name]
     
     async def add_date(self):
         date = datetime.datetime.now(pytz.timezone('GMT')).strftime('%Y.%m.%d')
@@ -62,5 +62,6 @@ class StatsManager:
         return await db.find_one({obj_filter})
 
 if __name__ == '__main__':
-    asyncio.run(Stats.add_date())
-    asyncio.run(Stats.add_path('/__demo/test'))
+    stats = StatsManager()
+    asyncio.run(stats.add_date())
+    asyncio.run(stats.add_path('/__demo/test'))

api/db/users.py

@@ -9,7 +9,7 @@ from motor.motor_asyncio import AsyncIOMotorClient
 
 load_dotenv()
 
-with open('config/credits.yml', encoding='utf8') as f:
+with open('config/config.yml', encoding='utf8') as f:
     credits_config = yaml.safe_load(f)
 
 ## MONGODB Setup

api/streaming.py

@@ -26,12 +26,12 @@ import yaml
 load_dotenv()
 
 ## Loads config which contains rate limits
-with open('config/credits.yml', encoding='utf8') as f:
+with open('config/config.yml', encoding='utf8') as f:
     config = yaml.safe_load(f)
 
 ## Where all rate limit requested data will be stored.
 # Rate limit data is **not persistent** (It will be deleted on server stop/restart).
-# user_last_request_time = {}
+user_last_request_time = {}
 
 DEMO_PAYLOAD = {
     'model': 'gpt-3.5-turbo',
@@ -93,18 +93,18 @@ async def stream(
     ## Rate limits user.
     # If rate limit is exceeded, error code 429. Otherwise, lets the user pass but notes down
     # last request time for future requests.
-    # if user:
-    #     role = user.get('role', 'default')
-    #     rate_limit = config['roles'][role]['rate_limit'].get(payload['model'], 10)
+    if user:
+        role = user.get('role', 'default')
+        rate_limit = config['roles'].get(role, 1)['rate_limit'].get(payload['model'], 1)
 
-    #     last_request_time = user_last_request_time.get(user['api_key'])
-    #     time_since_last_request = datetime.datetime.now() - last_request_time
+        last_request_time = user_last_request_time.get(user['api_key'])
+        time_since_last_request = datetime.now() - last_request_time
 
-    #     if time_since_last_request < datetime.timedelta(seconds=rate_limit):
-    #         yield await errors.yield_error(429, 'Rate limit exceeded', "You are making requests too quickly. Please wait and try again later. Ask a administrator if you think this shouldn't happen.")
-    #         return
-    #     else:
-    #         user_last_request_time[user['_id']] = datetime.datetime.now()
+        if time_since_last_request < datetime.timedelta(seconds=rate_limit):
+            yield await errors.yield_error(429, "Rate limit exceeded', 'You are making requests too quickly. Please wait and try again later. Ask a administrator if you think this shouldn't happen. ")
+            return
+        else:
+            user_last_request_time[user['_id']] = datetime.now()
 
     ## Setup managers
     db = UserManager()

api/transfer.py

@@ -16,8 +16,8 @@ load_dotenv()
 
 models_list = json.load(open('models.json'))
 
-with open('config/credits.yml', encoding='utf8') as f:
-    credits_config = yaml.safe_load(f)
+with open('config/config.yml', encoding='utf8') as f:
+    config = yaml.safe_load(f)
 
 async def handle(incoming_request):
     """
@@ -62,7 +62,7 @@ async def handle(incoming_request):
     if path_contains_models:
         return fastapi.responses.JSONResponse(content=models_list)
 
-    costs = credits_config['costs']
+    costs = config['costs']
     cost = costs['other']
 
     if 'chat/completions' in path:
@@ -77,7 +77,8 @@ async def handle(incoming_request):
     if policy_violation:
         return await errors.error(400, f'The request contains content which violates this model\'s policies for "{policy_violation}".', 'We currently don\'t support any NSFW models.')
 
-    role_cost_multiplier = credits_config['bonuses'].get(user['role'], 1)
+    role = user.get('role', 'default')
+    role_cost_multiplier = config['roles'].get(role, 1)['bonus']
     cost = round(cost * role_cost_multiplier)
 
     if user['credits'] < cost:

rewards/autocredits.py

@@ -1,15 +1,12 @@
-import os
+import sys
+# Weird hack because PYTHON IS GOOD LANGUAGE :))))
+sys.path.append('../')
+from api.db.users import UserManager
 
-from dotenv import load_dotenv
-
-load_dotenv()
-
-async def get_all_users(client):
-    users = client[os.getenv('MONGO_NAME', 'nova-test')]['users']
-    return users
 
 async def update_credits(pymongo_client, settings=None):
-    users = await get_all_users(pymongo_client)
+    manager = UserManager()
+    users = await manager.get_all_users(pymongo_client)
 
     if not settings:
         users.update_many({}, {'$inc': {'credits': 2500}})
@@ -17,8 +14,4 @@ async def update_credits(pymongo_client, settings=None):
     else:
         for key, value in settings.items():
             users.update_many(
-                {'level': key},
-                {'$inc':
-                    {'credits': int(value)}
-                }
-            )
+                {'level': key}, {'$inc': {'credits': int(value)}})

tests/__main__.py

@@ -10,16 +10,7 @@ from dotenv import load_dotenv
 load_dotenv()
 
 MODEL = 'gpt-3.5-turbo'
-# MESSAGES = [
-#     {
-#         'role': 'system',
-#         'content': 'Always answer with "3", no matter what the user asks for. No exceptions. Just answer with the number "3". Nothing else. Just "3". No punctuation.'
-#     },
-#     {
-#         'role': 'user',
-#         'content': '1+1=',
-#     },
-# ]
+
 MESSAGES = [
     {
         'role': 'user',