webscout 8.2.2__py3-none-any.whl → 2026.1.19__py3-none-any.whl

webscout/Provider/OPENAI/gradient.py

@@ -0,0 +1,448 @@
+"""
+Gradient Network OpenAI-compatible Provider
+Reverse engineered from https://chat.gradient.network/
+
+Provides OpenAI-compatible API interface for Gradient Network's distributed GPU clusters.
+"""
+
+import json
+import time
+import uuid
+from typing import Any, Dict, Generator, List, Optional, Union, cast
+
+import requests
+
+from webscout.Provider.OPENAI.base import (
+    BaseChat,
+    BaseCompletions,
+    OpenAICompatibleProvider,
+    SimpleModelList,
+)
+from webscout.Provider.OPENAI.utils import (
+    ChatCompletion,
+    ChatCompletionChunk,
+    ChatCompletionMessage,
+    Choice,
+    ChoiceDelta,
+    CompletionUsage,
+    count_tokens,
+)
+
+# ANSI escape codes for formatting
+BOLD = "\033[1m"
+RED = "\033[91m"
+RESET = "\033[0m"
+
+
+class Completions(BaseCompletions):
+    """Handles chat completions for Gradient Network."""
+
+    def __init__(self, client: "Gradient"):
+        self._client = client
+
+    def create(
+        self,
+        *,
+        model: str,
+        messages: List[Dict[str, str]],
+        max_tokens: Optional[int] = 2048,
+        stream: bool = False,
+        temperature: Optional[float] = None,
+        top_p: Optional[float] = None,
+        timeout: Optional[int] = None,
+        proxies: Optional[Dict[str, str]] = None,
+        enable_thinking: Optional[bool] = None,
+        cluster_mode: Optional[str] = None,
+        **kwargs: Any,
+    ) -> Union[ChatCompletion, Generator[ChatCompletionChunk, None, None]]:
+        """
+        Creates a model response for the given chat conversation.
+        Mimics openai.chat.completions.create
+
+        Args:
+            model: The model to use for completion
+            messages: List of message dicts with 'role' and 'content'
+            max_tokens: Maximum tokens to generate (not used by Gradient API)
+            stream: Whether to stream the response
+            temperature: Sampling temperature (not used by Gradient API)
+            top_p: Top-p sampling (not used by Gradient API)
+            timeout: Request timeout in seconds
+            proxies: Proxy configuration
+            enable_thinking: Enable thinking/reasoning mode (default: client setting)
+            cluster_mode: GPU cluster mode (auto-detected based on model if None)
+            **kwargs: Additional arguments
+
+        Returns:
+            ChatCompletion or Generator[ChatCompletionChunk] depending on stream
+        """
+        # Convert model name and get appropriate cluster mode
+        converted_model = self._client.convert_model_name(model)
+        actual_cluster_mode = cluster_mode or self._client.MODEL_CLUSTERS.get(
+            converted_model, self._client.cluster_mode
+        )
+
+        # Build the payload - pass messages directly as the API accepts them
+        payload = {
+            "model": converted_model,
+            "clusterMode": actual_cluster_mode,
+            "messages": messages,
+            "enableThinking": enable_thinking
+            if enable_thinking is not None
+            else self._client.enable_thinking,
+        }
+
+        request_id = f"chatcmpl-{uuid.uuid4()}"
+        created_time = int(time.time())
+
+        if stream:
+            return self._create_stream(request_id, created_time, model, payload, timeout, proxies)
+        else:
+            return self._create_non_stream(
+                request_id, created_time, model, payload, timeout, proxies
+            )
+
+    def _create_stream(
+        self,
+        request_id: str,
+        created_time: int,
+        model: str,
+        payload: Dict[str, Any],
+        timeout: Optional[int] = None,
+        proxies: Optional[Dict[str, str]] = None,
+    ) -> Generator[ChatCompletionChunk, None, None]:
+        """Handle streaming response from Gradient API."""
