4 files changed, 91 insertions, 9 deletions

diff --git a/g4f/Provider/DeepInfra.py b/g4f/Provider/DeepInfra.py
new file mode 100644
index 00000000..70045cae
--- /dev/null
+++ b/g4f/Provider/DeepInfra.py
@@ -0,0 +1,63 @@
+from __future__ import annotations
+
+import json
+from aiohttp import ClientSession
+
+from ..typing import AsyncResult, Messages
+from .base_provider import AsyncGeneratorProvider
+
+
+class DeepInfra(AsyncGeneratorProvider):
+    url = "https://deepinfra.com"
+    working = True
+
+    @classmethod
+    async def create_async_generator(
+        cls,
+        model: str,
+        messages: Messages,
+        proxy: str = None,
+        **kwargs
+    ) -> AsyncResult:
+        if not model:
+            model = "meta-llama/Llama-2-70b-chat-hf"
+        headers = {
+            "User-Agent": "Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:109.0) Gecko/20100101 Firefox/118.0",
+            "Accept": "text/event-stream",
+            "Accept-Language": "de,en-US;q=0.7,en;q=0.3",
+            "Accept-Encoding": "gzip, deflate, br",
+            "Referer": f"{cls.url}/",
+            "Content-Type": "application/json",
+            "X-Deepinfra-Source": "web-page",
+            "Origin": cls.url,
+            "Connection": "keep-alive",
+            "Sec-Fetch-Dest": "empty",
+            "Sec-Fetch-Mode": "cors",
+            "Sec-Fetch-Site": "same-site",
+            "Pragma": "no-cache",
+            "Cache-Control": "no-cache",
+        }
+        async with ClientSession(headers=headers) as session:
+            data = {
+                "model": model,
+                "messages": messages,
+                "stream": True,
+            }
+            async with session.post(
+                "https://api.deepinfra.com/v1/openai/chat/completions",
+                json=data,
+                proxy=proxy
+            ) as response:
+                response.raise_for_status()
+                first = True
+                async for line in response.content:
+                    if line.startswith(b"data: [DONE]"):
+                        break
+                    elif line.startswith(b"data: "):
+                        chunk = json.loads(line[6:])["choices"][0]["delta"].get("content")
+                        if chunk:
+                            if first:
+                                chunk = chunk.lstrip()
+                                if chunk:
+                                    first = False
+                            yield chunk
\ No newline at end of file
diff --git a/g4f/Provider/Llama2.py b/g4f/Provider/Llama2.py
index b59fde12..1b332f86 100644
--- a/g4f/Provider/Llama2.py
+++ b/g4f/Provider/Llama2.py
@@ -6,15 +6,14 @@ from ..typing import AsyncResult, Messages
 from .base_provider import AsyncGeneratorProvider
 
 models = {
-    "7B": {"name": "Llama 2 7B", "version": "d24902e3fa9b698cc208b5e63136c4e26e828659a9f09827ca6ec5bb83014381", "shortened":"7B"},
-    "13B": {"name": "Llama 2 13B", "version": "9dff94b1bed5af738655d4a7cbcdcde2bd503aa85c94334fe1f42af7f3dd5ee3", "shortened":"13B"},
-    "70B": {"name": "Llama 2 70B", "version": "2796ee9483c3fd7aa2e171d38f4ca12251a30609463dcfd4cd76703f22e96cdf", "shortened":"70B"},
+    "meta-llama/Llama-2-7b-chat-hf": {"name": "Llama 2 7B", "version": "d24902e3fa9b698cc208b5e63136c4e26e828659a9f09827ca6ec5bb83014381", "shortened":"7B"},
+    "meta-llama/Llama-2-13b-chat-hf": {"name": "Llama 2 13B", "version": "9dff94b1bed5af738655d4a7cbcdcde2bd503aa85c94334fe1f42af7f3dd5ee3", "shortened":"13B"},
+    "meta-llama/Llama-2-70b-chat-hf": {"name": "Llama 2 70B", "version": "2796ee9483c3fd7aa2e171d38f4ca12251a30609463dcfd4cd76703f22e96cdf", "shortened":"70B"},
     "Llava": {"name": "Llava 13B", "version": "6bc1c7bb0d2a34e413301fee8f7cc728d2d4e75bfab186aa995f63292bda92fc", "shortened":"Llava"}
 }
 
 class Llama2(AsyncGeneratorProvider):
     url                   = "https://www.llama2.ai"
-    supports_gpt_35_turbo = True
     working               = True
 
