0.3.5.4

29 files changed, 494 insertions, 328 deletions

diff --git a/etc/examples/api.py b/etc/examples/api.py
index f8f5d5ec..2485bade 100644
--- a/etc/examples/api.py
+++ b/etc/examples/api.py
@@ -1,13 +1,17 @@
 import requests
 import json
+import uuid
+
 url = "http://localhost:1337/v1/chat/completions"
+conversation_id = str(uuid.uuid4())
 body = {
     "model": "",
-    "provider": "",
+    "provider": "Copilot",
     "stream": True,
     "messages": [
-        {"role": "user", "content": "What can you do? Who are you?"}
-    ]
+        {"role": "user", "content": "Hello, i am Heiner. How are you?"}
+    ],
+    "conversation_id": conversation_id
 }
 response = requests.post(url, json=body, stream=True)
 response.raise_for_status()
@@ -21,4 +25,27 @@ for line in response.iter_lines():
             print(json_data.get("choices", [{"delta": {}}])[0]["delta"].get("content", ""), end="")
         except json.JSONDecodeError:
             pass
-print()
\ No newline at end of file
+print()
+print()
+print()
+body = {
+    "model": "",
+    "provider": "Copilot",
+    "stream": True,
+    "messages": [
+        {"role": "user", "content": "Tell me somethings about my name"}
+    ],
+    "conversation_id": conversation_id
+}
+response = requests.post(url, json=body, stream=True)
+response.raise_for_status()
+for line in response.iter_lines():
+    if line.startswith(b"data: "):
+        try:
+            json_data = json.loads(line[6:])
+            if json_data.get("error"):
+                print(json_data)
+                break
+            print(json_data.get("choices", [{"delta": {}}])[0]["delta"].get("content", ""), end="")
+        except json.JSONDecodeError:
+            pass
\ No newline at end of file
diff --git a/g4f/Provider/Airforce.py b/g4f/Provider/Airforce.py
index 54bb543b..f5bcfefa 100644
--- a/g4f/Provider/Airforce.py
+++ b/g4f/Provider/Airforce.py
@@ -20,7 +20,7 @@ class Airforce(AsyncGeneratorProvider, ProviderModelMixin):
     working = True
     supports_system_message = True
     supports_message_history = True
-    
+
     @classmethod
     def fetch_completions_models(cls):
         response = requests.get('https://api.airforce/models', verify=False)
@@ -34,19 +34,20 @@ class Airforce(AsyncGeneratorProvider, ProviderModelMixin):
         response.raise_for_status()
         return response.json()
 
-    completions_models = fetch_completions_models.__func__(None)
-    imagine_models = fetch_imagine_models.__func__(None)
-
     default_model = "gpt-4o-mini"
     default_image_model = "flux"
     additional_models_imagine = ["stable-diffusion-xl-base", "stable-diffusion-xl-lightning", "Flux-1.1-Pro"]
-    text_models = completions_models
-    image_models = [*imagine_models, *additional_models_imagine]
-    models = [
-        *text_models,
-        *image_models,
-    ]
-    
+
+    @classmethod
+    def get_models(cls):
+        if not cls.models:
+            cls.image_models = [*cls.fetch_imagine_models(), *cls.additional_models_imagine]
+            cls.models = [
+                *cls.fetch_completions_models(),
+                *cls.image_models
+            ]
+        return cls.models
+
     model_aliases = {        
         ### completions ###
         # openchat
@@ -100,7 +101,6 @@ class Airforce(AsyncGeneratorProvider, ProviderModelMixin):
         **kwargs
     ) -> AsyncResult:
         model = cls.get_model(model)
-
         if model in cls.image_models:
             return cls._generate_image(model, messages, proxy, seed, size)
         else:
diff --git a/g4f/Provider/Blackbox.py b/g4f/Provider/Blackbox.py
index b259b4aa..41905537 100644
--- a/g4f/Provider/Blackbox.py
+++ b/g4f/Provider/Blackbox.py
@@ -20,15 +20,15 @@ class Blackbox(AsyncGeneratorProvider, ProviderModelMixin):
     supports_system_message = True
     supports_message_history = True
     _last_validated_value = None
-    
+
     default_model = 'blackboxai'
     default_vision_model = default_model
     default_image_model = 'Image Generation' 
     image_models = ['Image Generation', 'repomap']
     vision_models = [default_model, 'gpt-4o', 'gemini-pro', 'gemini-1.5-flash', 'llama-3.1-8b', 'llama-3.1-70b', 'llama-3.1-405b']
-   
+
     userSelectedModel = ['gpt-4o', 'gemini-pro', 'claude-sonnet-3.5', 'blackboxai-pro']
-    
+
     agentMode = {
         'Image Generation': {'mode': True, 'id': "ImageGenerationLV45LJp", 'name': "Image Generation"},
     }
@@ -77,22 +77,21 @@ class Blackbox(AsyncGeneratorProvider, ProviderModelMixin):
     }
     
     additional_prefixes = {
-    'gpt-4o': '@gpt-4o',
-    'gemini-pro': '@gemini-pro',
-    'claude-sonnet-3.5': '@claude-sonnet'
-		}
+        'gpt-4o': '@gpt-4o',
+        'gemini-pro': '@gemini-pro',
+        'claude-sonnet-3.5': '@claude-sonnet'
+    }
     
     model_prefixes = {
-    **{mode: f"@{value['id']}" for mode, value in trendingAgentMode.items() 
-       if mode not in ["gemini-1.5-flash", "llama-3.1-8b", "llama-3.1-70b", "llama-3.1-405b", "repomap"]},
-    **additional_prefixes
-        }
+        **{
+            mode: f"@{value['id']}" for mode, value in trendingAgentMode.items() 
+            if mode not in ["gemini-1.5-flash", "llama-3.1-8b", "llama-3.1-70b", "llama-3.1-405b", "repomap"]
+        },
+        **additional_prefixes
+    }
 
-    
     models = list(dict.fromkeys([default_model, *userSelectedModel, *list(agentMode.keys()), *list(trendingAgentMode.keys())]))
 
-
-    
     model_aliases = {
         "gemini-flash": "gemini-1.5-flash",
         "claude-3.5-sonnet": "claude-sonnet-3.5",
@@ -131,12 +130,11 @@ class Blackbox(AsyncGeneratorProvider, ProviderModelMixin):
 
         return cls._last_validated_value
 
-
     @staticmethod
     def generate_id(length=7):
         characters = string.ascii_letters + string.digits
         return ''.join(random.choice(characters) for _ in range(length))
-        
+
     @classmethod
     def add_prefix_to_messages(cls, messages: Messages, model: str) -> Messages:
         prefix = cls.model_prefixes.get(model, "")
@@ -157,6 +155,7 @@ class Blackbox(AsyncGeneratorProvider, ProviderModelMixin):
         cls,
         model: str,
         messages: Messages,
+        prompt: str = None,
         proxy: str = None,
         web_search: bool = False,
         image: ImageType = None,
@@ -191,7 +190,7 @@ class Blackbox(AsyncGeneratorProvider, ProviderModelMixin):
             'sec-fetch-site': 'same-origin',
             'user-agent': 'Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/130.0.0.0 Safari/537.36'
         }
-        
+
         data = {
             "messages": messages,
             "id": message_id,
@@ -221,26 +220,25 @@ class Blackbox(AsyncGeneratorProvider, ProviderModelMixin):
             async with session.post(cls.api_endpoint, json=data, proxy=proxy) as response:
                 response.raise_for_status()
                 response_text = await response.text()
-                
+
                 if model in cls.image_models:
                     image_matches = re.findall(r'!\[.*?\]\((https?://[^\)]+)\)', response_text)
                     if image_matches:
                         image_url = image_matches[0]
-                        image_response = ImageResponse(images=[image_url], alt="Generated Image")
-                        yield image_response
+                        yield ImageResponse(image_url, prompt)
                         return
 
                 response_text = re.sub(r'Generated by BLACKBOX.AI, try unlimited chat https://www.blackbox.ai', '', response_text, flags=re.DOTALL)
-                
+
                 json_match = re.search(r'\$~~~\$(.*?)\$~~~\$', response_text, re.DOTALL)
                 if json_match:
                     search_results = json.loads(json_match.group(1))
                     answer = response_text.split('$~~~$')[-1].strip()
-                    
+
                     formatted_response = f"{answer}\n\n**Source:**"
                     for i, result in enumerate(search_results, 1):
                         formatted_response += f"\n{i}. {result['title']}: {result['link']}"
-                    
+
                     yield formatted_response
                 else:
                     yield response_text.strip()
diff --git a/g4f/Provider/Copilot.py b/g4f/Provider/Copilot.py
index e8eea0a5..e10a55e8 100644
--- a/g4f/Provider/Copilot.py
+++ b/g4f/Provider/Copilot.py
@@ -57,6 +57,7 @@ class Copilot(AbstractProvider):
         image: ImageType = None,
         conversation: Conversation = None,
         return_conversation: bool = False,
+        web_search: bool = True,
         **kwargs
     ) -> CreateResult:
         if not has_curl_cffi:
@@ -124,12 +125,14 @@ class Copilot(AbstractProvider):
             is_started = False
             msg = None
             image_prompt: str = None
+            last_msg = None
             while True:
                 try:
                     msg = wss.recv()[0]
                     msg = json.loads(msg)
                 except:
                     break
+                last_msg = msg
                 if msg.get("event") == "appendText":
                     is_started = True
                     yield msg.get("text")
@@ -139,8 +142,12 @@ class Copilot(AbstractProvider):
                     yield ImageResponse(msg.get("url"), image_prompt, {"preview": msg.get("thumbnailUrl")})
                 elif msg.get("event") == "done":
                     break
+                elif msg.get("event") == "error":
+                    raise RuntimeError(f"Error: {msg}")
+                elif msg.get("event") not in ["received", "startMessage", "citation", "partCompleted"]:
+                    debug.log(f"Copilot Message: {msg}")
             if not is_started:
-                raise RuntimeError(f"Last message: {msg}")
+                raise RuntimeError(f"Invalid response: {last_msg}")
 
