.gitignore

@@ -1,20 +1,7 @@
-# Environments
+*.zip
-# !!! KEEP THESE ENTRIES AT THE TOP OF THIS FILE BECAUSE THEY CONTAIN THE MOST SENSITIVE DATA !!!
-.env
-.venv
-env/
-venv/
-ENV/
-env.bak/
-venv.bak/
-env.old/
-.prod.env
-# !!! KEEP THESE ENTRIES AT THE TOP OF THIS FILE BECAUSE THEY CONTAIN THE MOST SENSITIVE DATA !!!
 
-rate_limited_keys.json
 last_update.txt
 
-*.zip
 *.log.json
 /logs
 /log
@@ -155,6 +142,15 @@ celerybeat.pid
 # SageMath parsed files
 *.sage.py
 
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
 # Spyder project settings
 .spyderproject
 .spyproject
@@ -184,8 +180,6 @@ cython_debug/
 #  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
 #  and can be added to the global gitignore or merged into this file.  For a more nuclear
 #  option (not recommended) you can uncomment the following to ignore the entire idea folder.
-.idea/
+#.idea/
 
-backups/
+backups/
-cache/
-api/cache/rate_limited_keys.json

.vscode/settings.json

@@ -7,7 +7,6 @@
         "**/.DS_Store": true,
         "**/Thumbs.db": true,
         "**/__pycache__": true,
-        "**/*.css.map": true,
         "**/.vscode": true,
         "**/*.map": true,
         "tests/__pycache__": true

PUSH_TO_PRODUCTION.sh

@@ -4,7 +4,7 @@
 # git commit -am "Auto-trigger - Production server started" && git push origin Production
 
 # backup database
-# /usr/local/bin/python /home/nova-api/api/backup_manager/main.py pre_prodpush
+/usr/local/bin/python /home/nova-api/api/backup_manager/main.py pre_prodpush
 
 # Kill production server
 fuser -k 2333/tcp
@@ -22,4 +22,4 @@ cp env/.prod.env /home/nova-prod/.env
 cd /home/nova-prod
 
 # Start screen
-screen -L -S nova-api python run prod && sleep 5
+screen -S nova-api python run prod && sleep 5

README.md

@@ -125,45 +125,14 @@ Set up a MongoDB database and set `MONGO_URI` to the MongoDB database connection
 Want to use a proxy list? See the according section!
 Keep in mind to set `USE_PROXY_LIST` to `True`! Otherwise, the proxy list won't be used.
 
-### Proxy Lists
+### `ACTUAL_IPS` (optional)
-To use proxy lists, navigate to `api/secret/proxies/` and create the following files:
-- `http.txt`
-- `socks4.txt`
-- `socks5.txt`
-
-Then, paste your proxies in the following format:
-
-```
-[username:password@]host:port
-```
-Whereas anything inside of `[]` is optional and the host can be an IP address or a hostname. Always specify the port.
-
-If you're using [iproyal.com](https://iproyal.com?r=307932)<sup>affiliate link</sup>, follow the following steps:
-- Order any type of proxy or proxies
-- In the *Product Info* tab:
-  - set *Select port* to `SOCKS5`
-  - and *Select format* to `USER:PASS@IP:PORT`
-
-#### Proxy List Examples
-```
-1.2.3.4:8080
-user:pass@127.0.0.1:1337
-aaaaaaaaaaaaa:bbbbbbbbbb@1.2.3.4:5555
-```
-
-In the proxy credential files, can use comments just like in Python.
-
-**Important:** to activate the proxy lists, you need to change the `USE_PROXY_LIST` environment variable to `True`!
-
-
-### ~~`ACTUAL_IPS` (optional)~~ (deprecated, might come back in the future)
 This is a security measure to make sure a proxy, VPN, Tor or any other IP hiding service is used by the host when accessing "Closed"AI's API.
 It is a space separated list of IP addresses that are allowed to access the API.
 You can also just add the *beginning* of an API address, like `12.123.` (without an asterisk!) to allow all IPs starting with `12.123.`.
 > To disable the warning if you don't have this feature enabled, set `ACTUAL_IPS` to `None`.
 
