From 688640b764c3dde1f3a85ec604adece760b6c62c Mon Sep 17 00:00:00 2001
From: Hexye <65314629+HexyeDEV@users.noreply.github.com>
Date: Fri, 28 Nov 2025 22:47:54 +0100
Subject: [PATCH 01/15] Add ItalyGPT provider

---
 g4f/Provider/ItalyGPT.py | 45 ++++++++++++++++++++++++++++++++++++++++
 1 file changed, 45 insertions(+)
 create mode 100644 g4f/Provider/ItalyGPT.py

diff --git a/g4f/Provider/ItalyGPT.py b/g4f/Provider/ItalyGPT.py
new file mode 100644
index 00000000..e13fc0ee
--- /dev/null
+++ b/g4f/Provider/ItalyGPT.py
@@ -0,0 +1,45 @@
+from .base_provider import AsyncGeneratorProvider, ProviderModelMixin
+from ..typing import AsyncResult, Messages
+from aiohttp import ClientSession
+
+class ItalyGPT(AsyncGeneratorProvider, ProviderModelMixin):
+    label = "ItalyGPT"
+    url = "https://italygpt.it"
+    working = True
+    supports_system_message = True
+    supports_message_history = True
+
+    default_model = "gpt-4o"
+    models = [default_model]
+
+    @classmethod
+    async def create_async_generator(
+        cls,
+        model: str,
+        messages: Messages,
+        stream: bool = True,
+        proxy: str = None,
+        **kwargs
+    ) -> AsyncResult:
+        model = cls.get_model(model)
+        headers = {
+            "content-type": "application/json",
+            "origin": "https://italygpt.it",
+            "referer": "https://italygpt.it/",
+            "user-agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/114.0.0.0 Safari/537.36"
+        }
+        payload = {
+            "messages": messages,
+            "stream": stream,
+        }
+        async with ClientSession() as session:
+            async with session.post(
+                f"{cls.url}/api/chat/",
+                json=payload,
+                headers=headers,
+                proxy=proxy,
+            ) as resp:
+                resp.raise_for_status()
+                async for chunk in resp.content:
+                    if chunk:
+                        yield chunk.decode()

From 6be76e3e843bf9b225007d06a097a39da3859093 Mon Sep 17 00:00:00 2001
From: hlohaus <983577+hlohaus@users.noreply.github.com>
Date: Sat, 29 Nov 2025 05:12:23 +0100
Subject: [PATCH 02/15] Add support for gemini-3-pro model in Gemini provider

---
 g4f/Provider/needs_auth/Gemini.py | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/g4f/Provider/needs_auth/Gemini.py b/g4f/Provider/needs_auth/Gemini.py
index fd7a4585..44e38728 100644
--- a/g4f/Provider/needs_auth/Gemini.py
+++ b/g4f/Provider/needs_auth/Gemini.py
@@ -71,6 +71,7 @@ models = {
     "gemini-2.0-flash-thinking": {"x-goog-ext-525001261-jspb": '[null,null,null,null,"9c17b1863f581b8a"]'},
     "gemini-2.0-flash-thinking-with-apps": {"x-goog-ext-525001261-jspb": '[null,null,null,null,"f8f8f5ea629f5d37"]'},
     # Currently used models
+    "gemini-3-pro": {"x-goog-ext-525001261-jspb": '[1,null,null,null,"9d8ca3786ebdfbea",null,null,0,[4]]'},
     "gemini-2.5-pro": {"x-goog-ext-525001261-jspb": '[1,null,null,null,"61530e79959ab139",null,null,null,[4]]'},
     "gemini-2.5-flash": {"x-goog-ext-525001261-jspb": '[1,null,null,null,"9ec249fc9ad08861",null,null,null,[4]]'},
     "gemini-audio": {}
@@ -89,7 +90,7 @@ class Gemini(AsyncGeneratorProvider, ProviderModelMixin):
     default_vision_model = default_model
     image_models = [default_image_model]
     models = [
-        default_model, "gemini-2.5-flash", "gemini-2.5-pro"
+        default_model, "gemini-3-pro", "gemini-2.5-flash", "gemini-2.5-pro"
     ]
 
     synthesize_content_type = "audio/vnd.wav"

From d76e56a66f70d7c7764975aa481927d046f62a5d Mon Sep 17 00:00:00 2001
From: Ammar <ammar.alkotb@gmail.com>
Date: Sat, 29 Nov 2025 06:15:05 +0200
Subject: [PATCH 03/15] fix error (#3265)

fix error
---
 g4f/Provider/needs_auth/OpenaiChat.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/g4f/Provider/needs_auth/OpenaiChat.py b/g4f/Provider/needs_auth/OpenaiChat.py
index 740c7c50..6e8b51f5 100644
--- a/g4f/Provider/needs_auth/OpenaiChat.py
+++ b/g4f/Provider/needs_auth/OpenaiChat.py
@@ -162,7 +162,7 @@ class OpenaiChat(AsyncAuthedProvider, ProviderModelMixin):
             hasher.update(data_bytes)
             image_hash = hasher.hexdigest()
             cache_file = ImagesCache.get(image_hash)
-            if cls.image_cache and file:
+            if cls.image_cache and cache_file:
                 debug.log("Using cached image")
                 return ImageRequest(cache_file)
             extension, mime_type = detect_file_type(data_bytes)