     @classmethod
     async def get_access_token_and_cookies(cls, proxy: str = None):
diff --git a/g4f/Provider/PollinationsAI.py b/g4f/Provider/PollinationsAI.py
index 57597bf1..a30f896d 100644
--- a/g4f/Provider/PollinationsAI.py
+++ b/g4f/Provider/PollinationsAI.py
@@ -3,7 +3,6 @@ from __future__ import annotations
 from urllib.parse import quote
 import random
 import requests
-from sys import maxsize
 from aiohttp import ClientSession
 
 from ..typing import AsyncResult, Messages
@@ -40,6 +39,7 @@ class PollinationsAI(OpenaiAPI):
         cls,
         model: str,
         messages: Messages,
+        prompt: str = None,
         api_base: str = "https://text.pollinations.ai/openai",
         api_key: str = None,
         proxy: str = None,
@@ -49,9 +49,10 @@ class PollinationsAI(OpenaiAPI):
         if model:
             model = cls.get_model(model)
         if model in cls.image_models:
-            prompt = messages[-1]["content"]
+            if prompt is None:
+                prompt = messages[-1]["content"]
             if seed is None:
-                seed = random.randint(0, maxsize)
+                seed = random.randint(0, 100000)
             image = f"https://image.pollinations.ai/prompt/{quote(prompt)}?width=1024&height=1024&seed={int(seed)}&nofeed=true&nologo=true&model={quote(model)}"
             yield ImageResponse(image, prompt)
             return
diff --git a/g4f/Provider/ReplicateHome.py b/g4f/Provider/ReplicateHome.py
index a7fc9b54..00de09e0 100644
--- a/g4f/Provider/ReplicateHome.py
+++ b/g4f/Provider/ReplicateHome.py
@@ -6,6 +6,8 @@ from aiohttp import ClientSession, ContentTypeError
 
 from ..typing import AsyncResult, Messages
 from .base_provider import AsyncGeneratorProvider, ProviderModelMixin
+from ..requests.aiohttp import get_connector
+from ..requests.raise_for_status import raise_for_status
 from .helper import format_prompt
 from ..image import ImageResponse
 
@@ -32,10 +34,8 @@ class ReplicateHome(AsyncGeneratorProvider, ProviderModelMixin):
         'yorickvp/llava-13b',
     ]
 
-
-
     models = text_models + image_models
-    
+
     model_aliases = {
         # image_models
         "sd-3": "stability-ai/stable-diffusion-3",
@@ -56,23 +56,14 @@ class ReplicateHome(AsyncGeneratorProvider, ProviderModelMixin):
         # text_models
         "google-deepmind/gemma-2b-it": "dff94eaf770e1fc211e425a50b51baa8e4cac6c39ef074681f9e39d778773626",
         "yorickvp/llava-13b": "80537f9eead1a5bfa72d5ac6ea6414379be41d4d4f6679fd776e9535d1eb58bb",
-        
     }
 
     @classmethod
-    def get_model(cls, model: str) -> str:
-        if model in cls.models:
-            return model
-        elif model in cls.model_aliases:
-            return cls.model_aliases[model]
-        else:
-            return cls.default_model
-
-    @classmethod
     async def create_async_generator(
         cls,
         model: str,
         messages: Messages,
+        prompt: str = None,
         proxy: str = None,
         **kwargs
     ) -> AsyncResult:
@@ -96,29 +87,30 @@ class ReplicateHome(AsyncGeneratorProvider, ProviderModelMixin):
             "user-agent": "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/128.0.0.0 Safari/537.36"
         }
         
-        async with ClientSession(headers=headers) as session:
-            if model in cls.image_models:
-                prompt = messages[-1]['content'] if messages else ""
-            else:
-                prompt = format_prompt(messages)
-            
+        async with ClientSession(headers=headers, connector=get_connector(proxy=proxy)) as session:
+            if prompt is None:
+                if model in cls.image_models:
+                    prompt = messages[-1]['content']
+                else:
+                    prompt = format_prompt(messages)
+
             data = {
                 "model": model,
                 "version": cls.model_versions[model],
                 "input": {"prompt": prompt},
             }
-            
-            async with session.post(cls.api_endpoint, json=data, proxy=proxy) as response:
-                response.raise_for_status()
+
+            async with session.post(cls.api_endpoint, json=data) as response:
+                await raise_for_status(response)
                 result = await response.json()
                 prediction_id = result['id']
-            
+
             poll_url = f"https://homepage.replicate.com/api/poll?id={prediction_id}"
             max_attempts = 30
             delay = 5
             for _ in range(max_attempts):
-                async with session.get(poll_url, proxy=proxy) as response:
-                    response.raise_for_status()
+                async with session.get(poll_url) as response:
+                    await raise_for_status(response)
                     try:
                         result = await response.json()
                     except ContentTypeError:
@@ -131,7 +123,7 @@ class ReplicateHome(AsyncGeneratorProvider, ProviderModelMixin):
                     if result['status'] == 'succeeded':
                         if model in cls.image_models:
                             image_url = result['output'][0]
-                            yield ImageResponse(image_url, "Generated image")
+                            yield ImageResponse(image_url, prompt)
                             return
                         else:
                             for chunk in result['output']:
@@ -140,6 +132,6 @@ class ReplicateHome(AsyncGeneratorProvider, ProviderModelMixin):
                     elif result['status'] == 'failed':
                         raise Exception(f"Prediction failed: {result.get('error')}")
                 await asyncio.sleep(delay)
-            
+
             if result['status'] != 'succeeded':
                 raise Exception("Prediction timed out")
diff --git a/g4f/Provider/RubiksAI.py b/g4f/Provider/RubiksAI.py
index c06e6c3d..816ea60c 100644
--- a/g4f/Provider/RubiksAI.py
+++ b/g4f/Provider/RubiksAI.py
@@ -9,7 +9,7 @@ from urllib.parse import urlencode
 from aiohttp import ClientSession
 
 from ..typing import AsyncResult, Messages
-from .base_provider import AsyncGeneratorProvider, ProviderModelMixin
+from .base_provider import AsyncGeneratorProvider, ProviderModelMixin, Sources
 from ..requests.raise_for_status import raise_for_status
 
 class RubiksAI(AsyncGeneratorProvider, ProviderModelMixin):
@@ -23,7 +23,6 @@ class RubiksAI(AsyncGeneratorProvider, ProviderModelMixin):
 
     default_model = 'gpt-4o-mini'
     models = [default_model, 'gpt-4o', 'o1-mini', 'claude-3.5-sonnet', 'grok-beta', 'gemini-1.5-pro', 'nova-pro']
-
     model_aliases = {
         "llama-3.1-70b": "llama-3.1-70b-versatile",
     }
@@ -118,7 +117,7 @@ class RubiksAI(AsyncGeneratorProvider, ProviderModelMixin):
 
                     if 'url' in json_data and 'title' in json_data:
                         if web_search:
-                            sources.append({'title': json_data['title'], 'url': json_data['url']})
+                            sources.append(json_data)
 
                     elif 'choices' in json_data:
                         for choice in json_data['choices']:
@@ -128,5 +127,4 @@ class RubiksAI(AsyncGeneratorProvider, ProviderModelMixin):
                                 yield content
 
                 if web_search and sources:
-                    sources_text = '\n'.join([f"{i+1}. [{s['title']}]: {s['url']}" for i, s in enumerate(sources)])
-                    yield f"\n\n**Source:**\n{sources_text}"
\ No newline at end of file
+                    yield Sources(sources)
\ No newline at end of file
diff --git a/g4f/Provider/base_provider.py b/g4f/Provider/base_provider.py
index 667f6964..c0d8edf0 100644
--- a/g4f/Provider/base_provider.py
+++ b/g4f/Provider/base_provider.py
@@ -1,4 +1,4 @@
 from ..providers.base_provider import *
-from ..providers.types import FinishReason, Streaming
-from ..providers.conversation import BaseConversation
+from ..providers.types import Streaming
+from ..providers.response import BaseConversation, Sources, FinishReason
 from .helper import get_cookies, format_prompt
\ No newline at end of file
diff --git a/g4f/Provider/bing/conversation.py b/g4f/Provider/bing/conversation.py
index b5c237f9..43bcbb4d 100644
--- a/g4f/Provider/bing/conversation.py
+++ b/g4f/Provider/bing/conversation.py
@@ -2,7 +2,7 @@ from __future__ import annotations
 