 ### Timeout
-`TRANSFER_TIMEOUT` seconds to wait until the program throws an exception for if the request takes too long. We recommend rather long times like `500` for 500 seconds.
+`TRANSFER_TIMEOUT` seconds to wait until the program throws an exception for if the request takes too long. We recommend rather long times like `120` for two minutes.
 
 ### Core Keys
 `CORE_API_KEY` specifies the **very secret key** for  which need to access the entire user database etc.
@@ -176,6 +145,29 @@ You can also just add the *beginning* of an API address, like `12.123.` (without
 ### Other
 `KEYGEN_INFIX` can be almost any string (avoid spaces or special characters) - this string will be put in the middle of every NovaAI API key which is generated. This is useful for identifying the source of the key using e.g. RegEx.
 
+## Proxy Lists
+To use proxy lists, navigate to `api/secret/proxies/` and create the following files:
+- `http.txt`
+- `socks4.txt`
+- `socks5.txt`
+
+Then, paste your proxies in the following format:
+
+```
+[username:password@]host:port
+```
+
+e.g.
+
+```
+1.2.3.4:8080
+user:pass@127.0.0.1:1337
+```
+
+You can use comments just like in Python.
+
+**Important:** to use the proxy lists, you need to change the `USE_PROXY_LIST` environment variable to `True`!
+
 ## Run
 > **Warning:** read the according section for production usage!
 
