Fix dangling asyncio tasks

api/db/logs.py

@@ -1,21 +1,17 @@
 import os
 import time
+import asyncio
 
 from dotenv import load_dotenv
 from motor.motor_asyncio import AsyncIOMotorClient
 
-from helpers import network
+try:
+    from helpers import network
+except ImportError:
+    pass
 
 load_dotenv()
 
-UA_SIMPLIFY = {
-    'Windows NT': 'W',
-    'Mozilla/5.0': 'M',
-    'Win64; x64': '64',
-    'Safari/537.36': 'S',
-    'AppleWebKit/537.36 (KHTML, like Gecko)': 'K',
-}
-
 ## MONGODB Setup
 
 conn = AsyncIOMotorClient(os.environ['MONGO_URI'])
@@ -30,18 +26,7 @@ async def replacer(text: str, dict_: dict) -> str:
     return text
 
 async def log_api_request(user: dict, incoming_request, target_url: str):
-    """Logs the API Request into the database.
-    No input prompt is logged, however data such as IP & useragent is noted.
-    This would be useful for security reasons. Other minor data is also collected.
-
-    Args:
-        user (dict): User dict object
-        incoming_request (_type_): Request
-        target_url (str): The URL the api request was targetted to.
-
-    Returns:
-        _type_: _description_
-    """
+    """Logs the API Request into the database."""
     db = await _get_collection('logs')
     payload = {}
 
@@ -53,7 +38,6 @@ async def log_api_request(user: dict, incoming_request, target_url: str):
 
     model = payload.get('model')
     ip_address = await network.get_ip(incoming_request)
-    useragent = await replacer(incoming_request.headers.get('User-Agent', ''), UA_SIMPLIFY)
 
     new_log_item = {
         'timestamp': time.time(),
@@ -62,7 +46,6 @@ async def log_api_request(user: dict, incoming_request, target_url: str):
         'user_id': str(user['_id']),
         'security': {
             'ip': ip_address,
-            'useragent': useragent,
         },
         'details': {
             'model': model,
@@ -90,5 +73,21 @@ async def delete_by_user_id(user_id: str):
     db = await _get_collection('logs')
     return await db.delete_many({'user_id': user_id})
 
+async def get_logs_time_range(start: int, end: int):
+    db = await _get_collection('logs')
+
+    entries = []
+    async for entry in db.find({'timestamp': {'$gte': start, '$lte': end}}):
+        entries.append(entry)
+
+    return entries
+
+async def main():
+    # how many requests in last 24 hours?
+    last_24_hours = time.time() - 86400
+    logs = await get_logs_time_range(last_24_hours, time.time())
+
+    print(f'Number of logs in last 24 hours: {len(logs)}')
+
 if __name__ == '__main__':
-    pass
+    asyncio.run(main())

api/db/providerkeys.py

@@ -98,7 +98,7 @@ manager = KeyManager()
 
 async def main():
     keys = await manager.get_possible_keys('closed')
-    print(len(keys))
+    print(keys)
 
 if __name__ == '__main__':
     asyncio.run(main())

api/db/stats.py

@@ -2,8 +2,6 @@ import os
 import pytz
 import asyncio
 import datetime
-import json
-import time
 
 from dotenv import load_dotenv
 from motor.motor_asyncio import AsyncIOMotorClient
@@ -15,13 +13,6 @@ load_dotenv()
 class StatsManager:
     """
     ### The manager for all statistics tracking
-    Stats tracked:
-    - Dates
-    - IPs
-    - Target URLs
-    - Tokens
-    - Models
-    - URL Paths
     """
 
     def __init__(self):

api/db/tester.py

@@ -1,6 +0,0 @@
-from stats import * 
-import asyncio
-
-manager = StatsManager()
-
-asyncio.run(manager.get_model_usage())

api/providers/__init__.py

@@ -1,12 +1,12 @@
 from . import \
-    azure, \
-    closed, \
-    closed4
+    azure \
+    # closed, \
+    # closed4
     # closed432
 
 MODULES = [
     azure,
-    closed,
-    closed4,
+    # closed,
+    # closed4,
     # closed432,
 ]

api/providers/azure.py

@@ -12,7 +12,7 @@ MODELS = [
     'gpt-4',
     'gpt-4-32k'
 ]
-MODELS = [f'{model}-azure' for model in MODELS]
+# MODELS = [f'{model}-azure' for model in MODELS]
 