 from ...requests import StreamSession, raise_for_status
 from ...errors import RateLimitError
-from ...providers.conversation import BaseConversation
+from ...providers.response import BaseConversation
 
 class Conversation(BaseConversation):
     """
diff --git a/g4f/Provider/needs_auth/BingCreateImages.py b/g4f/Provider/needs_auth/BingCreateImages.py
index 80984d40..b95a78c3 100644
--- a/g4f/Provider/needs_auth/BingCreateImages.py
+++ b/g4f/Provider/needs_auth/BingCreateImages.py
@@ -28,13 +28,14 @@ class BingCreateImages(AsyncGeneratorProvider, ProviderModelMixin):
         cls,
         model: str,
         messages: Messages,
+        prompt: str = None,
         api_key: str = None,
         cookies: Cookies = None,
         proxy: str = None,
         **kwargs
     ) -> AsyncResult:
         session = BingCreateImages(cookies, proxy, api_key)
-        yield await session.generate(messages[-1]["content"])
+        yield await session.generate(messages[-1]["content"] if prompt is None else prompt)
 
     async def generate(self, prompt: str) -> ImageResponse:
         """
diff --git a/g4f/Provider/needs_auth/DeepInfraImage.py b/g4f/Provider/needs_auth/DeepInfraImage.py
index 24df04e3..44790561 100644
--- a/g4f/Provider/needs_auth/DeepInfraImage.py
+++ b/g4f/Provider/needs_auth/DeepInfraImage.py
@@ -29,9 +29,10 @@ class DeepInfraImage(AsyncGeneratorProvider, ProviderModelMixin):
         cls,
         model: str,
         messages: Messages,
+        prompt: str = None,
         **kwargs
     ) -> AsyncResult:
-        yield await cls.create_async(messages[-1]["content"], model, **kwargs)
+        yield await cls.create_async(messages[-1]["content"] if prompt is None else prompt, model, **kwargs)
 
     @classmethod
     async def create_async(
diff --git a/g4f/Provider/needs_auth/OpenaiChat.py b/g4f/Provider/needs_auth/OpenaiChat.py
index 97515ec4..587c0a23 100644
--- a/g4f/Provider/needs_auth/OpenaiChat.py
+++ b/g4f/Provider/needs_auth/OpenaiChat.py
@@ -28,7 +28,7 @@ from ...requests.raise_for_status import raise_for_status
 from ...requests.aiohttp import StreamSession
 from ...image import ImageResponse, ImageRequest, to_image, to_bytes, is_accepted_format
 from ...errors import MissingAuthError, ResponseError
-from ...providers.conversation import BaseConversation
+from ...providers.response import BaseConversation
 from ..helper import format_cookies
 from ..openai.har_file import get_request_config, NoValidHarFileError
 from ..openai.har_file import RequestConfig, arkReq, arkose_url, start_url, conversation_url, backend_url, backend_anon_url
diff --git a/g4f/api/__init__.py b/g4f/api/__init__.py
index 02ba5260..21e69388 100644
--- a/g4f/api/__init__.py
+++ b/g4f/api/__init__.py
@@ -4,6 +4,7 @@ import logging
 import json
 import uvicorn
 import secrets
+import os
 
 from fastapi import FastAPI, Response, Request
 from fastapi.responses import StreamingResponse, RedirectResponse, HTMLResponse, JSONResponse
@@ -13,13 +14,16 @@ from starlette.exceptions import HTTPException
 from starlette.status import HTTP_422_UNPROCESSABLE_ENTITY, HTTP_401_UNAUTHORIZED, HTTP_403_FORBIDDEN
 from fastapi.encoders import jsonable_encoder
 from fastapi.middleware.cors import CORSMiddleware
+from starlette.responses import FileResponse
 from pydantic import BaseModel
 from typing import Union, Optional
 
 import g4f
 import g4f.debug
 from g4f.client import AsyncClient, ChatCompletion
+from g4f.providers.response import BaseConversation
 from g4f.client.helper import filter_none
+from g4f.image import is_accepted_format, images_dir
 from g4f.typing import Messages
 from g4f.cookies import read_cookie_files
 
@@ -63,6 +67,7 @@ class ChatCompletionsConfig(BaseModel):
     api_key: Optional[str] = None
     web_search: Optional[bool] = None
     proxy: Optional[str] = None
+    conversation_id: str = None
 
 class ImageGenerationConfig(BaseModel):
     prompt: str
@@ -98,6 +103,7 @@ class Api:
         self.client = AsyncClient()
         self.g4f_api_key = g4f_api_key
         self.get_g4f_api_key = APIKeyHeader(name="g4f-api-key")
+        self.conversations: dict[str, dict[str, BaseConversation]] = {}
 
     def register_authorization(self):
         @self.app.middleware("http")
@@ -179,12 +185,21 @@ class Api:
         async def chat_completions(config: ChatCompletionsConfig, request: Request = None, provider: str = None):
             try:
                 config.provider = provider if config.provider is None else config.provider
+                if config.provider is None:
+                    config.provider = AppConfig.provider
                 if config.api_key is None and request is not None:
                     auth_header = request.headers.get("Authorization")
                     if auth_header is not None:
-                        auth_header = auth_header.split(None, 1)[-1]
-                        if auth_header and auth_header != "Bearer":
-                            config.api_key = auth_header
+                        api_key = auth_header.split(None, 1)[-1]
+                        if api_key and api_key != "Bearer":
+                            config.api_key = api_key
+
+                conversation = return_conversation = None
+                if config.conversation_id is not None and config.provider is not None:
+                    return_conversation = True
+                    if config.conversation_id in self.conversations:
+                        if config.provider in self.conversations[config.conversation_id]:
+                            conversation = self.conversations[config.conversation_id][config.provider]
 
                 # Create the completion response
                 response = self.client.chat.completions.create(
@@ -194,6 +209,11 @@ class Api:
                             "provider": AppConfig.provider,
                             "proxy": AppConfig.proxy,
                             **config.dict(exclude_none=True),
+                            **{
+                                "conversation_id": None,
+                                "return_conversation": return_conversation,
+                                "conversation": conversation
+                            }
                         },
                         ignored=AppConfig.ignored_providers
                     ),
@@ -206,7 +226,13 @@ class Api:
                 async def streaming():
                     try:
                         async for chunk in response:
-                            yield f"data: {json.dumps(chunk.to_json())}\n\n"
+                            if isinstance(chunk, BaseConversation):
+                                if config.conversation_id is not None and config.provider is not None:
+                                    if config.conversation_id not in self.conversations:
+                                        self.conversations[config.conversation_id] = {}
+                                    self.conversations[config.conversation_id][config.provider] = chunk
+                            else:
+                                yield f"data: {json.dumps(chunk.to_json())}\n\n"
                     except GeneratorExit:
                         pass
                     except Exception as e:
@@ -222,7 +248,13 @@ class Api:
 
         @self.app.post("/v1/images/generate")
         @self.app.post("/v1/images/generations")
-        async def generate_image(config: ImageGenerationConfig):
+        async def generate_image(config: ImageGenerationConfig, request: Request):
+            if config.api_key is None:
+                auth_header = request.headers.get("Authorization")
+                if auth_header is not None:
+                    api_key = auth_header.split(None, 1)[-1]
+                    if api_key and api_key != "Bearer":
+                        config.api_key = api_key
             try:
                 response = await self.client.images.generate(
                     prompt=config.prompt,
@@ -234,6 +266,9 @@ class Api:
                         proxy = config.proxy
                     )
                 )
+                for image in response.data:
+                    if hasattr(image, "url") and image.url.startswith("/"):
+                        image.url = f"{request.base_url}{image.url.lstrip('/')}"
                 return JSONResponse(response.to_json())
             except Exception as e:
                 logger.exception(e)
@@ -243,6 +278,18 @@ class Api:
         async def completions():
             return Response(content=json.dumps({'info': 'Not working yet.'}, indent=4), media_type="application/json")
 
