Spaces:
Sleeping
Sleeping
Create groq_api.py
Browse files- groq_api.py +218 -0
groq_api.py
ADDED
|
@@ -0,0 +1,218 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os
|
| 2 |
+
import requests
|
| 3 |
+
from typing import List, Dict, Any
|
| 4 |
+
|
| 5 |
+
def grok_get_llm_response(
|
| 6 |
+
system_prompt: str,
|
| 7 |
+
user_input: str,
|
| 8 |
+
tools: List[Dict[str, Any]] = None,
|
| 9 |
+
tool_choice: str = "auto",
|
| 10 |
+
reasoning_effort: str = "default", #"default",json, text, verbose_json
|
| 11 |
+
response_format: Dict[str, Any] = None,
|
| 12 |
+
temperature: float = 0.3,
|
| 13 |
+
max_completion_tokens: int = 2000,
|
| 14 |
+
include_reasoning: str = False # True
|
| 15 |
+
) -> str:
|
| 16 |
+
"""
|
| 17 |
+
Make a request to the Grok API and return the response content, supporting tool usage and agentic features.
|
| 18 |
+
|
| 19 |
+
Args:
|
| 20 |
+
system_prompt (str): The system prompt to set the context.
|
| 21 |
+
user_input (str): The user input to process.
|
| 22 |
+
tools (List[Dict[str, Any]], optional): List of tool definitions for tool-calling.
|
| 23 |
+
tool_choice (str, optional): Controls tool usage ("none", "auto", "required"). Defaults to "auto".
|
| 24 |
+
reasoning_effort (str, optional): Reasoning mode for Qwen3 models ("none", "default"). Defaults to "default".
|
| 25 |
+
response_format (Dict[str, Any], optional): Format for structured outputs (e.g., JSON schema).
|
| 26 |
+
temperature (float, optional): Sampling temperature (0 to 2). Defaults to 0.7 for determinism.
|
| 27 |
+
max_completion_tokens (int, optional): Max tokens in response. Defaults to 1000.
|
| 28 |
+
|
| 29 |
+
Returns:
|
| 30 |
+
str: The content of the assistant's response or tool call results, or empty string on error.
|
| 31 |
+
"""
|
| 32 |
+
# Retrieve API key from environment
|
| 33 |
+
api_key = os.getenv("GROQ_API_KEY")
|
| 34 |
+
if not api_key:
|
| 35 |
+
print("Grok API error: GROQ_API_KEY environment variable not set")
|
| 36 |
+
return ""
|
| 37 |
+
|
| 38 |
+
# API endpoint
|
| 39 |
+
api_url = "https://api.groq.com/openai/v1/chat/completions"
|
| 40 |
+
|
| 41 |
+
# Prepare messages in Grok API format
|
| 42 |
+
messages = [
|
| 43 |
+
{"role": "system", "content": system_prompt},
|
| 44 |
+
{"role": "user", "content": user_input}
|
| 45 |
+
]
|
| 46 |
+
|
| 47 |
+
# Prepare payload
|
| 48 |
+
payload = {
|
| 49 |
+
"model": "qwen/qwen3-32b",
|
| 50 |
+
"messages": messages,
|
| 51 |
+
"temperature": max(0, min(temperature, 2)), # Clamp to valid range [0, 2]
|
| 52 |
+
"include_reasoning": include_reasoning,
|
| 53 |
+
"max_completion_tokens": max_completion_tokens
|
| 54 |
+
}
|
| 55 |
+
|
| 56 |
+
# Add tools and tool_choice if provided
|
| 57 |
+
if tools:
|
| 58 |
+
payload["tools"] = tools
|
| 59 |
+
if tool_choice in ["none", "auto", "required"]:
|
| 60 |
+
payload["tool_choice"] = tool_choice
|
| 61 |
+
else:
|
| 62 |
+
print(f"Grok API warning: Invalid tool_choice '{tool_choice}', defaulting to 'auto'")
|
| 63 |
+
payload["tool_choice"] = "auto"
|
| 64 |
+
|
| 65 |
+
# Add reasoning_effort for Qwen3 models
|
| 66 |
+
if reasoning_effort in ["none", "default"]:
|
| 67 |
+
payload["reasoning_effort"] = reasoning_effort
|
| 68 |
+
else:
|
| 69 |
+