@@ -194,20 +186,17 @@ python run 1337
 ```
 
 ## Adding a provider
-To be documented!]
 
-## Run tests
+## Test if it works
-Make sure the API server is running on the port you specified and run:
 `python checks`
 
-## Default Ports
+## Ports
 ```yml
-2332: Developement
+2332: Developement (default)
 2333: Production
 ```
 
 ## Production
-
 Make sure your server is secure and up to date.
 Check everything.
 

admintools/pruner.py

@@ -82,5 +82,8 @@
 
 # # ====================================================================================
 
+# def prune():
+#     # gets all users from 
+
 # if __name__ == '__main__':
 #     launch()

api/after_request.py

@@ -1,4 +1,4 @@
-from db import logs, stats, users, key_validation
+from db import logs, stats, users
 from helpers import network
 
 async def after_request(
@@ -23,8 +23,6 @@ async def after_request(
     await stats.manager.add_ip_address(ip_address)
     await stats.manager.add_path(path)
     await stats.manager.add_target(target_request['url'])
-    await key_validation.remove_rated_keys()
-    await key_validation.cache_all_keys()
 
     if is_chat:
         await stats.manager.add_model(model)

api/backup_manager/main.py

@@ -33,7 +33,7 @@ async def make_backup(output_dir: str):
             os.mkdir(f'{output_dir}/{database}')
 
         for collection in databases[database]:
-            print(f'Initiated database backup for {database}/{collection}')
+            print(f'Making backup for {database}/{collection}')
             await make_backup_for_collection(database, collection, output_dir)
 
 async def make_backup_for_collection(database, collection, output_dir):
@@ -52,4 +52,4 @@ if __name__ == '__main__':
         exit(1)
 
     output_dir = argv[1]
-    asyncio.run(main(output_dir))
+    asyncio.run(main(output_dir))

api/cache/crypto_prices.json

@@ -0,0 +1 @@
+{"LTC": 64.665, "_last_updated": 1695334741.4905503, "BTC": 26583.485, "MATIC": 0.52075, "XMR": 146.46058828041404, "ADA": 0.2455, "USDT": 1.000005, "ETH": 1586.115, "USD": 1.0, "EUR": 1.0662838016640013}

api/core.py

@@ -194,6 +194,4 @@ async def get_finances(incoming_request: fastapi.Request):
             amount_in_usd = await get_crypto_price(currency) * amount
             transactions[table][transactions[table].index(transaction)]['amount_usd'] = amount_in_usd
 
-    transactions['timestamp'] = time.time()
-
     return transactions

api/db/key_validation.py

@@ -1,84 +0,0 @@
-import os
-import time
-import asyncio
-import json
-
-from dotenv import load_dotenv
-from motor.motor_asyncio import AsyncIOMotorClient
-
-load_dotenv()
-
-MONGO_URI = os.getenv('MONGO_URI')
-
-async def log_rated_key(key: str) -> None:
-    """Logs a key that has been rate limited to the database."""
-
-    client = AsyncIOMotorClient(MONGO_URI)
-
-    scheme = {
-        'key': key,
-        'timestamp_added': int(time.time())
-    }
-
-    collection = client['Liabilities']['rate-limited-keys']
-    await collection.insert_one(scheme)
-
-
-async def key_is_rated(key: str) -> bool:
-    """Checks if a key is rate limited."""
-
-    client = AsyncIOMotorClient(MONGO_URI)
-    collection = client['Liabilities']['rate-limited-keys']
-
-    query = {
-        'key': key
-    }
-
-    result = await collection.find_one(query)
-    return result is not None
-
-
-async def cached_key_is_rated(key: str) -> bool:
-    path = os.path.join(os.getcwd(), 'cache', 'rate_limited_keys.json')
-
-    with open(path, 'r', encoding='utf8') as file:
-        keys = json.load(file)
-
-    return key in keys
-
-async def remove_rated_keys() -> None:
-    """Removes all keys that have been rate limited for more than a day."""
-
-    client = AsyncIOMotorClient(MONGO_URI)
-    collection = client['Liabilities']['rate-limited-keys']
-
-    keys = await collection.find().to_list(length=None)
-
-    marked_for_removal = []
-    for key in keys:
-        if int(time.time()) - key['timestamp_added'] > 86400:
-            marked_for_removal.append(key['_id'])
-
-    query = {
-        '_id': {
-            '$in': marked_for_removal
-        }
-    }
-
-    await collection.delete_many(query)
-
-async def cache_all_keys() -> None:
-    """Clones all keys from the database to the cache."""
-
-    client = AsyncIOMotorClient(MONGO_URI)
-    collection = client['Liabilities']['rate-limited-keys']
-
-    keys = await collection.find().to_list(length=None)
-    keys = [key['key'] for key in keys]
-
-    path = os.path.join(os.getcwd(), 'cache', 'rate_limited_keys.json')
-    with open(path, 'w') as file:
-        json.dump(keys, file)
-
-if __name__ == "__main__":
-    asyncio.run(remove_rated_keys())

api/handler.py