+        @self.app.get("/images/{filename}")
+        async def get_image(filename):
+            target = os.path.join(images_dir, filename)
+
+            if not os.path.isfile(target):
+                return Response(status_code=404)
+
+            with open(target, "rb") as f:
+                content_type = is_accepted_format(f.read(12))
+
+            return FileResponse(target, media_type=content_type)
+
 def format_exception(e: Exception, config: Union[ChatCompletionsConfig, ImageGenerationConfig], image: bool = False) -> str:
     last_provider = {} if not image else g4f.get_last_provider(True)
     provider = (AppConfig.image_provider if image else AppConfig.provider) if config.provider is None else config.provider
diff --git a/g4f/client/__init__.py b/g4f/client/__init__.py
index 549a244b..dea19a60 100644
--- a/g4f/client/__init__.py
+++ b/g4f/client/__init__.py
@@ -6,24 +6,27 @@ import random
 import string
 import asyncio
 import base64
-import aiohttp
 import logging
-from typing import Union, AsyncIterator, Iterator, Coroutine
+from typing import Union, AsyncIterator, Iterator, Coroutine, Optional
 
 from ..providers.base_provider import AsyncGeneratorProvider
-from ..image import ImageResponse, to_image, to_data_uri, is_accepted_format, EXTENSIONS_MAP
-from ..typing import Messages, Image
-from ..providers.types import ProviderType, FinishReason, BaseConversation
-from ..errors import NoImageResponseError
+from ..image import ImageResponse, copy_images, images_dir
+from ..typing import Messages, Image, ImageType
+from ..providers.types import ProviderType
+from ..providers.response import ResponseType, FinishReason, BaseConversation
+from ..errors import NoImageResponseError, ModelNotFoundError
 from ..providers.retry_provider import IterListProvider
+from ..providers.base_provider import get_running_loop
 from ..Provider.needs_auth.BingCreateImages import BingCreateImages
-from ..requests.aiohttp import get_connector
 from .stubs import ChatCompletion, ChatCompletionChunk, Image, ImagesResponse
 from .image_models import ImageModels
 from .types import IterResponse, ImageProvider, Client as BaseClient
 from .service import get_model_and_provider, get_last_provider, convert_to_provider
 from .helper import find_stop, filter_json, filter_none, safe_aclose, to_sync_iter, to_async_iterator
 
+ChatCompletionResponseType = Iterator[Union[ChatCompletion, ChatCompletionChunk, BaseConversation]]
+AsyncChatCompletionResponseType = AsyncIterator[Union[ChatCompletion, ChatCompletionChunk, BaseConversation]]
+
 try:
     anext # Python 3.8+
 except NameError:
@@ -35,12 +38,12 @@ except NameError:
 
 # Synchronous iter_response function
 def iter_response(
-    response: Union[Iterator[str], AsyncIterator[str]],
+    response: Union[Iterator[Union[str, ResponseType]]],
     stream: bool,
-    response_format: dict = None,
-    max_tokens: int = None,
-    stop: list = None
-) -> Iterator[Union[ChatCompletion, ChatCompletionChunk]]:
+    response_format: Optional[dict] = None,
+    max_tokens: Optional[int] = None,
+    stop: Optional[list[str]] = None
+) -> ChatCompletionResponseType:
     content = ""
     finish_reason = None
     completion_id = ''.join(random.choices(string.ascii_letters + string.digits, k=28))
@@ -88,22 +91,23 @@ def iter_response(
         yield ChatCompletion(content, finish_reason, completion_id, int(time.time()))
 
 # Synchronous iter_append_model_and_provider function
-def iter_append_model_and_provider(response: Iterator[ChatCompletionChunk]) -> Iterator[ChatCompletionChunk]:
+def iter_append_model_and_provider(response: ChatCompletionResponseType) -> ChatCompletionResponseType:
     last_provider = None
 
     for chunk in response:
-        last_provider = get_last_provider(True) if last_provider is None else last_provider
-        chunk.model = last_provider.get("model")
-        chunk.provider = last_provider.get("name")
-        yield chunk
+        if isinstance(chunk, (ChatCompletion, ChatCompletionChunk)):
+            last_provider = get_last_provider(True) if last_provider is None else last_provider
+            chunk.model = last_provider.get("model")
+            chunk.provider = last_provider.get("name")
+            yield chunk
 
 async def async_iter_response(
-    response: AsyncIterator[str],
+    response: AsyncIterator[Union[str, ResponseType]],
     stream: bool,
-    response_format: dict = None,
-    max_tokens: int = None,
-    stop: list = None
-) -> AsyncIterator[Union[ChatCompletion, ChatCompletionChunk]]:
+    response_format: Optional[dict] = None,
+    max_tokens: Optional[int] = None,
+    stop: Optional[list[str]] = None
+) -> AsyncChatCompletionResponseType:
     content = ""
     finish_reason = None
     completion_id = ''.join(random.choices(string.ascii_letters + string.digits, k=28))
@@ -149,13 +153,16 @@ async def async_iter_response(
         if hasattr(response, 'aclose'):
             await safe_aclose(response)
 
-async def async_iter_append_model_and_provider(response: AsyncIterator[ChatCompletionChunk]) -> AsyncIterator:
+async def async_iter_append_model_and_provider(
+        response: AsyncChatCompletionResponseType
+    ) -> AsyncChatCompletionResponseType:
     last_provider = None
     try:
         async for chunk in response:
-            last_provider = get_last_provider(True) if last_provider is None else last_provider
-            chunk.model = last_provider.get("model")
-            chunk.provider = last_provider.get("name")
+            if isinstance(chunk, (ChatCompletion, ChatCompletionChunk)):
+                last_provider = get_last_provider(True) if last_provider is None else last_provider
+                chunk.model = last_provider.get("model")
+                chunk.provider = last_provider.get("name")
             yield chunk
     finally:
         if hasattr(response, 'aclose'):
@@ -164,8 +171,8 @@ async def async_iter_append_model_and_provider(response: AsyncIterator[ChatCompl
 class Client(BaseClient):
     def __init__(
         self,
-        provider: ProviderType = None,
-        image_provider: ImageProvider = None,
+        provider: Optional[ProviderType] = None,
+        image_provider: Optional[ImageProvider] = None,
         **kwargs
     ) -> None:
         super().__init__(**kwargs)
@@ -173,7 +180,7 @@ class Client(BaseClient):
         self.images: Images = Images(self, image_provider)
 
 class Completions:
-    def __init__(self, client: Client, provider: ProviderType = None):
+    def __init__(self, client: Client, provider: Optional[ProviderType] = None):
         self.client: Client = client
         self.provider: ProviderType = provider
 
@@ -181,16 +188,16 @@ class Completions:
         self,
         messages: Messages,
         model: str,
-        provider: ProviderType = None,
-        stream: bool = False,
-        proxy: str = None,
-        response_format: dict = None,
-        max_tokens: int = None,
-        stop: Union[list[str], str] = None,
-        api_key: str = None,
-        ignored: list[str] = None,
-        ignore_working: bool = False,
-        ignore_stream: bool = False,
+        provider: Optional[ProviderType] = None,
+        stream: Optional[bool] = False,
+        proxy: Optional[str] = None,
+        response_format: Optional[dict] = None,
+        max_tokens: Optional[int] = None,
+        stop: Optional[Union[list[str], str]] = None,
+        api_key: Optional[str] = None,
+        ignored: Optional[list[str]] = None,
+        ignore_working: Optional[bool] = False,
+        ignore_stream: Optional[bool] = False,
         **kwargs
     ) -> IterResponse:
         model, provider = get_model_and_provider(
@@ -234,22 +241,38 @@ class Completions:
 class Chat:
     completions: Completions
 
-    def __init__(self, client: Client, provider: ProviderType = None):
+    def __init__(self, client: Client, provider: Optional[ProviderType] = None):
         self.completions = Completions(client, provider)
 
 class Images:
-    def __init__(self, client: Client, provider: ProviderType = None):
+    def __init__(self, client: Client, provider: Optional[ProviderType] = None):
         self.client: Client = client
-        self.provider: ProviderType = provider
+        self.provider: Optional[ProviderType] = provider
         self.models: ImageModels = ImageModels(client)
 
-    def generate(self, prompt: str, model: str = None, provider: ProviderType = None, response_format: str = "url", proxy: str = None, **kwargs) -> ImagesResponse:
+    def generate(
+        self,
+        prompt: str,
+        model: str = None,
+        provider: Optional[ProviderType] = None,
+        response_format: str = "url",
+        proxy: Optional[str] = None,
+        **kwargs
+    ) -> ImagesResponse:
         """
         Synchronous generate method that runs the async_generate method in an event loop.
         """
-        return asyncio.run(self.async_generate(prompt, model, provider, response_format=response_format, proxy=proxy, **kwargs))
+        return asyncio.run(self.async_generate(prompt, model, provider, response_format, proxy, **kwargs))
 