print(f"Grok API warning: Invalid reasoning_effort '{reasoning_effort}', defaulting to 'default'")
|
| 70 |
+
payload["reasoning_effort"] = "default"
|
| 71 |
+
|
| 72 |
+
# Add response_format if provided
|
| 73 |
+
if response_format:
|
| 74 |
+
payload["response_format"] = response_format
|
| 75 |
+
|
| 76 |
+
# Set headers
|
| 77 |
+
headers = {
|
| 78 |
+
"Content-Type": "application/json",
|
| 79 |
+
"Authorization": f"Bearer {api_key}"
|
| 80 |
+
}
|
| 81 |
+
|
| 82 |
+
try:
|
| 83 |
+
# Make API request
|
| 84 |
+
response = requests.post(api_url, headers=headers, json=payload, timeout=60)
|
| 85 |
+
response.raise_for_status()
|
| 86 |
+
|
| 87 |
+
# Parse response
|
| 88 |
+
result = response.json()
|
| 89 |
+
choice = result.get("choices", [{}])[0]
|
| 90 |
+
message = choice.get("message", {})
|
| 91 |
+
|
| 92 |
+
# Handle tool calls if present
|
| 93 |
+
if "tool_calls" in message:
|
| 94 |
+
tool_calls = message["tool_calls"]
|
| 95 |
+
tool_results = []
|
| 96 |
+
for tool_call in tool_calls:
|
| 97 |
+
tool_name = tool_call.get("function", {}).get("name", "")
|
| 98 |
+
tool_args = tool_call.get("function", {}).get("arguments", "{}")
|
| 99 |
+
tool_results.append(f"Tool Call: {tool_name} with args {tool_args}")
|
| 100 |
+
return "; ".join(tool_results) # Combine tool call results into a single string
|
| 101 |
+
|
| 102 |
+
# Return assistant content if no tool calls
|
| 103 |
+
content = message.get("content", "")
|
| 104 |
+
return content.strip()
|
| 105 |
+
|
| 106 |
+
except requests.exceptions.HTTPError as e:
|
| 107 |
+
print(f"Grok API error: HTTP {e.response.status_code} - {e.response.text}")
|
| 108 |
+
return ""
|
| 109 |
+
except requests.exceptions.RequestException as e:
|
| 110 |
+
print(f"Grok API error: Network error - {e}")
|
| 111 |
+
return ""
|
| 112 |
+
except (KeyError, ValueError) as e:
|
| 113 |
+
print(f"Grok API error: Unexpected response format - {e}")
|
| 114 |
+
return ""
|
| 115 |
+
except Exception as e:
|
| 116 |
+
print(f"Grok API error: Unexpected error - {e}")
|
| 117 |
+
return ""
|
| 118 |
+
|
| 119 |
+
|
| 120 |
+
|
| 121 |
+
def open_oss_get_llm_response(
|
| 122 |
+
system_prompt: str,
|
| 123 |
+
user_input: str,
|
| 124 |
+
tools: List[Dict[str, Any]] = None,
|
| 125 |
+
tool_choice: str = "auto",
|
| 126 |
+
temperature: float = 0.1,
|
| 127 |
+
max_completion_tokens: int = 3000
|
| 128 |
+
) -> str:
|
| 129 |
+
"""
|
| 130 |
+
Make a request to the Grok API and return the response content, supporting tool usage and agentic features.
|
| 131 |
+
|
| 132 |
+
Args:
|
| 133 |
+
system_prompt (str): The system prompt to set the context.
|
| 134 |
+
user_input (str): The user input to process.
|
| 135 |
+
tools (List[Dict[str, Any]], optional): List of tool definitions for tool-calling.
|
| 136 |
+
tool_choice (str, optional): Controls tool usage ("none", "auto", "required"). Defaults to "auto".
|
| 137 |
+
temperature (float, optional): Sampling temperature (0 to 2). Defaults to 0.7 for determinism.
|
| 138 |
+
max_completion_tokens (int, optional): Max tokens in response. Defaults to 1000.
|
| 139 |
+
|
| 140 |
+
Returns:
|
| 141 |
+
str: The content of the assistant's response or tool call results, or empty string on error.