From 7b32f89eca5b11032d72464717d346c96b34c262 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Sat, 29 Nov 2025 04:17:42 +0000
Subject: [PATCH 04/15] Initial plan


From f0ea4c5b95d67312e0c183e859ef7bfcecd0fdf6 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Sat, 29 Nov 2025 04:22:02 +0000
Subject: [PATCH 05/15] Add GradientNetwork provider for chat.gradient.network

Co-authored-by: hlohaus <983577+hlohaus@users.noreply.github.com>
---
 g4f/Provider/GradientNetwork.py | 131 ++++++++++++++++++++++++++++++++
 g4f/Provider/__init__.py        |   1 +
 2 files changed, 132 insertions(+)
 create mode 100644 g4f/Provider/GradientNetwork.py

diff --git a/g4f/Provider/GradientNetwork.py b/g4f/Provider/GradientNetwork.py
new file mode 100644
index 00000000..b9286d0b
--- /dev/null
+++ b/g4f/Provider/GradientNetwork.py
@@ -0,0 +1,131 @@
+from __future__ import annotations
+
+import json
+
+from aiohttp import ClientSession
+
+from ..typing import AsyncResult, Messages
+from ..providers.response import Reasoning
+from .base_provider import AsyncGeneratorProvider, ProviderModelMixin
+
+
+class GradientNetwork(AsyncGeneratorProvider, ProviderModelMixin):
+    """
+    Provider for chat.gradient.network
+    Supports streaming text generation with various Qwen models.
+    """
+    label = "Gradient Network"
+    url = "https://chat.gradient.network"
+    api_endpoint = "https://chat.gradient.network/api/generate"
+
+    working = True
+    needs_auth = False
+    supports_stream = True
+    supports_system_message = True
+    supports_message_history = True
+
+    default_model = "qwen3-235b"
+    models = [
+        default_model,
+        "qwen3-32b",
+        "deepseek-r1-0528",
+        "deepseek-v3-0324",
+        "llama-4-maverick",
+    ]
+    model_aliases = {
+        "qwen-3-235b": "qwen3-235b",
+        "deepseek-r1": "deepseek-r1-0528",
+        "deepseek-v3": "deepseek-v3-0324",
+    }
+
+    @classmethod
+    async def create_async_generator(
+        cls,
+        model: str,
+        messages: Messages,
+        proxy: str = None,
+        temperature: float = None,
+        max_tokens: int = None,
+        enable_thinking: bool = False,
+        **kwargs
+    ) -> AsyncResult:
+        """
+        Create an async generator for streaming chat responses.
+
+        Args:
+            model: The model name to use
+            messages: List of message dictionaries
+            proxy: Optional proxy URL
+            temperature: Optional temperature parameter
+            max_tokens: Optional max tokens parameter
+            enable_thinking: Enable the thinking/analysis channel
+            **kwargs: Additional arguments
+
+        Yields:
+            str: Content chunks from the response
+            Reasoning: Thinking content when enable_thinking is True
+        """
+        model = cls.get_model(model)
+
+        headers = {
+            "Accept": "application/x-ndjson",
+            "Content-Type": "application/json",
+            "User-Agent": "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36",
+            "Origin": cls.url,
+            "Referer": f"{cls.url}/",
+        }
+
+        payload = {
+            "model": model,
+            "messages": messages,
+        }
+
+        if temperature is not None:
+            payload["temperature"] = temperature
+        if max_tokens is not None:
+            payload["max_tokens"] = max_tokens
+        if enable_thinking:
+            payload["enableThinking"] = True
+
+        async with ClientSession(headers=headers) as session:
+            async with session.post(
+                cls.api_endpoint,
+                json=payload,
+                proxy=proxy
+            ) as response:
+                response.raise_for_status()
+
+                async for line_bytes in response.content:
+                    if not line_bytes:
+                        continue
+
+                    line = line_bytes.decode("utf-8").strip()
+                    if not line:
+                        continue
+
+                    try:
+                        data = json.loads(line)
+                        msg_type = data.get("type")
+
+                        if msg_type == "text":
+                            # Regular text content
+                            content = data.get("data")
+                            if content:
+                                yield content
+
+                        elif msg_type == "thinking":
+                            # Thinking/reasoning content
+                            content = data.get("data")
+                            if content:
+                                yield Reasoning(content)
+
+                        elif msg_type == "done":
+                            # Stream complete
+                            break
+
+                        # Ignore clusterInfo and blockUpdate messages
+                        # as they are for GPU cluster visualization only
+
+                    except json.JSONDecodeError:
+                        # Skip non-JSON lines
+                        continue
diff --git a/g4f/Provider/__init__.py b/g4f/Provider/__init__.py
index 826ac4c4..ea51e380 100644
--- a/g4f/Provider/__init__.py
+++ b/g4f/Provider/__init__.py
@@ -48,6 +48,7 @@ from .Copilot              import Copilot
 from .DeepInfra            import DeepInfra
 from .EasyChat             import EasyChat
 from .GLM                  import GLM
+from .GradientNetwork      import GradientNetwork
 from .LambdaChat           import LambdaChat
 from .Mintlify             import Mintlify
 from .OIVSCodeSer          import OIVSCodeSer2, OIVSCodeSer0501