+        try:
+            response = self._client.session.post(
+                self._client.base_url,
+                headers=self._client.headers,
+                json=payload,
+                stream=True,
+                timeout=timeout or self._client.timeout,
+                proxies=proxies or self._client.proxies or None,
+            )
+            response.raise_for_status()
+
+            completion_tokens = 0
+            prompt_tokens = count_tokens(str(payload.get("messages", [])))
+            first_chunk = True
+
+            for line in response.iter_lines():
+                if not line:
+                    continue
+
+                try:
+                    # Decode bytes to string
+                    decoded_line = line.decode("utf-8") if isinstance(line, bytes) else line
+                    decoded_line = decoded_line.strip()
+                    if not decoded_line:
+                        continue
+
+                    # Parse JSON response
+                    data = json.loads(decoded_line)
+
+                    # Only process "reply" type chunks
+                    chunk_type = data.get("type")
+                    if chunk_type != "reply":
+                        continue
+
+                    # Extract content - prefer "content" over "reasoningContent"
+                    reply_data = data.get("data", {})
+                    content = reply_data.get("content") or reply_data.get("reasoningContent")
+
+                    if content:
+                        completion_tokens += count_tokens(content)
+
+                        delta = ChoiceDelta(
+                            content=content, role="assistant" if first_chunk else None
+                        )
+                        first_chunk = False
+
+                        choice = Choice(index=0, delta=delta, finish_reason=None, logprobs=None)
+
+                        chunk = ChatCompletionChunk(
+                            id=request_id,
+                            choices=[choice],
+                            created=created_time,
+                            model=model,
+                            system_fingerprint=None,
+                        )
+                        chunk.usage = {
+                            "prompt_tokens": prompt_tokens,
+                            "completion_tokens": completion_tokens,
+                            "total_tokens": prompt_tokens + completion_tokens,
+                            "estimated_cost": None,
+                        }
+                        yield chunk
+
+                except json.JSONDecodeError:
+                    continue
+                except Exception:
+                    continue
+
+            # Final chunk with finish_reason="stop"
+            delta = ChoiceDelta(content=None, role=None, tool_calls=None)
+            choice = Choice(index=0, delta=delta, finish_reason="stop", logprobs=None)
+            chunk = ChatCompletionChunk(
+                id=request_id,
+                choices=[choice],
+                created=created_time,
+                model=model,
+                system_fingerprint=None,
+            )
+            chunk.usage = {
+                "prompt_tokens": prompt_tokens,
+                "completion_tokens": completion_tokens,
+                "total_tokens": prompt_tokens + completion_tokens,
+                "estimated_cost": None,
+            }
+            yield chunk
+
+        except requests.exceptions.RequestException as e:
+            print(f"{RED}Error during Gradient stream request: {e}{RESET}")
+            raise IOError(f"Gradient request failed: {e}") from e
+        except Exception as e:
+            print(f"{RED}Error during Gradient stream request: {type(e).__name__}: {e}{RESET}")
+            raise IOError(f"Gradient request failed: {e}") from e
+
+    def _create_non_stream(
+        self,
+        request_id: str,
+        created_time: int,
+        model: str,
+        payload: Dict[str, Any],
+        timeout: Optional[int] = None,
+        proxies: Optional[Dict[str, str]] = None,
+    ) -> ChatCompletion:
+        """Handle non-streaming response from Gradient API."""
+        try:
+            # Collect all chunks from streaming
+            full_content = ""
+            prompt_tokens = count_tokens(str(payload.get("messages", [])))
+
+            response = self._client.session.post(
+                self._client.base_url,
+                headers=self._client.headers,
+                json=payload,
+                stream=True,
+                timeout=timeout or self._client.timeout,
+                proxies=proxies or self._client.proxies or None,
+            )
+            response.raise_for_status()
+
+            for line in response.iter_lines():
+                if not line:
+                    continue
+
+                try:
+                    # Decode bytes to string
+                    decoded_line = line.decode("utf-8") if isinstance(line, bytes) else line
+                    decoded_line = decoded_line.strip()
+                    if not decoded_line:
+                        continue
+
+                    data = json.loads(decoded_line)
+
+                    # Only process "reply" type chunks
+                    chunk_type = data.get("type")
+                    if chunk_type != "reply":
+                        continue
+
+                    reply_data = data.get("data", {})
+                    # Prefer "content" over "reasoningContent"
+                    content = reply_data.get("content") or reply_data.get("reasoningContent")
+                    if content:
+                        full_content += content
+
+                except json.JSONDecodeError:
+                    continue
+                except Exception:
+                    continue
+
+            completion_tokens = count_tokens(full_content)
+            total_tokens = prompt_tokens + completion_tokens
+
+            usage = CompletionUsage(
+                prompt_tokens=prompt_tokens,
+                completion_tokens=completion_tokens,
+                total_tokens=total_tokens,
+            )
+
+            message = ChatCompletionMessage(role="assistant", content=full_content)
+
+            choice = Choice(index=0, message=message, finish_reason="stop")
+
+            completion = ChatCompletion(
+                id=request_id,
+                choices=[choice],
+                created=created_time,
+                model=model,
+                usage=usage,
+            )
+
+            return completion
+
+        except requests.exceptions.RequestException as e:
+            print(f"{RED}Error during Gradient non-stream request: {e}{RESET}")
+            raise IOError(f"Gradient request failed: {e}") from e
+        except Exception as e:
+            print(f"{RED}Error during Gradient non-stream request: {type(e).__name__}: {e}{RESET}")
+            raise IOError(f"Gradient request failed: {e}") from e
+
+
+class Chat(BaseChat):
+    """Chat interface for Gradient Network."""
+
+    def __init__(self, client: "Gradient"):
+        self.completions = Completions(client)
+
+
+class Gradient(OpenAICompatibleProvider):
+    """
+    OpenAI-compatible client for Gradient Network API.