|
| 142 |
+
"""
|
| 143 |
+
# Retrieve API key from environment
|
| 144 |
+
api_key = os.getenv("GROQ_API_KEY")
|
| 145 |
+
if not api_key:
|
| 146 |
+
print("Grok API error: GROQ_API_KEY environment variable not set")
|
| 147 |
+
return ""
|
| 148 |
+
|
| 149 |
+
# API endpoint
|
| 150 |
+
api_url = "https://api.groq.com/openai/v1/chat/completions"
|
| 151 |
+
|
| 152 |
+
# Prepare messages in Grok API format
|
| 153 |
+
messages = [
|
| 154 |
+
{"role": "system", "content": system_prompt},
|
| 155 |
+
{"role": "user", "content": user_input}
|
| 156 |
+
]
|
| 157 |
+
|
| 158 |
+
# Prepare payload
|
| 159 |
+
payload = {
|
| 160 |
+
"model": "openai/gpt-oss-20b",
|
| 161 |
+
"messages": messages,
|
| 162 |
+
"temperature": max(0, min(temperature, 2)), # Clamp to valid range [0, 2]
|
| 163 |
+
"max_completion_tokens": max_completion_tokens,
|
| 164 |
+
"reasoning_effort": "medium"
|
| 165 |
+
}
|
| 166 |
+
|
| 167 |
+
# Add tools and tool_choice if provided
|
| 168 |
+
if tools:
|
| 169 |
+
payload["tools"] = tools
|
| 170 |
+
if tool_choice in ["none", "auto", "required"]:
|
| 171 |
+
payload["tool_choice"] = tool_choice
|
| 172 |
+
else:
|
| 173 |
+
print(f"Grok API warning: Invalid tool_choice '{tool_choice}', defaulting to 'auto'")
|
| 174 |
+
payload["tool_choice"] = "auto"
|
| 175 |
+
|
| 176 |
+
|
| 177 |
+
# Set headers
|
| 178 |
+
headers = {
|
| 179 |
+
"Content-Type": "application/json",
|
| 180 |
+
"Authorization": f"Bearer {api_key}"
|
| 181 |
+
}
|
| 182 |
+
|
| 183 |
+
try:
|
| 184 |
+
# Make API request
|
| 185 |
+
response = requests.post(api_url, headers=headers, json=payload, timeout=60)
|
| 186 |
+
response.raise_for_status()
|
| 187 |
+
|
| 188 |
+
# Parse response
|
| 189 |
+
result = response.json()
|
| 190 |
+
choice = result.get("choices", [{}])[0]
|
| 191 |
+
message = choice.get("message", {})
|
| 192 |
+
|
| 193 |
+
# Handle tool calls if present
|
| 194 |
+
if "tool_calls" in message:
|
| 195 |
+
tool_calls = message["tool_calls"]
|
| 196 |
+
tool_results = []
|
| 197 |
+
for tool_call in tool_calls:
|
| 198 |
+
tool_name = tool_call.get("function", {}).get("name", "")
|
| 199 |
+
tool_args = tool_call.get("function", {}).get("arguments", "{}")
|
| 200 |
+
tool_results.append(f"Tool Call: {tool_name} with args {tool_args}")
|
| 201 |
+
return "; ".join(tool_results) # Combine tool call results into a single string
|
| 202 |
+
|
| 203 |
+
# Return assistant content if no tool calls
|
| 204 |
+
content = message.get("content", "")
|
| 205 |
+
return content.strip()
|
| 206 |
+
|
| 207 |
+
except requests.exceptions.HTTPError as e:
|
| 208 |
+
print(f"Grok API error: HTTP {e.response.status_code} - {e.response.text}")
|
| 209 |
+
return ""
|
| 210 |
+
except requests.exceptions.RequestException as e:
|
| 211 |
+
print(f"Grok API error: Network error - {e}")
|
| 212 |
+
return ""
|
| 213 |
+
except (KeyError, ValueError) as e:
|
| 214 |
+
print(f"Grok API error: Unexpected response format - {e}")
|
| 215 |
+
return ""
|
| 216 |
+
except Exception as e:
|
| 217 |
+
print(f"Grok API error: Unexpected error - {e}")
|
| 218 |
+
return ""
|