-    async def async_generate(self, prompt: str, model: str = None, provider: ProviderType = None, response_format: str = "url", proxy: str = None, **kwargs) -> ImagesResponse:
+    async def async_generate(
+        self,
+        prompt: str,
+        model: Optional[str] = None,
+        provider: Optional[ProviderType] = None,
+        response_format: Optional[str] = "url",
+        proxy: Optional[str] = None,
+        **kwargs
+    ) -> ImagesResponse:
         if provider is None:
             provider_handler = self.models.get(model, provider or self.provider or BingCreateImages)
         elif isinstance(provider, str):
@@ -257,97 +280,73 @@ class Images:
         else:
             provider_handler = provider
         if provider_handler is None:
-            raise ValueError(f"Unknown model: {model}")
-        if proxy is None:
-            proxy = self.client.proxy
-
+            raise ModelNotFoundError(f"Unknown model: {model}")
         if isinstance(provider_handler, IterListProvider):
             if provider_handler.providers:
                 provider_handler = provider_handler.providers[0]
             else:
-                raise ValueError(f"IterListProvider for model {model} has no providers")
+                raise ModelNotFoundError(f"IterListProvider for model {model} has no providers")
+        if proxy is None:
+            proxy = self.client.proxy
 
         response = None
-        if hasattr(provider_handler, "create_async_generator"):
-            messages = [{"role": "user", "content": prompt}]
-            async for item in provider_handler.create_async_generator(model, messages, **kwargs):
+        if isinstance(provider, type) and issubclass(provider, AsyncGeneratorProvider):
+            messages = [{"role": "user", "content": f"Generate a image: {prompt}"}]
+            async for item in provider_handler.create_async_generator(model, messages, prompt=prompt, **kwargs):
                 if isinstance(item, ImageResponse):
                     response = item
                     break
-        elif hasattr(provider, 'create'):
+        elif hasattr(provider_handler, 'create'):
             if asyncio.iscoroutinefunction(provider_handler.create):
                 response = await provider_handler.create(prompt)
             else:
                 response = provider_handler.create(prompt) 
             if isinstance(response, str):
                 response = ImageResponse([response], prompt)
+        elif hasattr(provider_handler, "create_completion"):
+            get_running_loop(check_nested=True)
+            messages = [{"role": "user", "content": f"Generate a image: {prompt}"}]
+            for item in provider_handler.create_completion(model, messages, prompt=prompt, **kwargs):
+                if isinstance(item, ImageResponse):
+                    response = item
+                    break
         else:
             raise ValueError(f"Provider {provider} does not support image generation")
         if isinstance(response, ImageResponse):
-            return await self._process_image_response(response, response_format, proxy, model=model, provider=provider)
-
+            return await self._process_image_response(
+                response,
+                response_format,
+                proxy,
+                model,
+                getattr(provider_handler, "__name__", None)
+            )
         raise NoImageResponseError(f"Unexpected response type: {type(response)}")
 
-    async def _process_image_response(self, response: ImageResponse, response_format: str, proxy: str = None, model: str = None, provider: str = None) -> ImagesResponse:
-            async def process_image_item(session: aiohttp.ClientSession, image_data: str):
-                image_data_bytes = None
-                if image_data.startswith("http://") or image_data.startswith("https://"):
-                    if response_format == "url":
-                        return Image(url=image_data, revised_prompt=response.alt)
-                    elif response_format == "b64_json":
-                        # Fetch the image data and convert it to base64
-                        image_data_bytes = await self._fetch_image(session, image_data)
-                        b64_json = base64.b64encode(image_data_bytes).decode("utf-8")
-                        return Image(b64_json=b64_json, url=image_data, revised_prompt=response.alt)
-                else:
-                    # Assume image_data is base64 data or binary
-                    if response_format == "url":
-                        if image_data.startswith("data:image"):
-                            # Remove the data URL scheme and get the base64 data
-                            base64_data = image_data.split(",", 1)[-1]
-                        else:
-                            base64_data = image_data
-                        # Decode the base64 data
-                        image_data_bytes = base64.b64decode(base64_data)
-                if image_data_bytes:
-                    file_name = self._save_image(image_data_bytes)
-                    return Image(url=file_name, revised_prompt=response.alt)
-                else:
-                    raise ValueError("Unable to process image data")
-
-            last_provider = get_last_provider(True)
-            async with aiohttp.ClientSession(cookies=response.get("cookies"), connector=get_connector(proxy=proxy)) as session:
-                return ImagesResponse(
-                    await asyncio.gather(*[process_image_item(session, image_data) for image_data in response.get_list()]),
-                    model=last_provider.get("model") if model is None else model,
-                    provider=last_provider.get("name") if provider is None else provider
-                )
-
-    async def _fetch_image(self, session: aiohttp.ClientSession, url: str) -> bytes:
-        # Asynchronously fetch image data from the URL
-        async with session.get(url) as resp:
-            if resp.status == 200:
-                return await resp.read()
-            else:
-                raise RuntimeError(f"Failed to fetch image from {url}, status code {resp.status}")
-
-    def _save_image(self, image_data_bytes: bytes) -> str:
-        os.makedirs('generated_images', exist_ok=True)
-        image = to_image(image_data_bytes)
-        file_name = f"generated_images/image_{int(time.time())}_{random.randint(0, 10000)}.{EXTENSIONS_MAP[is_accepted_format(image_data_bytes)]}"
-        image.save(file_name)
-        return file_name
-
-    def create_variation(self, image: Union[str, bytes], model: str = None, provider: ProviderType = None, response_format: str = "url", **kwargs) -> ImagesResponse:
+    def create_variation(
+        self,
+        image: Union[str, bytes],
+        model: str = None,
+        provider: Optional[ProviderType] = None,
+        response_format: str = "url",
+        **kwargs
+    ) -> ImagesResponse:
         return asyncio.run(self.async_create_variation(
            image, model, provider, response_format, **kwargs
         ))
 
-    async def async_create_variation(self, image: Union[str, bytes], model: str = None, provider: ProviderType = None, response_format: str = "url", proxy: str = None, **kwargs) -> ImagesResponse:
+    async def async_create_variation(
+        self,
+        image: ImageType,
+        model: Optional[str] = None,
+        provider: Optional[ProviderType] = None,
+        response_format: str = "url",
+        proxy: Optional[str] = None,
+        **kwargs
+    ) -> ImagesResponse:
         if provider is None:
             provider = self.models.get(model, provider or self.provider or BingCreateImages)
             if provider is None:
-                raise ValueError(f"Unknown model: {model}")
+                raise ModelNotFoundError(f"Unknown model: {model}")
         if isinstance(provider, str):
             provider = convert_to_provider(provider)
         if proxy is None:
@@ -355,38 +354,61 @@ class Images:
 
         if isinstance(provider, type) and issubclass(provider, AsyncGeneratorProvider):
             messages = [{"role": "user", "content": "create a variation of this image"}]
-            image_data = to_data_uri(image)
             generator = None
             try:
-                generator = provider.create_async_generator(model, messages, image=image_data, response_format=response_format, proxy=proxy, **kwargs)
-                async for response in generator:
-                    if isinstance(response, ImageResponse):
-                        return self._process_image_response(response)
-            except RuntimeError as e:
-                if "async generator ignored GeneratorExit" in str(e):
-                    logging.warning("Generator ignored GeneratorExit in create_variation, handling gracefully")
-                else:
-                    raise
+                generator = provider.create_async_generator(model, messages, image=image, response_format=response_format, proxy=proxy, **kwargs)
+                async for chunk in generator:
+                    if isinstance(chunk, ImageResponse):
+                        response = chunk
+                        break
             finally:
                 if generator and hasattr(generator, 'aclose'):
                     await safe_aclose(generator)
-                logging.info("AsyncGeneratorProvider processing completed in create_variation")
         elif hasattr(provider, 'create_variation'):
             if asyncio.iscoroutinefunction(provider.create_variation):
                 response = await provider.create_variation(image, model=model, response_format=response_format, proxy=proxy, **kwargs)
             else:
                 response = provider.create_variation(image, model=model, response_format=response_format, proxy=proxy, **kwargs)
-            if isinstance(response, str):
-                response = ImageResponse([response])
-            return self._process_image_response(response)
         else:
-            raise ValueError(f"Provider {provider} does not support image variation")
+            raise NoImageResponseError(f"Provider {provider} does not support image variation")
+    
+        if isinstance(response, str):
+            response = ImageResponse([response])
+        if isinstance(response, ImageResponse):
+            return self._process_image_response(response, response_format, proxy, model, getattr(provider, "__name__", None))
+        raise NoImageResponseError(f"Unexpected response type: {type(response)}")
+
+    async def _process_image_response(
+        self,
+        response: ImageResponse,
+        response_format: str,
+        proxy: str = None,
+        model: Optional[str] = None,
+        provider: Optional[str] = None
+    ) -> list[Image]:
+        if response_format in ("url", "b64_json"):
+            images = await copy_images(response.get_list(), response.options.get("cookies"), proxy)
+            async def process_image_item(image_file: str) -> Image:
+                if response_format == "b64_json":
+                    with open(os.path.join(images_dir, os.path.basename(image_file)), "rb") as file:
+                        image_data = base64.b64encode(file.read()).decode()
+                        return Image(url=image_file, b64_json=image_data, revised_prompt=response.alt)
+                return Image(url=image_file, revised_prompt=response.alt)
+            images = await asyncio.gather(*[process_image_item(image) for image in images])
+        else:
+            images = [Image(url=image, revised_prompt=response.alt) for image in response.get_list()]
+        last_provider = get_last_provider(True)
+        return ImagesResponse(
+            images,
+            model=last_provider.get("model") if model is None else model,
+            provider=last_provider.get("name") if provider is None else provider
+        )
 
 class AsyncClient(BaseClient):
     def __init__(
         self,
-        provider: ProviderType = None,
-        image_provider: ImageProvider = None,
+        provider: Optional[ProviderType] = None,
+        image_provider: Optional[ImageProvider] = None,
         **kwargs
     ) -> None:
         super().__init__(**kwargs)
@@ -396,11 +418,11 @@ class AsyncClient(BaseClient):
 class AsyncChat:
     completions: AsyncCompletions
 
