diff options
author | kqlio67 <kqlio67@users.noreply.github.com> | 2024-10-11 08:33:30 +0200 |
---|---|---|
committer | kqlio67 <kqlio67@users.noreply.github.com> | 2024-10-11 08:33:30 +0200 |
commit | a9bc67362f2be529fe9165ebb13347195ba1ddcf (patch) | |
tree | 1a91836eaa94f14c18ad5d55f687ff8a2118c357 /g4f/Provider/nexra/NexraGeminiPro.py | |
parent | feat(g4f/Provider/__init__.py): add new providers and update imports (diff) | |
download | gpt4free-a9bc67362f2be529fe9165ebb13347195ba1ddcf.tar gpt4free-a9bc67362f2be529fe9165ebb13347195ba1ddcf.tar.gz gpt4free-a9bc67362f2be529fe9165ebb13347195ba1ddcf.tar.bz2 gpt4free-a9bc67362f2be529fe9165ebb13347195ba1ddcf.tar.lz gpt4free-a9bc67362f2be529fe9165ebb13347195ba1ddcf.tar.xz gpt4free-a9bc67362f2be529fe9165ebb13347195ba1ddcf.tar.zst gpt4free-a9bc67362f2be529fe9165ebb13347195ba1ddcf.zip |
Diffstat (limited to 'g4f/Provider/nexra/NexraGeminiPro.py')
-rw-r--r-- | g4f/Provider/nexra/NexraGeminiPro.py | 70 |
1 files changed, 43 insertions, 27 deletions
diff --git a/g4f/Provider/nexra/NexraGeminiPro.py b/g4f/Provider/nexra/NexraGeminiPro.py index a57daed4..651f7cb4 100644 --- a/g4f/Provider/nexra/NexraGeminiPro.py +++ b/g4f/Provider/nexra/NexraGeminiPro.py @@ -1,17 +1,25 @@ from __future__ import annotations -import json from aiohttp import ClientSession - -from ...typing import AsyncResult, Messages +import json from ..base_provider import AsyncGeneratorProvider, ProviderModelMixin from ..helper import format_prompt +from ...typing import AsyncResult, Messages class NexraGeminiPro(AsyncGeneratorProvider, ProviderModelMixin): label = "Nexra Gemini PRO" + url = "https://nexra.aryahcr.cc/documentation/gemini-pro/en" api_endpoint = "https://nexra.aryahcr.cc/api/chat/complements" - models = ['gemini-pro'] + working = True + supports_stream = True + + default_model = 'gemini-pro' + models = [default_model] + + @classmethod + def get_model(cls, model: str) -> str: + return cls.default_model @classmethod async def create_async_generator( @@ -19,34 +27,42 @@ class NexraGeminiPro(AsyncGeneratorProvider, ProviderModelMixin): model: str, messages: Messages, proxy: str = None, + stream: bool = False, + markdown: bool = False, **kwargs ) -> AsyncResult: + model = cls.get_model(model) + headers = { "Content-Type": "application/json" } + + data = { + "messages": [ + { + "role": "user", + "content": format_prompt(messages) + } + ], + "markdown": markdown, + "stream": stream, + "model": model + } + async with ClientSession(headers=headers) as session: - data = { - "messages": [ - {'role': 'assistant', 'content': ''}, - {'role': 'user', 'content': format_prompt(messages)} - ], - "markdown": False, - "stream": True, - "model": model - } async with session.post(cls.api_endpoint, json=data, proxy=proxy) as response: response.raise_for_status() - full_response = '' - async for line in response.content: - if line: - messages = line.decode('utf-8').split('\x1e') - for message_str in messages: - try: - message = json.loads(message_str) - if message.get('message'): - full_response = message['message'] - if message.get('finish'): - yield full_response.strip() - return - except json.JSONDecodeError: - pass + buffer = "" + async for chunk in response.content.iter_any(): + if chunk.strip(): # Check if chunk is not empty + buffer += chunk.decode() + while '\x1e' in buffer: + part, buffer = buffer.split('\x1e', 1) + if part.strip(): + try: + response_json = json.loads(part) + message = response_json.get("message", "") + if message: + yield message + except json.JSONDecodeError as e: + print(f"JSONDecodeError: {e}") |