     @classmethod
@@ -26,8 +25,8 @@ class Llama2(AsyncGeneratorProvider):
         **kwargs
     ) -> AsyncResult:
         if not model:
-            model = "70B"
-        if model not in models:
+            model = "meta-llama/Llama-2-70b-chat-hf"
+        elif model not in models:
             raise ValueError(f"Model are not supported: {model}")
         version = models[model]["version"]
         headers = {
@@ -54,7 +53,7 @@ class Llama2(AsyncGeneratorProvider):
                 "systemPrompt": kwargs.get("system_message", "You are a helpful assistant."),
                 "temperature": kwargs.get("temperature", 0.75),
                 "topP": kwargs.get("top_p", 0.9),
-                "maxTokens": kwargs.get("max_tokens", 1024),
+                "maxTokens": kwargs.get("max_tokens", 8000),
                 "image": None
             }
             started = False
@@ -68,9 +67,9 @@ class Llama2(AsyncGeneratorProvider):
             
 def format_prompt(messages: Messages):
     messages = [
-        f"[INST]{message['content']}[/INST]"
+        f"[INST] {message['content']} [/INST]"
         if message["role"] == "user"
         else message["content"]
         for message in messages
     ]
-    return "\n".join(messages)
\ No newline at end of file
+    return "\n".join(messages) + "\n"
\ No newline at end of file
diff --git a/g4f/Provider/__init__.py b/g4f/Provider/__init__.py
index 653b6026..60d3bd25 100644
--- a/g4f/Provider/__init__.py
+++ b/g4f/Provider/__init__.py
@@ -17,6 +17,7 @@ from .ChatgptFree     import ChatgptFree
 from .ChatgptLogin    import ChatgptLogin
 from .ChatgptX        import ChatgptX
 from .Cromicle        import Cromicle
+from .DeepInfra       import DeepInfra
 from .FakeGpt         import FakeGpt
 from .FreeGpt         import FreeGpt
 from .GPTalk          import GPTalk
@@ -70,6 +71,7 @@ class ProviderUtils:
         'ChatgptX': ChatgptX,
         'CodeLinkAva': CodeLinkAva,
         'Cromicle': Cromicle,
+        'DeepInfra': DeepInfra,
         'DfeHub': DfeHub,
         'EasyChat': EasyChat,
         'Equing': Equing,
@@ -144,6 +146,7 @@ __all__ = [
     'ChatgptLogin',
     'ChatgptX',
     'Cromicle',
+    'DeepInfra',
     'CodeLinkAva',
     'DfeHub',
     'EasyChat',
diff --git a/g4f/models.py b/g4f/models.py
index 7eee917a..45799602 100644
--- a/g4f/models.py
+++ b/g4f/models.py
@@ -6,12 +6,14 @@ from .Provider   import (
     GptForLove,
     ChatgptAi,
     GptChatly,
+    DeepInfra,
     ChatgptX,
     ChatBase,
     GeekGpt,
     FakeGpt,
     FreeGpt,
     NoowAi,
+    Llama2,
     Vercel, 
     Aichat,
     GPTalk,
@@ -74,6 +76,21 @@ gpt_4 = Model(
     ])
 )
 
+llama2_7b = Model(
+    name          = "meta-llama/Llama-2-7b-chat-hf",
+    base_provider = 'huggingface',
+    best_provider = RetryProvider([Llama2, DeepInfra]))
+
+llama2_13b = Model(
+    name          ="meta-llama/Llama-2-13b-chat-hf",
+    base_provider = 'huggingface',
+    best_provider = RetryProvider([Llama2, DeepInfra]))
+
+llama2_70b = Model(
+    name          = "meta-llama/Llama-2-70b-chat-hf",
+    base_provider = "huggingface",
+    best_provider = RetryProvider([Llama2, DeepInfra]))
+
 # Bard
 palm = Model(
     name          = 'palm',