 AZURE_API = '2023-07-01-preview'
 

api/providers/helpers/utils.py

@@ -34,7 +34,4 @@ async def conversation_to_prompt(conversation: list) -> str:
     return text
 
 async def random_secret_for(name: str) -> str:
-    try:
-        return await providerkeys.manager.get_key(name)
-    except ValueError:
-        raise ValueError(f'Keys missing for "{name}" <no_keys>')
+    return await providerkeys.manager.get_key(name)

api/responder.py

@@ -2,11 +2,12 @@
 
 import os
 import json
-import logging
+import ujson
 import aiohttp
 import asyncio
 import starlette
 
+from typing import Any, Coroutine, Set
 from rich import print
 from dotenv import load_dotenv
 
@@ -23,6 +24,19 @@ CRITICAL_API_ERRORS = ['invalid_api_key', 'account_deactivated']
 
 keymanager = providerkeys.manager
 
+background_tasks: Set[asyncio.Task[Any]] = set()
+
+
+def create_background_task(coro: Coroutine[Any, Any, Any]) -> None:
+    """asyncio.create_task, which prevents the task from being garbage collected.
+
+    https://docs.python.org/3/library/asyncio-task.html#asyncio.create_task
+    """
+    task = asyncio.create_task(coro)
+    background_tasks.add(task)
+    task.add_done_callback(background_tasks.discard)
+
+
 async def respond(
     path: str='/v1/chat/completions',
     user: dict=None,
@@ -49,9 +63,7 @@ async def respond(
         'Content-Type': 'application/json'
     }
 
-    for i in range(20):
-        print(i)
-        # Load balancing: randomly selecting a suitable provider
+    for i in range(1):
         try:
             if is_chat:
                 target_request = await load_balancing.balance_chat_request(payload)
@@ -151,13 +163,21 @@ async def respond(
 
                         async for chunk in response.content.iter_any():
                             chunk = chunk.decode('utf8').strip()
+
+                            if 'azure' in provider_name:
+                                chunk = chunk.strip().replace('data: ', '')
+
+                                if not chunk or 'prompt_filter_results' in chunk:
+                                    continue
+
                             yield chunk + '\n\n'
 
                     break
 
             except Exception as exc:
                 print('[!] exception', exc)
-                continue
+                # continue
+                raise exc
 
     else:
         yield await errors.yield_error(500, 'Sorry, our API seems to have issues connecting to our provider(s).', 'This most likely isn\'t your fault. Please try again later.')
@@ -166,7 +186,7 @@ async def respond(
     if (not is_stream) and server_json_response:
         yield json.dumps(server_json_response)
 
-    asyncio.create_task(
+    create_background_task(
         after_request.after_request(
             incoming_request=incoming_request,
             target_request=target_request,

checks/client.py

@@ -215,7 +215,10 @@ async def demo():
         else:
             raise ConnectionError('API Server is not running.')
 
-        for func in [test_chat_non_stream_gpt4, test_chat_stream_gpt3]:
+        for func in [
+            # test_chat_non_stream_gpt4,
+            test_chat_stream_gpt3
+        ]:
             print(f'[*] {func.__name__}')
             result = await func()
             print(f'[+] {func.__name__} - {result}')

playground/notes.txt

@@ -0,0 +1,10 @@
+--- EXPECTED ---
+
+data: {"id":"custom-chatcmpl-nUSiapqELukaPT7vEnGcXkbvrS1fR","object":"chat.completion.chunk","created":1696716717,"model":"gpt-3.5-turbo-0613","choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null}]}
+
+data: {"id":"custom-chatcmpl-nUSiapqELukaPT7vEnGcXkbvrS1fR","object":"chat.completion.chunk","created":1696716717,"model":"gpt-3.5-turbo-0613","choices":[{"index":0,"delta":{"content":"123"},"finish_reason":null}]}
+
+data: {"id":"custom-chatcmpl-nUSiapqELukaPT7vEnGcXkbvrS1fR","object":"chat.completion.chunk","created":1696716717,"model":"gpt-3.5-turbo-0613","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}
+
+data: [DONE]
+