From 07883bc9f07afc6c6633815f5e4c34ea64aeb170 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Sat, 29 Nov 2025 04:23:46 +0000
Subject: [PATCH 06/15] Initial plan


From da4d7d118dcffbf59a5a7c2a7f495e04bdfd9319 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Sat, 29 Nov 2025 04:24:09 +0000
Subject: [PATCH 07/15] Use StreamSession for proper line-by-line NDJSON
 parsing

Co-authored-by: hlohaus <983577+hlohaus@users.noreply.github.com>
---
 g4f/Provider/GradientNetwork.py | 16 +++++-----------
 1 file changed, 5 insertions(+), 11 deletions(-)

diff --git a/g4f/Provider/GradientNetwork.py b/g4f/Provider/GradientNetwork.py
index b9286d0b..d5dc1919 100644
--- a/g4f/Provider/GradientNetwork.py
+++ b/g4f/Provider/GradientNetwork.py
@@ -2,10 +2,9 @@ from __future__ import annotations
 
 import json
 
-from aiohttp import ClientSession
-
 from ..typing import AsyncResult, Messages
 from ..providers.response import Reasoning
+from ..requests import StreamSession
 from .base_provider import AsyncGeneratorProvider, ProviderModelMixin
 
 
@@ -58,7 +57,7 @@ class GradientNetwork(AsyncGeneratorProvider, ProviderModelMixin):
             proxy: Optional proxy URL
             temperature: Optional temperature parameter
             max_tokens: Optional max tokens parameter
-            enable_thinking: Enable the thinking/analysis channel
+            enable_thinking: Enable the thinking/analysis channel (maps to enableThinking in API)
             **kwargs: Additional arguments
 
         Yields:
@@ -87,19 +86,14 @@ class GradientNetwork(AsyncGeneratorProvider, ProviderModelMixin):
         if enable_thinking:
             payload["enableThinking"] = True
 
-        async with ClientSession(headers=headers) as session:
+        async with StreamSession(headers=headers, proxy=proxy) as session:
             async with session.post(
                 cls.api_endpoint,
                 json=payload,
-                proxy=proxy
             ) as response:
                 response.raise_for_status()
 
-                async for line_bytes in response.content:
-                    if not line_bytes:
-                        continue
-
-                    line = line_bytes.decode("utf-8").strip()
+                async for line in response.iter_lines():
                     if not line:
                         continue
 
@@ -127,5 +121,5 @@ class GradientNetwork(AsyncGeneratorProvider, ProviderModelMixin):
                         # as they are for GPU cluster visualization only
 
                     except json.JSONDecodeError:
-                        # Skip non-JSON lines
+                        # Skip non-JSON lines (may be partial data or empty)
                         continue

From f57663cbe81b9e93399aeb29bdfa39929f532146 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Sat, 29 Nov 2025 04:25:22 +0000
Subject: [PATCH 08/15] Address code review: pass enable_thinking value
 directly, explicit skip for cluster messages

Co-authored-by: hlohaus <983577+hlohaus@users.noreply.github.com>
---
 g4f/Provider/GradientNetwork.py | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/g4f/Provider/GradientNetwork.py b/g4f/Provider/GradientNetwork.py
index d5dc1919..b6aec90c 100644
--- a/g4f/Provider/GradientNetwork.py
+++ b/g4f/Provider/GradientNetwork.py
@@ -84,7 +84,7 @@ class GradientNetwork(AsyncGeneratorProvider, ProviderModelMixin):
         if max_tokens is not None:
             payload["max_tokens"] = max_tokens
         if enable_thinking:
-            payload["enableThinking"] = True
+            payload["enableThinking"] = enable_thinking
 
         async with StreamSession(headers=headers, proxy=proxy) as session:
             async with session.post(
@@ -117,8 +117,9 @@ class GradientNetwork(AsyncGeneratorProvider, ProviderModelMixin):
                             # Stream complete
                             break
 
-                        # Ignore clusterInfo and blockUpdate messages
-                        # as they are for GPU cluster visualization only
+                        elif msg_type in ("clusterInfo", "blockUpdate"):
+                            # Skip GPU cluster visualization messages
+                            continue
 
                     except json.JSONDecodeError:
                         # Skip non-JSON lines (may be partial data or empty)

From c364425250e3070eaf0fc63c45bd9c18c5c1ab4e Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Sat, 29 Nov 2025 04:32:32 +0000
Subject: [PATCH 09/15] Add BAAI_Ling provider for Ling-1T model