-    def __init__(self, client: AsyncClient, provider: ProviderType = None):
+    def __init__(self, client: AsyncClient, provider: Optional[ProviderType] = None):
         self.completions = AsyncCompletions(client, provider)
 
 class AsyncCompletions:
-    def __init__(self, client: AsyncClient, provider: ProviderType = None):
+    def __init__(self, client: AsyncClient, provider: Optional[ProviderType] = None):
         self.client: AsyncClient = client
         self.provider: ProviderType = provider
 
@@ -408,18 +430,18 @@ class AsyncCompletions:
         self,
         messages: Messages,
         model: str,
-        provider: ProviderType = None,
-        stream: bool = False,
-        proxy: str = None,
-        response_format: dict = None,
-        max_tokens: int = None,
-        stop: Union[list[str], str] = None,
-        api_key: str = None,
-        ignored: list[str] = None,
-        ignore_working: bool = False,
-        ignore_stream: bool = False,
+        provider: Optional[ProviderType] = None,
+        stream: Optional[bool] = False,
+        proxy: Optional[str] = None,
+        response_format: Optional[dict] = None,
+        max_tokens: Optional[int] = None,
+        stop: Optional[Union[list[str], str]] = None,
+        api_key: Optional[str] = None,
+        ignored: Optional[list[str]] = None,
+        ignore_working: Optional[bool] = False,
+        ignore_stream: Optional[bool] = False,
         **kwargs
-    ) -> Union[Coroutine[ChatCompletion], AsyncIterator[ChatCompletionChunk]]:
+    ) -> Union[Coroutine[ChatCompletion], AsyncIterator[ChatCompletionChunk, BaseConversation]]:
         model, provider = get_model_and_provider(
             model,
             self.provider if provider is None else provider,
@@ -450,15 +472,29 @@ class AsyncCompletions:
         return response if stream else anext(response)
 
 class AsyncImages(Images):
-    def __init__(self, client: AsyncClient, provider: ImageProvider = None):
+    def __init__(self, client: AsyncClient, provider: Optional[ProviderType] = None):
         self.client: AsyncClient = client
-        self.provider: ImageProvider = provider
+        self.provider: Optional[ProviderType] = provider
         self.models: ImageModels = ImageModels(client)
 
-    async def generate(self, prompt: str, model: str = None, provider: ProviderType = None, response_format: str = "url", **kwargs) -> ImagesResponse:
+    async def generate(
+        self,
+        prompt: str,
+        model: Optional[str] = None,
+        provider: Optional[ProviderType] = None,
+        response_format: str = "url",
+        **kwargs
+    ) -> ImagesResponse:
         return await self.async_generate(prompt, model, provider, response_format, **kwargs)
 
-    async def create_variation(self, image: Union[str, bytes], model: str = None, provider: ProviderType = None, response_format: str = "url", **kwargs) -> ImagesResponse:
+    async def create_variation(
+        self,
+        image: ImageType,
+        model: str = None,
+        provider: ProviderType = None,
+        response_format: str = "url",
+        **kwargs
+    ) -> ImagesResponse:
         return await self.async_create_variation(
            image, model, provider, response_format, **kwargs
-        )
+        )
\ No newline at end of file
diff --git a/g4f/client/helper.py b/g4f/client/helper.py
index 71bfd38a..93588c07 100644
--- a/g4f/client/helper.py
+++ b/g4f/client/helper.py
@@ -6,7 +6,7 @@ import threading
 import logging
 import asyncio
 
-from typing import AsyncIterator, Iterator, AsyncGenerator
+from typing import AsyncIterator, Iterator, AsyncGenerator, Optional
 
 def filter_json(text: str) -> str:
     """
@@ -23,7 +23,7 @@ def filter_json(text: str) -> str:
         return match.group("code")
     return text
 
-def find_stop(stop, content: str, chunk: str = None):
+def find_stop(stop: Optional[list[str]], content: str, chunk: str = None):
     first = -1
     word = None
     if stop is not None:
diff --git a/g4f/client/image_models.py b/g4f/client/image_models.py
index edaa4592..0b97a56b 100644
--- a/g4f/client/image_models.py
+++ b/g4f/client/image_models.py
@@ -1,7 +1,5 @@
 from __future__ import annotations
 
-from .types import Client, ImageProvider
-
 from ..models import ModelUtils
 
 class ImageModels():
diff --git a/g4f/client/types.py b/g4f/client/types.py
index 4f252ba9..5010e098 100644
--- a/g4f/client/types.py
+++ b/g4f/client/types.py
@@ -3,7 +3,7 @@ from __future__ import annotations
 import os
 
 from .stubs import ChatCompletion, ChatCompletionChunk
-from ..providers.types import BaseProvider, ProviderType, FinishReason
+from ..providers.types import BaseProvider
 from typing import Union, Iterator, AsyncIterator
 
 ImageProvider = Union[BaseProvider, object]
diff --git a/g4f/gui/client/index.html b/g4f/gui/client/index.html
index 48214093..6c2ad8b6 100644
--- a/g4f/gui/client/index.html
+++ b/g4f/gui/client/index.html
@@ -112,6 +112,11 @@
                 <label for="hide-systemPrompt" class="toogle" title="For more space on phones"></label>
             </div>
             <div class="field">
+                <span class="label">Download generated images</span>
+                <input type="checkbox" id="download_images" checked/>
+                <label for="download_images" class="toogle" title="Download and save generated images to /generated_images"></label>
+            </div>
+            <div class="field">
                 <span class="label">Auto continue in ChatGPT</span>
                 <input id="auto_continue" type="checkbox" name="auto_continue" checked/>
                 <label for="auto_continue" class="toogle" title="Continue large responses in OpenaiChat"></label>
diff --git a/g4f/gui/client/static/css/style.css b/g4f/gui/client/static/css/style.css
index 76399703..c4b61d87 100644
--- a/g4f/gui/client/static/css/style.css
+++ b/g4f/gui/client/static/css/style.css
@@ -498,6 +498,7 @@ body {
     gap: 12px;
     cursor: pointer;
     animation: show_popup 0.4s;
+    height: 28px;
 }
 
 .toolbar .regenerate {
diff --git a/g4f/gui/client/static/js/chat.v1.js b/g4f/gui/client/static/js/chat.v1.js
index a3e94ee2..0136f9c4 100644
--- a/g4f/gui/client/static/js/chat.v1.js
+++ b/g4f/gui/client/static/js/chat.v1.js
@@ -428,10 +428,14 @@ async function add_message_chunk(message, message_id) {
         p.innerText = message.log;
         log_storage.appendChild(p);
     }
-    window.scrollTo(0, 0);
-    if (message_box.scrollTop >= message_box.scrollHeight - message_box.clientHeight - 100) {
-        message_box.scrollTo({ top: message_box.scrollHeight, behavior: "auto" });
+    let scroll_down = ()=>{
+        if (message_box.scrollTop >= message_box.scrollHeight - message_box.clientHeight - 100) {
+            window.scrollTo(0, 0);
+            message_box.scrollTo({ top: message_box.scrollHeight, behavior: "auto" });
+        }
     }
+    setTimeout(scroll_down, 200);
+    setTimeout(scroll_down, 1000);
 }
 
 cameraInput?.addEventListener("click", (e) => {
@@ -492,6 +496,7 @@ const ask_gpt = async (message_index = -1, message_id) => {
         const file = input && input.files.length > 0 ? input.files[0] : null;
         const provider = providerSelect.options[providerSelect.selectedIndex].value;
         const auto_continue = document.getElementById("auto_continue")?.checked;
+        const download_images = document.getElementById("download_images")?.checked;
         let api_key = get_api_key_by_provider(provider);
         await api("conversation", {
             id: message_id,
@@ -501,13 +506,13 @@ const ask_gpt = async (message_index = -1, message_id) => {
             provider: provider,
             messages: messages,
             auto_continue: auto_continue,
+            download_images: download_images,
             api_key: api_key,
         }, file, message_id);
         if (!error_storage[message_id]) {
             html = markdown_render(message_storage[message_id]);
             content_map.inner.innerHTML = html;
             highlight(content_map.inner);
-
             if (imageInput) imageInput.value = "";
             if (cameraInput) cameraInput.value = "";
             if (fileInput) fileInput.value = "";
@@ -1239,8 +1244,7 @@ async function load_version() {
     if (versions["version"] != versions["latest_version"]) {
         let release_url = 'https://github.com/xtekky/gpt4free/releases/tag/' + versions["latest_version"];
         let title = `New version: ${versions["latest_version"]}`;
-        text += `<a href="${release_url}" target="_blank" title="${title}">${versions["version"]}</a> `;
-        text += `<i class="fa-solid fa-rotate"></i>`
+        text += `<a href="${release_url}" target="_blank" title="${title}">${versions["version"]}</a> 🆕`;
     } else {
         text += versions["version"];
     }
diff --git a/g4f/gui/server/api.py b/g4f/gui/server/api.py
index 29fc34e2..00eb7182 100644
--- a/g4f/gui/server/api.py
+++ b/g4f/gui/server/api.py
@@ -2,34 +2,22 @@ from __future__ import annotations
 
