gpt4free/g4f/Provider/needs_auth/Grok.py
kqlio67 9bac34fc88 refactor: reorganize providers and update model configurations
- Rename DeepInfraChat to DeepInfra across all files
- Move DeepInfra from needs_auth to main Provider directory
- Rename LMArenaBeta to LMArena throughout codebase
- Move search-related providers to new search subdirectory (GoogleSearch, SearXNG, YouTube)
- Move deprecated providers to not_working directory (Free2GPT, LegacyLMArena, PenguinAI, ImageLabs, har)
- Add new Mintlify provider with custom AI assistant implementation
- Update Anthropic provider with Claude 4 models and Opus 4.1 parameter handling
- Update Grok provider with Grok 4 models and improved streaming support
- Update GithubCopilot with expanded model list including o3-mini, o4-mini, gpt-5 previews
- Update LambdaChat default model from deepseek-r1 to deepseek-llama3.3-70b
- Update TeachAnything default model from gemini-1.5-pro to gemma
- Remove DeepInfra from needs_auth directory
- Update all model_map references from DeepInfraChat to DeepInfra
- Update all model_map references from LMArenaBeta to LMArena
- Add beta_headers support to Anthropic for special features
- Improve Mintlify provider with system prompt handling and streaming
- Update model configurations in models.py to reflect provider changes
2025-08-25 23:50:53 +03:00

276 lines
12 KiB
Python

from __future__ import annotations
import os
import json
import time
import asyncio
import uuid
from typing import Dict, Any, AsyncIterator
try:
import nodriver
except ImportError:
pass
from ...typing import Messages, AsyncResult
from ...providers.response import JsonConversation, Reasoning, ImagePreview, ImageResponse, TitleGeneration, AuthResult, RequestLogin
from ...requests import StreamSession, get_nodriver, DEFAULT_HEADERS, merge_cookies
from ...requests.raise_for_status import raise_for_status
from ...errors import MissingAuthError
from ..base_provider import AsyncAuthedProvider, ProviderModelMixin
from ..helper import format_prompt, get_last_user_message
class Conversation(JsonConversation):
def __init__(self,
conversation_id: str
) -> None:
self.conversation_id = conversation_id
class Grok(AsyncAuthedProvider, ProviderModelMixin):
label = "Grok AI"
url = "https://grok.com"
cookie_domain = ".grok.com"
assets_url = "https://assets.grok.com"
conversation_url = "https://grok.com/rest/app-chat/conversations"
needs_auth = True
working = True
# Updated to Grok 4 as default
default_model = "grok-4"
# Updated model list with latest Grok 4 and 3 models
models = [
# Grok 4 models
"grok-4",
"grok-4-heavy",
"grok-4-reasoning",
# Grok 3 models
"grok-3",
"grok-3-reasoning",
"grok-3-mini",
"grok-3-mini-reasoning",
# Legacy Grok 2 (still supported)
"grok-2",
"grok-2-image",
# Latest aliases
"grok-latest",
]
model_aliases = {
# Grok 3 aliases
"grok-3-thinking": "grok-3-reasoning",
"grok-3-r1": "grok-3-reasoning",
"grok-3-mini-thinking": "grok-3-mini-reasoning",
# Latest alias
"grok": "grok-latest",
}
@classmethod
async def on_auth_async(cls, proxy: str = None, **kwargs) -> AsyncIterator:
auth_result = AuthResult(headers=DEFAULT_HEADERS, impersonate="chrome")
auth_result.headers["referer"] = cls.url + "/"
browser, stop_browser = await get_nodriver(proxy=proxy)
yield RequestLogin(cls.__name__, os.environ.get("G4F_LOGIN_URL") or "")
try:
page = await browser.get(cls.url)
has_headers = False
def on_request(event: nodriver.cdp.network.RequestWillBeSent, page=None):
nonlocal has_headers
if event.request.url.startswith(cls.conversation_url + "/new"):
for key, value in event.request.headers.items():
auth_result.headers[key.lower()] = value
has_headers = True
await page.send(nodriver.cdp.network.enable())
page.add_handler(nodriver.cdp.network.RequestWillBeSent, on_request)
await page.reload()
auth_result.headers["user-agent"] = await page.evaluate("window.navigator.userAgent", return_by_value=True)
while True:
if has_headers:
break
textarea = await page.select("textarea", 180)
await textarea.send_keys("Hello")
await asyncio.sleep(1)
button = await page.select("button[type='submit']")
if button:
await button.click()
await asyncio.sleep(1)
auth_result.cookies = {}
for c in await page.send(nodriver.cdp.network.get_cookies([cls.url])):
auth_result.cookies[c.name] = c.value
await page.close()
finally:
stop_browser()
yield auth_result
@classmethod
async def _prepare_payload(cls, model: str, message: str) -> Dict[str, Any]:
# Map model names to API model names
api_model = "grok-latest"
if model in ["grok-4", "grok-4-heavy", "grok-4-reasoning"]:
api_model = model
elif model == "grok-3":
api_model = "grok-3"
elif model in ["grok-3-mini", "grok-3-mini-reasoning"]:
api_model = "grok-3-mini"
elif model == "grok-2":
api_model = "grok-2"
# Check if it's a reasoning model
is_reasoning = model.endswith("-reasoning") or model.endswith("-thinking") or model.endswith("-r1")
# Enable Big Brain mode for heavy models
enable_big_brain = "heavy" in model or "big-brain" in model
# Enable DeepSearch for Grok 3+ models
enable_deep_search = not model.startswith("grok-2")
return {
"temporary": True,
"modelName": api_model,
"message": message,
"fileAttachments": [],
"imageAttachments": [],
"disableSearch": False,
"enableImageGeneration": model == "grok-2-image" or model == "grok-4",
"returnImageBytes": False,
"returnRawGrokInXaiRequest": False,
"enableImageStreaming": True,
"imageGenerationCount": 2,
"forceConcise": False,
"toolOverrides": {},
"enableSideBySide": True,
"isPreset": False,
"sendFinalMetadata": True,
"customInstructions": "",
"deepsearchPreset": "enabled" if enable_deep_search else "",
"isReasoning": is_reasoning,
"enableBigBrain": enable_big_brain,
"enableLiveSearch": False, # Real-time search for Grok 4
"contextWindow": 256000 if model.startswith("grok-4") else 131072, # 256k for Grok 4, 128k for others
}
@classmethod
async def create_authed(
cls,
model: str,
messages: Messages,
auth_result: AuthResult,
conversation: Conversation = None,
**kwargs
) -> AsyncResult:
conversation_id = None if conversation is None else conversation.conversation_id
prompt = format_prompt(messages) if conversation_id is None else get_last_user_message(messages)
async with StreamSession(
**auth_result.get_dict()
) as session:
payload = await cls._prepare_payload(model, prompt)
# Add voice mode support flag (for future use)
if kwargs.get("enable_voice", False):
payload["enableVoiceMode"] = True
if conversation_id is None:
url = f"{cls.conversation_url}/new"
else:
url = f"{cls.conversation_url}/{conversation_id}/responses"
async with session.post(url, json=payload, headers={"x-xai-request-id": str(uuid.uuid4())}) as response:
if response.status == 403:
raise MissingAuthError("Invalid secrets")
auth_result.cookies = merge_cookies(auth_result.cookies, response)
await raise_for_status(response)
thinking_duration = None
deep_search_active = False
async for line in response.iter_lines():
if line:
try:
json_data = json.loads(line)
result = json_data.get("result", {})
if conversation_id is None:
conversation_id = result.get("conversation", {}).get("conversationId")
response_data = result.get("response", {})
# Handle DeepSearch status
deep_search = response_data.get("deepSearchStatus")
if deep_search:
if not deep_search_active:
deep_search_active = True
yield Reasoning(status="🔍 Deep searching...")
if deep_search.get("completed"):
deep_search_active = False
yield Reasoning(status="Deep search completed")
# Handle image generation (Aurora for Grok 3+)
image = response_data.get("streamingImageGenerationResponse", None)
if image is not None:
image_url = image.get("imageUrl")
if image_url:
yield ImagePreview(
f'{cls.assets_url}/{image_url}',
"",
{"cookies": auth_result.cookies, "headers": auth_result.headers}
)
# Handle text tokens
token = response_data.get("token", result.get("token"))
is_thinking = response_data.get("isThinking", result.get("isThinking"))
if token:
if is_thinking:
if thinking_duration is None:
thinking_duration = time.time()
# Different status for different models
if "grok-4" in model:
status = "🧠 Grok 4 is processing..."
elif "big-brain" in payload and payload["enableBigBrain"]:
status = "🧠 Big Brain mode active..."
else:
status = "🤔 Is thinking..."
yield Reasoning(status=status)
yield Reasoning(token)
else:
if thinking_duration is not None:
thinking_duration = time.time() - thinking_duration
status = f"Thought for {thinking_duration:.2f}s" if thinking_duration > 1 else ""
thinking_duration = None
yield Reasoning(status=status)
yield token
# Handle generated images
generated_images = response_data.get("modelResponse", {}).get("generatedImageUrls", None)
if generated_images:
yield ImageResponse(
[f'{cls.assets_url}/{image}' for image in generated_images],
"",
{"cookies": auth_result.cookies, "headers": auth_result.headers}
)
# Handle title generation
title = result.get("title", {}).get("newTitle", "")
if title:
yield TitleGeneration(title)
# Handle tool usage information (Grok 4)
tool_usage = response_data.get("toolUsage")
if tool_usage:
tools_used = tool_usage.get("toolsUsed", [])
if tools_used:
yield Reasoning(status=f"Used tools: {', '.join(tools_used)}")
except json.JSONDecodeError:
continue
# Return conversation ID for continuation
if conversation_id is not None and kwargs.get("return_conversation", False):
yield Conversation(conversation_id)