Co-authored-by: hlohaus <983577+hlohaus@users.noreply.github.com>
---
 g4f/Provider/hf_space/BAAI_Ling.py | 104 +++++++++++++++++++++++++++++
 g4f/Provider/hf_space/__init__.py  |   2 +
 2 files changed, 106 insertions(+)
 create mode 100644 g4f/Provider/hf_space/BAAI_Ling.py

diff --git a/g4f/Provider/hf_space/BAAI_Ling.py b/g4f/Provider/hf_space/BAAI_Ling.py
new file mode 100644
index 00000000..725fb75a
--- /dev/null
+++ b/g4f/Provider/hf_space/BAAI_Ling.py
@@ -0,0 +1,104 @@
+from __future__ import annotations
+
+import aiohttp
+import json
+import uuid
+
+from ...typing import AsyncResult, Messages
+from ...providers.response import JsonConversation, Reasoning
+from ...requests.raise_for_status import raise_for_status
+from ..base_provider import AsyncGeneratorProvider, ProviderModelMixin
+from ..helper import format_prompt, get_last_user_message
+from ... import debug
+
+class BAAI_Ling(AsyncGeneratorProvider, ProviderModelMixin):
+    label = "BAAI Ling"
+    url = "https://instspace-ling-playground.hf.space"
+    api_endpoint = f"{url}/gradio_api/queue/join"
+
+    working = True
+    supports_stream = True
+    supports_system_message = True
+    supports_message_history = False
+
+    default_model = "ling-1t"
+    models = [default_model]
+    model_aliases = {
+        "ling-1t": default_model,
+        "ling": default_model,
+    }
+
+    @classmethod
+    async def create_async_generator(
+        cls,
+        model: str,
+        messages: Messages,
+        proxy: str = None,
+        conversation: JsonConversation = None,
+        **kwargs
+    ) -> AsyncResult:
+        if conversation is None or not hasattr(conversation, 'session_hash'):
+            conversation = JsonConversation(session_hash=str(uuid.uuid4()).replace('-', '')[:12])
+
+        prompt = format_prompt(messages) if conversation is None else get_last_user_message(messages)
+
+        headers = {
+            'accept': '*/*',
+            'accept-language': 'en-US,en;q=0.9',
+            'content-type': 'application/json',
+            'origin': cls.url,
+            'referer': f'{cls.url}/',
+            'user-agent': 'Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/132.0.0.0 Safari/537.36'
+        }
+
+        payload = {
+            "data": [prompt],
+            "event_data": None,
+            "fn_index": 0,
+            "trigger_id": 5,
+            "session_hash": conversation.session_hash
+        }
+
+        async with aiohttp.ClientSession() as session:
+            async with session.post(cls.api_endpoint, headers=headers, json=payload, proxy=proxy) as response:
+                await raise_for_status(response)
+                await response.json()
+
+            data_url = f'{cls.url}/gradio_api/queue/data?session_hash={conversation.session_hash}'
+            headers_data = {
+                'accept': 'text/event-stream',
+                'referer': f'{cls.url}/',
+                'user-agent': 'Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/132.0.0.0 Safari/537.36'
+            }
+
+            async with session.get(data_url, headers=headers_data, proxy=proxy) as response:
+                full_response = ""
+                async for line in response.content:
+                    decoded_line = line.decode('utf-8')
+                    if decoded_line.startswith('data: '):
+                        try:
+                            json_data = json.loads(decoded_line[6:])
+
+                            if json_data.get('msg') == 'process_generating':
+                                if 'output' in json_data and 'data' in json_data['output']:
+                                    output_data = json_data['output']['data']
+                                    if output_data and len(output_data) > 0:
+                                        text = output_data[0]
+                                        if isinstance(text, str) and text.startswith(full_response):
+                                            yield text[len(full_response):]
+                                            full_response = text
+                                        elif isinstance(text, str):
+                                            yield text
+                                            full_response = text
+
+                            elif json_data.get('msg') == 'process_completed':
+                                if 'output' in json_data and 'data' in json_data['output']:
+                                    output_data = json_data['output']['data']
+                                    if output_data and len(output_data) > 0:
+                                        final_text = output_data[0]
+                                        if isinstance(final_text, str) and len(final_text) > len(full_response):
+                                            yield final_text[len(full_response):]
+                                break
+
+                        except json.JSONDecodeError:
+                            debug.log("Could not parse JSON:", decoded_line)
diff --git a/g4f/Provider/hf_space/__init__.py b/g4f/Provider/hf_space/__init__.py
index 150467b0..3ad57263 100644
--- a/g4f/Provider/hf_space/__init__.py
+++ b/g4f/Provider/hf_space/__init__.py
@@ -6,6 +6,7 @@ from ...typing import AsyncResult, Messages, MediaListType
 from ...errors import ResponseError
 from ..base_provider import AsyncGeneratorProvider, ProviderModelMixin
 
+from .BAAI_Ling                       import BAAI_Ling
 from .BlackForestLabs_Flux1Dev       import BlackForestLabs_Flux1Dev
 from .BlackForestLabs_Flux1KontextDev import BlackForestLabs_Flux1KontextDev
 from .CohereForAI_C4AI_Command       import CohereForAI_C4AI_Command
@@ -27,6 +28,7 @@ class HuggingSpace(AsyncGeneratorProvider, ProviderModelMixin):
     default_image_model = BlackForestLabs_Flux1Dev.default_model
     default_vision_model = Microsoft_Phi_4_Multimodal.default_model
     providers = [
+        BAAI_Ling,
         BlackForestLabs_Flux1Dev,
         BlackForestLabs_Flux1KontextDev,
         CohereForAI_C4AI_Command,