+
+    Gradient Network provides access to distributed GPU clusters running large language models.
+    This provider supports real-time streaming responses.
+
+    Note: GPT OSS 120B works on "nvidia" cluster, Qwen3 235B works on "hybrid" cluster.
+    Cluster mode is auto-detected based on model selection.
+
+    Usage:
+        client = Gradient()
+        response = client.chat.completions.create(
+            model="GPT OSS 120B",
+            messages=[{"role": "user", "content": "Hello!"}]
+        )
+        print(response.choices[0].message.content)
+
+        # Streaming
+        for chunk in client.chat.completions.create(
+            model="Qwen3 235B",
+            messages=[{"role": "user", "content": "Tell me a story"}],
+            stream=True
+        ):
+            if chunk.choices[0].delta.content:
+                print(chunk.choices[0].delta.content, end="", flush=True)
+    """
+
+    required_auth = False
+
+    AVAILABLE_MODELS = [
+        "GPT OSS 120B",
+        "Qwen3 235B",
+    ]
+
+    # Model to cluster mapping
+    MODEL_CLUSTERS = {
+        "GPT OSS 120B": "nvidia",
+        "Qwen3 235B": "hybrid",
+    }
+
+    def __init__(
+        self,
+        timeout: int = 60,
+        cluster_mode: str = "nvidia",
+        enable_thinking: bool = True,
+        proxies: Optional[Dict[str, str]] = None,
+    ):
+        """
+        Initialize the Gradient client.
+
+        Args:
+            timeout: Request timeout in seconds (default: 60)
+            cluster_mode: Default GPU cluster mode (default: "nvidia", auto-detected per model)
+            enable_thinking: Enable thinking/reasoning mode (default: True)
+            proxies: Optional proxy configuration
+        """
+        self.timeout = timeout
+        self.cluster_mode = cluster_mode
+        self.enable_thinking = enable_thinking
+        self.proxies = proxies or {}
+
+        self.base_url = "https://chat.gradient.network/api/generate"
+        self.session = requests.Session()
+
+        # Set up headers matching the curl request
+        self.headers = {
+            "accept": "*/*",
+            "accept-language": "en-US,en;q=0.9,en-IN;q=0.8",
+            "content-type": "application/json",
+            "dnt": "1",
+            "origin": "https://chat.gradient.network",
+            "priority": "u=1, i",
+            "referer": "https://chat.gradient.network/",
+            "sec-ch-ua": '"Microsoft Edge";v="143", "Chromium";v="143", "Not A(Brand";v="24"',
+            "sec-ch-ua-mobile": "?0",
+            "sec-ch-ua-platform": '"Windows"',
+            "sec-fetch-dest": "empty",
+            "sec-fetch-mode": "cors",
+            "sec-fetch-site": "same-origin",
+            "sec-gpc": "1",
+            "user-agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/143.0.0.0 Safari/537.36 Edg/143.0.0.0",
+        }
+
+        self.session.headers.update(self.headers)
+
+        # Initialize the chat interface
+        self.chat = Chat(self)
+
+    def convert_model_name(self, model: str) -> str:
+        """
+        Ensure the model name is in the correct format.
+
+        Args:
+            model: Model name to convert
+
+        Returns:
+            Valid model name
+        """
+        if model in self.AVAILABLE_MODELS:
+            return model
+
+        # Try case-insensitive matching with dash to space conversion
+        model_lower = model.lower().replace("-", " ")
+        for available_model in self.AVAILABLE_MODELS:
+            if model_lower == available_model.lower():
+                return available_model
+
+        # Default to first available model
+        print(
+            f"{BOLD}Warning: Model '{model}' not found, using default model '{self.AVAILABLE_MODELS[0]}'{RESET}"
+        )
+        return self.AVAILABLE_MODELS[0]
+
+    @property
+    def models(self) -> SimpleModelList:
+        return SimpleModelList(type(self).AVAILABLE_MODELS)
+
+
+if __name__ == "__main__":
+    print("-" * 80)
+    print(f"{'Model':<50} {'Status':<10} {'Response'}")
+    print("-" * 80)
+
+    for model in Gradient.AVAILABLE_MODELS:
+        try:
+            client = Gradient(timeout=120)
+            response = client.chat.completions.create(
+                model=model,
+                messages=[
+                    {"role": "user", "content": "Say 'Hello' in one word"},
+                ],
+                stream=False,
+            )
+
+            if (
+                isinstance(response, ChatCompletion)
+                and response.choices
+                and response.choices[0].message
+                and response.choices[0].message.content
+            ):
+                status = "✓"
+                display_text = response.choices[0].message.content.strip()
+                display_text = display_text[:50] + "..." if len(display_text) > 50 else display_text
+            else:
+                status = "✗"
+                display_text = "Empty or invalid response"
+            print(f"{model:<50} {status:<10} {display_text}")
+        except Exception as e:
+            print(f"{model:<50} {'✗':<10} {str(e)[:50]}")