@@ -28,15 +28,15 @@ with open('config/config.yml', encoding='utf8') as f:
 moderation_debug_key_key = os.getenv('MODERATION_DEBUG_KEY')
 
 async def handle(incoming_request: fastapi.Request):
-    """Transfer a streaming response 
+    """
+    ### Transfer a streaming response 
     Takes the request from the incoming request to the target endpoint.
     Checks method, token amount, auth and cost along with if request is NSFW.
     """
-
+    path = incoming_request.url.path.replace('v1/v1', 'v1').replace('//', '/')
-    path = incoming_request.url.path
-    path = path.replace('/v1/v1', '/v1')
 
     ip_address = await network.get_ip(incoming_request)
+    print(f'[bold green]>{ip_address}[/bold green]')
 
     if '/models' in path:
         return fastapi.responses.JSONResponse(content=models_list)
@@ -65,17 +65,12 @@ async def handle(incoming_request: fastapi.Request):
         return await errors.error(418, 'Invalid or inactive NovaAI API key!', 'Create a new NovaOSS API key or reactivate your account.')
 
     if user.get('auth', {}).get('discord'):
-        print(f'[bold green]>{ip_address} ({user["auth"]["discord"]})[/bold green]')
+        print(f'[bold green]>Discord[/bold green] {user["auth"]["discord"]}')
 
     ban_reason = user['status']['ban_reason']
     if ban_reason:
         return await errors.error(403, f'Your NovaAI account has been banned. Reason: \'{ban_reason}\'.', 'Contact the staff for an appeal.')
 
-    # Checking for enterprise status
-    enterprise_keys = os.environ.get('NO_RATELIMIT_KEYS')
-    if '/enterprise' in path and user.get('api_key') not in enterprise_keys:
-        return await errors.error(403, 'Enterprise API is not available.', 'Contact the staff for an upgrade.')
-
     if 'account/credits' in path:
         return fastapi.responses.JSONResponse({'credits': user['credits']})
 

api/helpers/network.py

@@ -1,7 +1,7 @@
 import os
 import time
+
 from dotenv import load_dotenv
-from slowapi.util import get_remote_address
 
 load_dotenv()
 
@@ -24,10 +24,22 @@ async def get_ip(request) -> str:
 
 def get_ratelimit_key(request) -> str:
     """Get the IP address of the incoming request."""
-    custom = os.environ('NO_RATELIMIT_IPS')
-    ip = get_remote_address(request)
 
-    if ip in custom:
+    xff = None
-        return f'enterprise_{ip}'
+    if request.headers.get('x-forwarded-for'):
+        xff, *_ = request.headers['x-forwarded-for'].split(', ')
 
-    return ip
+    possible_ips = [
+        xff,
+        request.headers.get('cf-connecting-ip'),
+        request.client.host
+    ]
+
+    detected_ip = next((i for i in possible_ips if i), None)
+
+    for whitelisted_ip in os.getenv('NO_RATELIMIT_IPS', '').split():
+        if whitelisted_ip in detected_ip:
+            custom_key = f'whitelisted-{time.time()}'
+            return custom_key
+
+    return detected_ip

api/load_balancing.py

@@ -31,6 +31,7 @@ async def balance_chat_request(payload: dict) -> dict:
 
     provider = random.choice(providers_available)
     target = await provider.chat_completion(**payload)
+
     module_name = await _get_module_name(provider)
     target['module'] = module_name
 

api/main.py

@@ -11,9 +11,10 @@ from bson.objectid import ObjectId
 from slowapi.errors import RateLimitExceeded
 from slowapi.middleware import SlowAPIMiddleware
 from fastapi.middleware.cors import CORSMiddleware
-from slowapi.util import get_remote_address
 from slowapi import Limiter, _rate_limit_exceeded_handler
 
+from helpers import network
+
 import core
 import handler
 
@@ -33,13 +34,11 @@ app.include_router(core.router)
 
 limiter = Limiter(
     swallow_errors=True,
-    key_func=get_remote_address,
+    key_func=network.get_ratelimit_key, default_limits=[
-    default_limits=[
+    '2/second',
-    '1/second',
     '20/minute',
     '300/hour'
 ])
-
 app.state.limiter = limiter
 app.add_exception_handler(RateLimitExceeded, _rate_limit_exceeded_handler)
 app.add_middleware(SlowAPIMiddleware)
@@ -67,10 +66,4 @@ async def root():
 @app.route('/v1/{path:path}', methods=['GET', 'POST', 'PUT', 'DELETE', 'PATCH'])
 async def v1_handler(request: fastapi.Request):
     res = await handler.handle(incoming_request=request)
-    return res
+    return res
-
-@limiter.limit('100/minute', '1000/hour') 
-@app.route('/enterprise/v1/{path:path}', methods=['GET', 'POST', 'PUT', 'DELETE', 'PATCH'])
-async def enterprise_handler(request: fastapi.Request):
-    res = await handler.handle(incoming_request=request)
-    return res

api/provider_auth.py

@@ -49,7 +49,7 @@ async def invalidate_key(provider_and_key: str) -> None:
     with open(f'secret/{provider}.invalid.txt', 'a', encoding='utf8') as f:
         f.write(key + '\n')
 