From 04e300d7a6ae4bfaa0edaa71241a59a7457827d9 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Sat, 29 Nov 2025 04:35:54 +0000
Subject: [PATCH 10/15] Fix code review issues in BAAI_Ling provider

Co-authored-by: hlohaus <983577+hlohaus@users.noreply.github.com>
---
 g4f/Provider/hf_space/BAAI_Ling.py | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/g4f/Provider/hf_space/BAAI_Ling.py b/g4f/Provider/hf_space/BAAI_Ling.py
index 725fb75a..dbd1fbb2 100644
--- a/g4f/Provider/hf_space/BAAI_Ling.py
+++ b/g4f/Provider/hf_space/BAAI_Ling.py
@@ -5,7 +5,7 @@ import json
 import uuid
 
 from ...typing import AsyncResult, Messages
-from ...providers.response import JsonConversation, Reasoning
+from ...providers.response import JsonConversation
 from ...requests.raise_for_status import raise_for_status
 from ..base_provider import AsyncGeneratorProvider, ProviderModelMixin
 from ..helper import format_prompt, get_last_user_message
@@ -22,11 +22,10 @@ class BAAI_Ling(AsyncGeneratorProvider, ProviderModelMixin):
     supports_message_history = False
 
     default_model = "ling-1t"
-    models = [default_model]
     model_aliases = {
-        "ling-1t": default_model,
         "ling": default_model,
     }
+    models = [default_model]
 
     @classmethod
     async def create_async_generator(
@@ -37,10 +36,11 @@ class BAAI_Ling(AsyncGeneratorProvider, ProviderModelMixin):
         conversation: JsonConversation = None,
         **kwargs
     ) -> AsyncResult:
-        if conversation is None or not hasattr(conversation, 'session_hash'):
+        is_new_conversation = conversation is None or not hasattr(conversation, 'session_hash')
+        if is_new_conversation:
             conversation = JsonConversation(session_hash=str(uuid.uuid4()).replace('-', '')[:12])
 
-        prompt = format_prompt(messages) if conversation is None else get_last_user_message(messages)
+        prompt = format_prompt(messages) if is_new_conversation else get_last_user_message(messages)
 
         headers = {
             'accept': '*/*',
@@ -62,6 +62,7 @@ class BAAI_Ling(AsyncGeneratorProvider, ProviderModelMixin):
         async with aiohttp.ClientSession() as session:
             async with session.post(cls.api_endpoint, headers=headers, json=payload, proxy=proxy) as response:
                 await raise_for_status(response)
+                # Response body must be consumed for the request to complete
                 await response.json()
 
             data_url = f'{cls.url}/gradio_api/queue/data?session_hash={conversation.session_hash}'

From 098b2401eae4afe654a3934fd80e1c5996433b48 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Sat, 29 Nov 2025 04:36:25 +0000
Subject: [PATCH 11/15] Fix response parsing: use type "reply" with
 data.content/reasoningContent, update models

Co-authored-by: hlohaus <983577+hlohaus@users.noreply.github.com>
---
 g4f/Provider/GradientNetwork.py | 38 +++++++++++++--------------------
 1 file changed, 15 insertions(+), 23 deletions(-)

diff --git a/g4f/Provider/GradientNetwork.py b/g4f/Provider/GradientNetwork.py
index b6aec90c..e253a4bb 100644
--- a/g4f/Provider/GradientNetwork.py
+++ b/g4f/Provider/GradientNetwork.py
@@ -11,7 +11,7 @@ from .base_provider import AsyncGeneratorProvider, ProviderModelMixin
 class GradientNetwork(AsyncGeneratorProvider, ProviderModelMixin):
     """
     Provider for chat.gradient.network
-    Supports streaming text generation with various Qwen models.
+    Supports streaming text generation with Qwen and GPT OSS models.
     """
     label = "Gradient Network"
     url = "https://chat.gradient.network"
@@ -23,18 +23,15 @@ class GradientNetwork(AsyncGeneratorProvider, ProviderModelMixin):
     supports_system_message = True
     supports_message_history = True
 
-    default_model = "qwen3-235b"
+    default_model = "Qwen3 235B"
     models = [
         default_model,
-        "qwen3-32b",
-        "deepseek-r1-0528",
-        "deepseek-v3-0324",
-        "llama-4-maverick",
+        "GPT OSS 120B",
     ]
     model_aliases = {
-        "qwen-3-235b": "qwen3-235b",
-        "deepseek-r1": "deepseek-r1-0528",
-        "deepseek-v3": "deepseek-v3-0324",
+        "qwen-3-235b": "Qwen3 235B",
+        "qwen3-235b": "Qwen3 235B",
+        "gpt-oss-120b": "GPT OSS 120B",
     }
 
     @classmethod
@@ -62,7 +59,7 @@ class GradientNetwork(AsyncGeneratorProvider, ProviderModelMixin):
 
         Yields:
             str: Content chunks from the response
-            Reasoning: Thinking content when enable_thinking is True
+            Reasoning: Reasoning content when enable_thinking is True
         """
         model = cls.get_model(model)
 