 import logging
 import os
-import uuid
 import asyncio
-import time
-from aiohttp import ClientSession
-from typing import Iterator, Optional
+from typing import Iterator
 from flask import send_from_directory
+from inspect import signature
 
 from g4f import version, models
 from g4f import get_last_provider, ChatCompletion
 from g4f.errors import VersionNotFoundError
-from g4f.typing import Cookies
-from g4f.image import ImagePreview, ImageResponse, is_accepted_format, extract_data_uri
-from g4f.requests.aiohttp import get_connector
+from g4f.image import ImagePreview, ImageResponse, copy_images, ensure_images_dir, images_dir
 from g4f.Provider import ProviderType, __providers__, __map__
-from g4f.providers.base_provider import ProviderModelMixin, FinishReason
-from g4f.providers.conversation import BaseConversation
+from g4f.providers.base_provider import ProviderModelMixin
+from g4f.providers.response import BaseConversation, FinishReason
+from g4f.client.service import convert_to_provider
 from g4f import debug
 
 logger = logging.getLogger(__name__)
-
-# Define the directory for generated images
-images_dir = "./generated_images"
-
-# Function to ensure the images directory exists
-def ensure_images_dir():
-    if not os.path.exists(images_dir):
-        os.makedirs(images_dir)
-
 conversations: dict[dict[str, BaseConversation]] = {}
 
 class Api:
@@ -42,7 +30,10 @@ class Api:
         if provider in __map__:
             provider: ProviderType = __map__[provider]
             if issubclass(provider, ProviderModelMixin):
-                models = provider.get_models() if api_key is None else provider.get_models(api_key=api_key)
+                if api_key is not None and "api_key" in signature(provider.get_models).parameters:
+                    models = provider.get_models(api_key=api_key)
+                else:
+                    models = provider.get_models()
                 return [
                     {
                         "model": model,
@@ -90,7 +81,7 @@ class Api:
     def get_providers() -> list[str]:
         return {
             provider.__name__: (provider.label if hasattr(provider, "label") else provider.__name__)
-            + (" (Image Generation)" if hasattr(provider, "image_models") else "")
+            + (" (Image Generation)" if getattr(provider, "image_models", None) else "")
             + (" (Image Upload)" if getattr(provider, "default_vision_model", None) else "")
             + (" (WebDriver)" if "webdriver" in provider.get_parameters() else "")
             + (" (Auth)" if provider.needs_auth else "")
@@ -120,16 +111,23 @@ class Api:
         api_key = json_data.get("api_key")
         if api_key is not None:
             kwargs["api_key"] = api_key
-        if json_data.get('web_search'):
-            if provider:
-                kwargs['web_search'] = True
-            else:
-                from .internet import get_search_message
-                messages[-1]["content"] = get_search_message(messages[-1]["content"])
+        do_web_search = json_data.get('web_search')
+        if do_web_search and provider:
+            provider_handler = convert_to_provider(provider)
+            if hasattr(provider_handler, "get_parameters"):
+                if "web_search" in provider_handler.get_parameters():
+                    kwargs['web_search'] = True
+                    do_web_search = False
+        if do_web_search:
+            from .internet import get_search_message
+            messages[-1]["content"] = get_search_message(messages[-1]["content"])
+        if json_data.get("auto_continue"):
+            kwargs['auto_continue'] = True
 
         conversation_id = json_data.get("conversation_id")
-        if conversation_id and provider in conversations and conversation_id in conversations[provider]:
-            kwargs["conversation"] = conversations[provider][conversation_id]
+        if conversation_id and provider:
+            if provider in conversations and conversation_id in conversations[provider]:
+                kwargs["conversation"] = conversations[provider][conversation_id]
 
         return {
             "model": model,
@@ -141,7 +139,7 @@ class Api:
             **kwargs
         }
 
-    def _create_response_stream(self, kwargs: dict, conversation_id: str, provider: str) -> Iterator:
+    def _create_response_stream(self, kwargs: dict, conversation_id: str, provider: str, download_images: bool = True) -> Iterator:
         if debug.logging:
             debug.logs = []
             print_callback = debug.log_handler
@@ -163,18 +161,22 @@ class Api:
                         first = False
                         yield self._format_json("provider", get_last_provider(True))
                     if isinstance(chunk, BaseConversation):
-                        if provider not in conversations:
-                            conversations[provider] = {}
-                        conversations[provider][conversation_id] = chunk
-                        yield self._format_json("conversation", conversation_id)
+                        if provider:
+                            if provider not in conversations:
+                                conversations[provider] = {}
+                            conversations[provider][conversation_id] = chunk
+                            yield self._format_json("conversation", conversation_id)
                     elif isinstance(chunk, Exception):
                         logger.exception(chunk)
                         yield self._format_json("message", get_error_message(chunk))
                     elif isinstance(chunk, ImagePreview):
                         yield self._format_json("preview", chunk.to_string())
                     elif isinstance(chunk, ImageResponse):
-                        images = asyncio.run(self._copy_images(chunk.get_list(), chunk.options.get("cookies")))
-                        yield self._format_json("content", str(ImageResponse(images, chunk.alt)))
+                        images = chunk
+                        if download_images:
+                            images = asyncio.run(copy_images(chunk.get_list(), chunk.options.get("cookies")))
+                            images = ImageResponse(images, chunk.alt)
+                        yield self._format_json("content", str(images))
                     elif not isinstance(chunk, FinishReason):
                         yield self._format_json("content", str(chunk))
                     if debug.logs:
@@ -185,31 +187,6 @@ class Api:
             logger.exception(e)
             yield self._format_json('error', get_error_message(e))
 
-    async def _copy_images(self, images: list[str], cookies: Optional[Cookies] = None):
-        ensure_images_dir()
-        async with ClientSession(
-            connector=get_connector(None, os.environ.get("G4F_PROXY")),
-            cookies=cookies
-        ) as session:
-            async def copy_image(image: str) -> str:
-                target = os.path.join(images_dir, f"{int(time.time())}_{str(uuid.uuid4())}")
-                if image.startswith("data:"):
-                    with open(target, "wb") as f:
-                        f.write(extract_data_uri(image))
-                else:
-                    async with session.get(image) as response:
-                        with open(target, "wb") as f:
-                            async for chunk in response.content.iter_any():
-                                f.write(chunk)
-                with open(target, "rb") as f:
-                    extension = is_accepted_format(f.read(12)).split("/")[-1]
-                    extension = "jpg" if extension == "jpeg" else extension
-                new_target = f"{target}.{extension}"
-                os.rename(target, new_target)
-                return f"/images/{os.path.basename(new_target)}"
-
-            return await asyncio.gather(*[copy_image(image) for image in images])
-
     def _format_json(self, response_type: str, content):
         return {
             'type': response_type,
@@ -221,4 +198,4 @@ def get_error_message(exception: Exception) -> str:
     provider = get_last_provider()
     if provider is None:
         return message
-    return f"{provider.__name__}: {message}"
+    return f"{provider.__name__}: {message}"
\ No newline at end of file
diff --git a/g4f/gui/server/backend.py b/g4f/gui/server/backend.py
index 020e49ef..917d779e 100644
--- a/g4f/gui/server/backend.py
+++ b/g4f/gui/server/backend.py
@@ -89,7 +89,12 @@ class Backend_Api(Api):
         kwargs = self._prepare_conversation_kwargs(json_data, kwargs)
 
         return self.app.response_class(
-            self._create_response_stream(kwargs, json_data.get("conversation_id"), json_data.get("provider")),
+            self._create_response_stream(
+                kwargs,
+                json_data.get("conversation_id"),
+                json_data.get("provider"),
+                json_data.get("download_images", True),
+            ),
             mimetype='text/event-stream'
         )
 
diff --git a/g4f/gui/server/internet.py b/g4f/gui/server/internet.py
index b41b5eae..bafa3af7 100644
--- a/g4f/gui/server/internet.py
+++ b/g4f/gui/server/internet.py
@@ -8,12 +8,14 @@ try:
 except ImportError:
     has_requirements = False
 from ...errors import MissingRequirementsError
-    
+from ... import debug
+
 import asyncio
 
 class SearchResults():
-    def __init__(self, results: list):
+    def __init__(self, results: list, used_words: int):
         self.results = results
+        self.used_words = used_words
 
     def __iter__(self):
         yield from self.results
@@ -104,7 +106,8 @@ async def search(query: str, n_results: int = 5, max_words: int = 2500, add_text
                 region="wt-wt",
                 safesearch="moderate",
                 timelimit="y",
-                max_results=n_results
+                max_results=n_results,
+                backend="html"
             ):
             results.append(SearchResultEntry(
                 result["title"],
@@ -120,6 +123,7 @@ async def search(query: str, n_results: int = 5, max_words: int = 2500, add_text
                 texts = await asyncio.gather(*requests)
 
         formatted_results = []
+        used_words = 0
         left_words = max_words
         for i, entry in enumerate(results):
             if add_text:
@@ -132,13 +136,14 @@ async def search(query: str, n_results: int = 5, max_words: int = 2500, add_text
                     left_words -= entry.snippet.count(" ")
                 if 0 > left_words:
                     break
+            used_words = max_words - left_words
             formatted_results.append(entry)
 