-    # await invalidation_webhook(provider_and_key)
+    await invalidation_webhook(provider_and_key)
 
 if __name__ == '__main__':
     asyncio.run(invalidate_key('closed>demo-...'))

api/proxies.py

@@ -37,20 +37,15 @@ class Proxy:
             url = url.split('://')[1]
 
             if '@' in url:
-                username = url.split('@')[0].split(':')[0]
+                username = url.split('@')[1].split(':')[0]
-                password = url.split('@')[0].split(':')[1]
+                password = url.split('@')[1].split(':')[1]
 
-            host_or_ip = url.split('@')[-1].split(':')[0]
+            host_or_ip = url.split(':')[0]
-            port = int(url.split('@')[-1].split(':')[1])
+            port = url.split(':')[1]
 
         self.proxy_type = proxy_type
         self.host_or_ip = host_or_ip
-
+        self.ip_address = socket.gethostbyname(self.host_or_ip) # get ip address from host
-        try:
-            self.ip_address = socket.gethostbyname(self.host_or_ip) # get ip address from host
-        except socket.gaierror:
-            self.ip_address = self.host_or_ip
-
         self.host = self.host_or_ip
         self.port = port
         self.username = username
@@ -83,7 +78,7 @@ class Proxy:
 
         return aiohttp_socks.ProxyConnector(
             proxy_type=proxy_types[self.proxy_type],
-            host=self.host,
+            host=self.ip_address,
             port=self.port,
             rdns=False, 
             username=self.username,
@@ -94,15 +89,16 @@ class Proxy:
 
 proxies_in_files = []
 
-for proxy_type in ['http', 'socks4', 'socks5']:
+try:
-    try:
+    for proxy_type in ['http', 'socks4', 'socks5']:
         with open(f'secret/proxies/{proxy_type}.txt') as f:
             for line in f:
                 clean_line = line.split('#', 1)[0].strip()
                 if clean_line:
                     proxies_in_files.append(f'{proxy_type}://{clean_line}')
-    except FileNotFoundError:
+
-        pass
+except FileNotFoundError:
+    pass
 
 ## Manages the proxy list
 
@@ -118,7 +114,7 @@ def get_proxy() -> Proxy:
     ### Returns a Proxy object
     The proxy is either from the proxy list or from the environment variables.
     """
-
+    
     if USE_PROXY_LIST:
         return ProxyLists().get_random
 
@@ -129,6 +125,3 @@ def get_proxy() -> Proxy:
         username=os.getenv('PROXY_USER'),
         password=os.getenv('PROXY_PASS')
     )
-
-if __name__ == '__main__':
-    print(get_proxy().url)

api/responder.py

@@ -2,7 +2,9 @@
 
 import os
 import json
-import random
+import yaml
+import dhooks
+import asyncio
 import aiohttp
 import starlette
 
@@ -14,9 +16,7 @@ import provider_auth
 import after_request
 import load_balancing
 
-from helpers import errors
+from helpers import network, chat, errors
-
-from db import key_validation
 
 load_dotenv()
 
@@ -44,15 +44,21 @@ async def respond(
     json_response = {}
 
     headers = {
-        'Content-Type': 'application/json'
+        'Content-Type': 'application/json',
+        'User-Agent': 'axios/0.21.1',
     }
 
     for _ in range(10):
         # Load balancing: randomly selecting a suitable provider
+        # If the request is a chat completion, then we need to load balance between chat providers
+        # If the request is an organic request, then we need to load balance between organic providers
         try:
             if is_chat:
                 target_request = await load_balancing.balance_chat_request(payload)
             else:
+                
+                # In this case we are doing a organic request. "organic" means that it's not using a reverse engineered front-end, but rather ClosedAI's API directly
+                # churchless.tech is an example of an organic provider, because it redirects the request to ClosedAI.
                 target_request = await load_balancing.balance_organic_request({
                     'method': incoming_request.method,
                     'path': path,
@@ -61,7 +67,10 @@ async def respond(
                     'cookies': incoming_request.cookies
                 })
         except ValueError as exc:
-            yield await errors.yield_error(500, f'Sorry, the API has no active API keys for {model}.', 'Please use a different model.')
+            if model in ['gpt-3.5-turbo', 'gpt-4', 'gpt-4-32k']:
+                webhook = dhooks.Webhook(os.environ['DISCORD_WEBHOOK__API_ISSUE'])
+                webhook.send(content=f'API Issue: **`{exc}`**\nhttps://i.imgflip.com/7uv122.jpg')
+                yield await errors.yield_error(500, 'Sorry, the API has no working keys anymore.', 'The admins have been messaged automatically.')
             return
 