@@ -101,22 +98,17 @@ class GradientNetwork(AsyncGeneratorProvider, ProviderModelMixin):
                         data = json.loads(line)
                         msg_type = data.get("type")
 
-                        if msg_type == "text":
-                            # Regular text content
-                            content = data.get("data")
+                        if msg_type == "reply":
+                            # Response chunks with content or reasoningContent
+                            reply_data = data.get("data", {})
+                            content = reply_data.get("content")
+                            reasoning_content = reply_data.get("reasoningContent")
+
+                            if reasoning_content:
+                                yield Reasoning(reasoning_content)
                             if content:
                                 yield content
 
-                        elif msg_type == "thinking":
-                            # Thinking/reasoning content
-                            content = data.get("data")
-                            if content:
-                                yield Reasoning(content)
-
-                        elif msg_type == "done":
-                            # Stream complete
-                            break
-
                         elif msg_type in ("clusterInfo", "blockUpdate"):
                             # Skip GPU cluster visualization messages
                             continue

From 21113c51a64d2e6804cd787e1dd918c2c68e33cf Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Sat, 29 Nov 2025 04:39:45 +0000
Subject: [PATCH 12/15] Remove redundant continue statement for cluster message
 handling

Co-authored-by: hlohaus <983577+hlohaus@users.noreply.github.com>
---
 g4f/Provider/GradientNetwork.py | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/g4f/Provider/GradientNetwork.py b/g4f/Provider/GradientNetwork.py
index e253a4bb..2a7f20a8 100644
--- a/g4f/Provider/GradientNetwork.py
+++ b/g4f/Provider/GradientNetwork.py
@@ -109,9 +109,7 @@ class GradientNetwork(AsyncGeneratorProvider, ProviderModelMixin):
                             if content:
                                 yield content
 
-                        elif msg_type in ("clusterInfo", "blockUpdate"):
-                            # Skip GPU cluster visualization messages
-                            continue
+                        # Skip clusterInfo and blockUpdate GPU visualization messages
 
                     except json.JSONDecodeError:
                         # Skip non-JSON lines (may be partial data or empty)

From 1fd9b8d116804d095146e4065f0ad916bb7a89da Mon Sep 17 00:00:00 2001
From: hlohaus <983577+hlohaus@users.noreply.github.com>
Date: Sun, 30 Nov 2025 11:20:29 +0100
Subject: [PATCH 13/15] Refactor GradientNetwork and ItalyGPT providers; update
 BAAI_Ling for improved functionality and model handling

---
 g4f/Provider/GradientNetwork.py    | 23 ++++----------
 g4f/Provider/ItalyGPT.py           |  7 ++--
 g4f/Provider/__init__.py           |  1 +
 g4f/Provider/hf_space/BAAI_Ling.py | 51 +++++++++++++++++-------------
 4 files changed, 41 insertions(+), 41 deletions(-)

diff --git a/g4f/Provider/GradientNetwork.py b/g4f/Provider/GradientNetwork.py
index 2a7f20a8..679adfb2 100644
--- a/g4f/Provider/GradientNetwork.py
+++ b/g4f/Provider/GradientNetwork.py
@@ -3,7 +3,7 @@ from __future__ import annotations
 import json
 
 from ..typing import AsyncResult, Messages
-from ..providers.response import Reasoning
+from ..providers.response import Reasoning, JsonResponse
 from ..requests import StreamSession
 from .base_provider import AsyncGeneratorProvider, ProviderModelMixin
 
@@ -23,7 +23,7 @@ class GradientNetwork(AsyncGeneratorProvider, ProviderModelMixin):
     supports_system_message = True
     supports_message_history = True
 
-    default_model = "Qwen3 235B"
+    default_model = "GPT OSS 120B"
     models = [
         default_model,
         "GPT OSS 120B",
@@ -40,9 +40,7 @@ class GradientNetwork(AsyncGeneratorProvider, ProviderModelMixin):
         model: str,
         messages: Messages,
         proxy: str = None,
-        temperature: float = None,
-        max_tokens: int = None,
-        enable_thinking: bool = False,
+        enable_thinking: bool = True,
         **kwargs
     ) -> AsyncResult:
         """
@@ -52,8 +50,6 @@ class GradientNetwork(AsyncGeneratorProvider, ProviderModelMixin):
             model: The model name to use
             messages: List of message dictionaries
             proxy: Optional proxy URL
-            temperature: Optional temperature parameter
-            max_tokens: Optional max tokens parameter
             enable_thinking: Enable the thinking/analysis channel (maps to enableThinking in API)
             **kwargs: Additional arguments
 
@@ -66,24 +62,18 @@ class GradientNetwork(AsyncGeneratorProvider, ProviderModelMixin):
         headers = {
             "Accept": "application/x-ndjson",
             "Content-Type": "application/json",
-            "User-Agent": "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36",
             "Origin": cls.url,
             "Referer": f"{cls.url}/",
         }
 
         payload = {
+            "clusterMode": "nvidia" if "GPT OSS" in model else "hybrid",
             "model": model,
             "messages": messages,
         }
-
-        if temperature is not None:
-            payload["temperature"] = temperature
-        if max_tokens is not None:
-            payload["max_tokens"] = max_tokens
         if enable_thinking:
             payload["enableThinking"] = enable_thinking
-
-        async with StreamSession(headers=headers, proxy=proxy) as session:
+        async with StreamSession(headers=headers, proxy=proxy, impersonate="chrome") as session:
             async with session.post(
                 cls.api_endpoint,
                 json=payload,
@@ -96,6 +86,7 @@ class GradientNetwork(AsyncGeneratorProvider, ProviderModelMixin):
 
                     try:
                         data = json.loads(line)
+                        yield JsonResponse.from_dict(data)
                         msg_type = data.get("type")
 