-        return SearchResults(formatted_results)
+        return SearchResults(formatted_results, used_words)
 
-def get_search_message(prompt) -> str:
+def get_search_message(prompt, n_results: int = 5, max_words: int = 2500) -> str:
     try:
-        search_results = asyncio.run(search(prompt))
+        search_results = asyncio.run(search(prompt, n_results, max_words))
         message = f"""
 {search_results}
 
@@ -149,7 +154,8 @@ Make sure to add the sources of cites using [[Number]](Url) notation after the r
 User request:
 {prompt}
 """
+        debug.log(f"Web search: '{prompt.strip()[:50]}...' {search_results.used_words} Words")
         return message
     except Exception as e:
-        print("Couldn't do web search:", e)
+        debug.log(f"Couldn't do web search: {e.__class__.__name__}: {e}")
         return prompt
\ No newline at end of file
diff --git a/g4f/image.py b/g4f/image.py
index 8a3d7a74..114dcc13 100644
--- a/g4f/image.py
+++ b/g4f/image.py
@@ -1,9 +1,13 @@
 from __future__ import annotations
 
+import os
 import re
+import time
+import uuid
 from io import BytesIO
 import base64
-from .typing import ImageType, Union, Image
+import asyncio
+from aiohttp import ClientSession
 
 try:
     from PIL.Image import open as open_image, new as new_image
@@ -12,7 +16,10 @@ try:
 except ImportError:
     has_requirements = False
 
+from .typing import ImageType, Union, Image, Optional, Cookies
 from .errors import MissingRequirementsError
+from .providers.response import ResponseType
+from .requests.aiohttp import get_connector
 
 ALLOWED_EXTENSIONS = {'png', 'jpg', 'jpeg', 'gif', 'webp', 'svg'}
 
@@ -23,10 +30,12 @@ EXTENSIONS_MAP: dict[str, str] = {
     "image/webp": "webp",
 }
 
+# Define the directory for generated images
+images_dir = "./generated_images"
+
 def fix_url(url:str) -> str:
     """ replace ' ' by '+' (to be markdown compliant)"""
     return url.replace(" ","+")
-    
 
 def to_image(image: ImageType, is_svg: bool = False) -> Image:
     """
@@ -223,7 +232,6 @@ def format_images_markdown(images: Union[str, list], alt: str, preview: Union[st
             preview = [preview.replace('{image}', image) if preview else image for image in images]
         result = "\n".join(
             f"[![#{idx+1} {alt}]({fix_url(preview[idx])})]({fix_url(image)})"
-            #f'[<img src="{preview[idx]}" width="200" alt="#{idx+1} {alt}">]({image})'
             for idx, image in enumerate(images)
         )
     start_flag = "<!-- generated images start -->\n"
@@ -260,7 +268,39 @@ def to_data_uri(image: ImageType) -> str:
         return f"data:{is_accepted_format(data)};base64,{data_base64}"
     return image
 
-class ImageResponse:
+# Function to ensure the images directory exists
+def ensure_images_dir():
+    if not os.path.exists(images_dir):
+        os.makedirs(images_dir)
+
+async def copy_images(images: list[str], cookies: Optional[Cookies] = None, proxy: Optional[str] = None):
+    ensure_images_dir()
+    async with ClientSession(
+        connector=get_connector(
+            proxy=os.environ.get("G4F_PROXY") if proxy is None else proxy
+        ),
+        cookies=cookies
+    ) as session:
+        async def copy_image(image: str) -> str:
+            target = os.path.join(images_dir, f"{int(time.time())}_{str(uuid.uuid4())}")
+            if image.startswith("data:"):
+                with open(target, "wb") as f:
+                    f.write(extract_data_uri(image))
+            else:
+                async with session.get(image) as response:
+                    with open(target, "wb") as f:
+                        async for chunk in response.content.iter_chunked(4096):
+                            f.write(chunk)
+            with open(target, "rb") as f:
+                extension = is_accepted_format(f.read(12)).split("/")[-1]
+                extension = "jpg" if extension == "jpeg" else extension
+            new_target = f"{target}.{extension}"
+            os.rename(target, new_target)
+            return f"/images/{os.path.basename(new_target)}"
+
+        return await asyncio.gather(*[copy_image(image) for image in images])
+
+class ImageResponse(ResponseType):
     def __init__(
         self,
         images: Union[str, list],
diff --git a/g4f/providers/base_provider.py b/g4f/providers/base_provider.py
index 9fa17fc3..b6df48e8 100644
--- a/g4f/providers/base_provider.py
+++ b/g4f/providers/base_provider.py
@@ -10,7 +10,8 @@ from inspect import signature, Parameter
 from typing import Callable, Union
 
 from ..typing import CreateResult, AsyncResult, Messages
-from .types import BaseProvider, FinishReason
+from .types import BaseProvider
+from .response import FinishReason, BaseConversation
 from ..errors import NestAsyncioError, ModelNotSupportedError
 from .. import debug
 
@@ -100,7 +101,7 @@ class AbstractProvider(BaseProvider):
         )
 
     @classmethod
-    def get_parameters(cls) -> dict:
+    def get_parameters(cls) -> dict[str, Parameter]:
         return signature(
             cls.create_async_generator if issubclass(cls, AsyncGeneratorProvider) else
             cls.create_async if issubclass(cls, AsyncProvider) else
@@ -258,7 +259,7 @@ class AsyncGeneratorProvider(AsyncProvider):
         """
         return "".join([
             str(chunk) async for chunk in cls.create_async_generator(model, messages, stream=False, **kwargs) 
-            if not isinstance(chunk, (Exception, FinishReason))
+            if not isinstance(chunk, (Exception, FinishReason, BaseConversation))
         ])
 
     @staticmethod
@@ -307,4 +308,4 @@ class ProviderModelMixin:
         elif model not in cls.get_models() and cls.models:
             raise ModelNotSupportedError(f"Model is not supported: {model} in: {cls.__name__}")
         debug.last_model = model
-        return model
+        return model
\ No newline at end of file
diff --git a/g4f/providers/conversation.py b/g4f/providers/conversation.py
deleted file mode 100644
index 921810d3..00000000
--- a/g4f/providers/conversation.py
+++ /dev/null
@@ -1,2 +0,0 @@
-class BaseConversation:
-    ...
\ No newline at end of file
diff --git a/g4f/providers/response.py b/g4f/providers/response.py
new file mode 100644
index 00000000..a4d1467a
--- /dev/null
+++ b/g4f/providers/response.py
@@ -0,0 +1,26 @@
+from __future__ import annotations
+
+from abc import abstractmethod
+
+class ResponseType:
+    @abstractmethod
+    def __str__(self) -> str:
+        pass
+
+class FinishReason():
+    def __init__(self, reason: str):
+        self.reason = reason
+
+    def __str__(self) -> str:
+        return ""
+
+class Sources(ResponseType):
+    def __init__(self, sources: list[dict[str, str]]) -> None:
+        self.list = sources
+
+    def __str__(self) -> str:
+        return "\n\n" + ("\n".join([f"{idx+1}. [{link['title']}]({link['url']})" for idx, link in enumerate(self.list)]))
+
+class BaseConversation(ResponseType):
+    def __str__(self) -> str:
+        return ""
\ No newline at end of file
diff --git a/g4f/providers/types.py b/g4f/providers/types.py
index e7ca32ee..d6d7cccd 100644
--- a/g4f/providers/types.py
+++ b/g4f/providers/types.py
@@ -3,7 +3,6 @@ from __future__ import annotations
 from abc import ABC, abstractmethod
 from typing import Union, Dict, Type
 from ..typing import Messages, CreateResult
-from .conversation import BaseConversation
 
 class BaseProvider(ABC):
     """
@@ -98,10 +97,6 @@ class BaseRetryProvider(BaseProvider):
 
 ProviderType = Union[Type[BaseProvider], BaseRetryProvider]
 
-class FinishReason():
-    def __init__(self, reason: str):
-        self.reason = reason
-
 class Streaming():
     def __init__(self, data: str) -> None:
         self.data = data
diff --git a/g4f/typing.py b/g4f/typing.py
index 710fde4b..1bc71141 100644
--- a/g4f/typing.py
+++ b/g4f/typing.py
@@ -10,11 +10,13 @@ if sys.version_info >= (3, 8):
     from typing import TypedDict
 else:
     from typing_extensions import TypedDict
+    
+from .providers.response import ResponseType
 
 SHA256 = NewType('sha_256_hash', str)
-CreateResult = Iterator[str]
-AsyncResult = AsyncIterator[str]
-Messages = List[Dict[str, Union[str,List[Dict[str,Union[str,Dict[str,str]]]]]]]
+CreateResult = Iterator[Union[str, ResponseType]]
+AsyncResult = AsyncIterator[Union[str, ResponseType]]
+Messages = List[Dict[str, Union[str, List[Dict[str, Union[str, Dict[str, str]]]]]]]
 Cookies = Dict[str, str]
 ImageType = Union[str, bytes, IO, Image, None]