         target_request['headers'].update(target_request.get('headers', {}))
@@ -82,56 +91,34 @@ async def respond(
                     cookies=target_request.get('cookies'),
                     ssl=False,
                     timeout=aiohttp.ClientTimeout(
-                        connect=1.0,
+                        connect=0.3,
                         total=float(os.getenv('TRANSFER_TIMEOUT', '500'))
                     ),
                 ) as response:
                     is_stream = response.content_type == 'text/event-stream'
 
                     if response.status == 429:
-                        await key_validation.log_rated_key(target_request.get('provider_auth'))
                         continue
 
                     if response.content_type == 'application/json':
                         data = await response.json()
 
-                        error = data.get('error')
-                        match error:
-                            case None:
-                                pass
-
-                            case _:
-                                key = target_request.get('provider_auth')
-
-                                match error.get('code'):
-                                    case 'invalid_api_key':
-                                        await key_validation.log_rated_key(key)
-                                        print('[!] invalid key', key)
-
-                                    case _:
-                                        print('[!] unknown error with key: ', key, error)
-
                         if 'method_not_supported' in str(data):
                             await errors.error(500, 'Sorry, this endpoint does not support this method.', data['error']['message'])
 
                         if 'invalid_api_key' in str(data) or 'account_deactivated' in str(data):
+                            print('[!] invalid api key', target_request.get('provider_auth'))
                             await provider_auth.invalidate_key(target_request.get('provider_auth'))
                             continue
 
                         if response.ok:
                             json_response = data
 
-                        else:
-                            print('[!] error', data)
-                            continue
-
-
                     if is_stream:
                         try:
                             response.raise_for_status()
                         except Exception as exc:
                             if 'Too Many Requests' in str(exc):
-                                print('[!] too many requests')
                                 continue
 
                         async for chunk in response.content.iter_any():
@@ -141,18 +128,20 @@ async def respond(
                     break
 
             except Exception as exc:
-                if 'too many requests' in str(exc):
-                    await key_validation.log_rated_key(key)
-
                 continue
 
+            if (not json_response) and is_chat:
+                print('[!] chat response is empty')
+                continue
     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.')
+        yield await errors.yield_error(500, 'Sorry, the provider is not responding. We\'re possibly getting rate-limited.', 'Please try again later.')
         return
 
     if (not is_stream) and json_response:
         yield json.dumps(json_response)
 
+    print(f'[+] {path} -> {model or ""}')
+
     await after_request.after_request(
         incoming_request=incoming_request,
         target_request=target_request,

checks/client.py

@@ -164,7 +164,7 @@ async def test_function_calling():
             url=f'{api_endpoint}/chat/completions',
             headers=HEADERS,
             json=json_data,
-            timeout=15,
+            timeout=10,
         )
         response.raise_for_status()
 
@@ -208,8 +208,8 @@ async def demo():
         else:
             raise ConnectionError('API Server is not running.')
 
-        # print('[lightblue]Checking if function calling works...')
+        print('[lightblue]Checking if function calling works...')
-        # print(await test_function_calling())
+        print(await test_function_calling())
 
         print('Checking non-streamed chat completions...')
         print(await test_chat_non_stream_gpt4())
@@ -220,8 +220,8 @@ async def demo():
         # print('[lightblue]Checking if  image generation works...')
         # print(await test_image_generation())
 
-        # print('Checking the models endpoint...')
+        print('Checking the models endpoint...')
-        # print(await test_models())
+        print(await test_models())
 
     except Exception as exc:
         print('[red]Error: ' + str(exc))