                         if msg_type == "reply":
@@ -113,4 +104,4 @@ class GradientNetwork(AsyncGeneratorProvider, ProviderModelMixin):
 
                     except json.JSONDecodeError:
                         # Skip non-JSON lines (may be partial data or empty)
-                        continue
+                        raise
diff --git a/g4f/Provider/ItalyGPT.py b/g4f/Provider/ItalyGPT.py
index e13fc0ee..43e9eba0 100644
--- a/g4f/Provider/ItalyGPT.py
+++ b/g4f/Provider/ItalyGPT.py
@@ -1,5 +1,6 @@
 from .base_provider import AsyncGeneratorProvider, ProviderModelMixin
 from ..typing import AsyncResult, Messages
+from ..requests import DEFAULT_HEADERS
 from aiohttp import ClientSession
 
 class ItalyGPT(AsyncGeneratorProvider, ProviderModelMixin):
@@ -23,10 +24,10 @@ class ItalyGPT(AsyncGeneratorProvider, ProviderModelMixin):
     ) -> AsyncResult:
         model = cls.get_model(model)
         headers = {
+            **DEFAULT_HEADERS,
             "content-type": "application/json",
             "origin": "https://italygpt.it",
             "referer": "https://italygpt.it/",
-            "user-agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/114.0.0.0 Safari/537.36"
         }
         payload = {
             "messages": messages,
@@ -34,12 +35,12 @@ class ItalyGPT(AsyncGeneratorProvider, ProviderModelMixin):
         }
         async with ClientSession() as session:
             async with session.post(
-                f"{cls.url}/api/chat/",
+                f"{cls.url}/api/chat",
                 json=payload,
                 headers=headers,
                 proxy=proxy,
             ) as resp:
                 resp.raise_for_status()
-                async for chunk in resp.content:
+                async for chunk in resp.content.iter_any():
                     if chunk:
                         yield chunk.decode()
diff --git a/g4f/Provider/__init__.py b/g4f/Provider/__init__.py
index ea51e380..1f97f9ce 100644
--- a/g4f/Provider/__init__.py
+++ b/g4f/Provider/__init__.py
@@ -49,6 +49,7 @@ from .DeepInfra            import DeepInfra
 from .EasyChat             import EasyChat
 from .GLM                  import GLM
 from .GradientNetwork      import GradientNetwork
+from .ItalyGPT             import ItalyGPT
 from .LambdaChat           import LambdaChat
 from .Mintlify             import Mintlify
 from .OIVSCodeSer          import OIVSCodeSer2, OIVSCodeSer0501
diff --git a/g4f/Provider/hf_space/BAAI_Ling.py b/g4f/Provider/hf_space/BAAI_Ling.py
index dbd1fbb2..cff02dd0 100644
--- a/g4f/Provider/hf_space/BAAI_Ling.py
+++ b/g4f/Provider/hf_space/BAAI_Ling.py
@@ -8,12 +8,12 @@ from ...typing import AsyncResult, Messages
 from ...providers.response import JsonConversation
 from ...requests.raise_for_status import raise_for_status
 from ..base_provider import AsyncGeneratorProvider, ProviderModelMixin
-from ..helper import format_prompt, get_last_user_message
+from ..helper import format_prompt, get_last_user_message, get_system_prompt
 from ... import debug
 
 class BAAI_Ling(AsyncGeneratorProvider, ProviderModelMixin):
-    label = "BAAI Ling"
-    url = "https://instspace-ling-playground.hf.space"
+    label = "Ling & Ring Playground"
+    url = "https://cafe3310-ling-playground.hf.space"
     api_endpoint = f"{url}/gradio_api/queue/join"
 
     working = True
@@ -25,7 +25,7 @@ class BAAI_Ling(AsyncGeneratorProvider, ProviderModelMixin):
     model_aliases = {
         "ling": default_model,
     }
-    models = [default_model]
+    models = ['ling-mini-2.0', 'ling-1t', 'ling-flash-2.0', 'ring-1t', 'ring-flash-2.0', 'ring-mini-2.0']
 
     @classmethod
     async def create_async_generator(
@@ -40,6 +40,7 @@ class BAAI_Ling(AsyncGeneratorProvider, ProviderModelMixin):
         if is_new_conversation:
             conversation = JsonConversation(session_hash=str(uuid.uuid4()).replace('-', '')[:12])
 
+        model = cls.get_model(model)
         prompt = format_prompt(messages) if is_new_conversation else get_last_user_message(messages)
 
         headers = {
@@ -52,10 +53,21 @@ class BAAI_Ling(AsyncGeneratorProvider, ProviderModelMixin):
         }
 
         payload = {
-            "data": [prompt],
+            "data": [
+                prompt,
+                [
+                    [
+                        None,
+                        "Hello! I'm Ling. Try selecting a scenario and a message example below to get started."
+                    ]
+                ],
+                get_system_prompt(messages),
+                1,
+                model
+            ],
             "event_data": None,
-            "fn_index": 0,
-            "trigger_id": 5,
+            "fn_index": 11,
+            "trigger_id": 14,
             "session_hash": conversation.session_hash
         }
 
@@ -79,27 +91,22 @@ class BAAI_Ling(AsyncGeneratorProvider, ProviderModelMixin):
                     if decoded_line.startswith('data: '):
                         try:
                             json_data = json.loads(decoded_line[6:])
-
                             if json_data.get('msg') == 'process_generating':
                                 if 'output' in json_data and 'data' in json_data['output']:
                                     output_data = json_data['output']['data']
                                     if output_data and len(output_data) > 0:
-                                        text = output_data[0]
-                                        if isinstance(text, str) and text.startswith(full_response):
-                                            yield text[len(full_response):]
-                                            full_response = text
-                                        elif isinstance(text, str):
-                                            yield text
-                                            full_response = text
+                                        parts = output_data[0][0]
+                                        if len(parts) == 2:
+                                            new_text = output_data[0][1].pop()
+                                            full_response += new_text
+                                            yield new_text
+                                        if len(parts) > 2:
+                                            new_text = parts[2]
+                                            full_response += new_text
+                                            yield new_text
 
                             elif json_data.get('msg') == 'process_completed':
-                                if 'output' in json_data and 'data' in json_data['output']:
-                                    output_data = json_data['output']['data']
-                                    if output_data and len(output_data) > 0:
-                                        final_text = output_data[0]
-                                        if isinstance(final_text, str) and len(final_text) > len(full_response):
-                                            yield final_text[len(full_response):]
-                                break
+                               break
 
                         except json.JSONDecodeError:
                             debug.log("Could not parse JSON:", decoded_line)

From 468dc7bd671306fb01f34b6ebc0525c0e4ede66b Mon Sep 17 00:00:00 2001
From: hlohaus <983577+hlohaus@users.noreply.github.com>
Date: Sun, 30 Nov 2025 22:30:14 +0100
Subject: [PATCH 14/15] Add "gemini-3-pro-preview" model to GeminiCLI provider

---
 g4f/Provider/needs_auth/GeminiCLI.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/g4f/Provider/needs_auth/GeminiCLI.py b/g4f/Provider/needs_auth/GeminiCLI.py
index 8797a1f4..09aec55c 100644
--- a/g4f/Provider/needs_auth/GeminiCLI.py
+++ b/g4f/Provider/needs_auth/GeminiCLI.py
@@ -500,6 +500,7 @@ class GeminiCLI(AsyncGeneratorProvider, ProviderModelMixin):
     models = [
         "gemini-2.5-pro",
         "gemini-2.5-flash",
+        "gemini-3-pro-preview"
     ]
 
     working = True

From 957d73a76ece5284c1920431bfccea7edf96a502 Mon Sep 17 00:00:00 2001
From: hlohaus <983577+hlohaus@users.noreply.github.com>
Date: Fri, 5 Dec 2025 23:34:10 +0100
Subject: [PATCH 15/15] Add UTF-8 encoding to file writes in Backend_Api class

---
 g4f/gui/server/backend_api.py | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/g4f/gui/server/backend_api.py b/g4f/gui/server/backend_api.py
index 2d509a06..a391349d 100644
--- a/g4f/gui/server/backend_api.py
+++ b/g4f/gui/server/backend_api.py
@@ -440,7 +440,7 @@ class Backend_Api(Api):
                     os.remove(copyfile)
                     continue
                 if not is_media and result:
-                    with open(os.path.join(bucket_dir, f"{filename}.md"), 'w') as f:
+                    with open(os.path.join(bucket_dir, f"{filename}.md"), 'w', encoding="utf-8") as f:
                         f.write(f"{result}\n")
                     filenames.append(f"{filename}.md")
                 if is_media:
@@ -477,7 +477,7 @@ class Backend_Api(Api):
                 except OSError:
                     shutil.copyfile(copyfile, newfile)
                     os.remove(copyfile)
-            with open(os.path.join(bucket_dir, "files.txt"), 'w') as f:
+            with open(os.path.join(bucket_dir, "files.txt"), 'w', encoding="utf-8") as f:
                 for filename in filenames:
                     f.write(f"{filename}\n")
             return {"bucket_id": bucket_id, "files": filenames, "media": media}
@@ -572,7 +572,7 @@ class Backend_Api(Api):
             share_id = secure_filename(share_id)
             bucket_dir = get_bucket_dir(share_id)
             os.makedirs(bucket_dir, exist_ok=True)
-            with open(os.path.join(bucket_dir, "chat.json"), 'w') as f:
+            with open(os.path.join(bucket_dir, "chat.json"), 'w', encoding="utf-8") as f:
                 json.dump(chat_data, f)
             self.chat_cache[share_id] = updated
             return {"share_id": share_id}