Spaces:
Sleeping
Sleeping
File size: 9,252 Bytes
3193174 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 | """
Tool integration with LLMs via native function calling.
This module provides:
- LLMResponse and LLMToolCall for parsing LLM responses
- OpenAICaller — a SINGLE caller for all cases (with and without tools)
- Response parsers for OpenAI and Anthropic
Usage:
# One caller for everything
caller = create_openai_caller(api_key="...", model="gpt-4")
# Without tools — returns str
response = caller("Hello!")
# With tools — returns LLMResponse
response = caller("Calculate fib(10)", tools=[...])
if response.has_tool_calls:
for tc in response.tool_calls:
print(tc.name, tc.arguments)
"""
import json
from dataclasses import dataclass, field
from typing import Any
from .base import ToolCall
@dataclass
class LLMToolCall:
"""
Structured tool call from the LLM.
Represents a tool call returned by the LLM via native function calling.
"""
id: str
name: str
arguments: dict[str, Any]
def to_tool_call(self) -> ToolCall:
"""Convert to ToolCall for execution."""
return ToolCall(name=self.name, arguments=self.arguments)
@dataclass
class LLMResponse:
"""
LLM response with tool call support.
Attributes:
content: Text content of the response.
tool_calls: List of tool calls (if requested by the LLM).
raw_response: Original API response (for debugging).
"""
content: str = ""
tool_calls: list[LLMToolCall] = field(default_factory=list)
raw_response: Any = None
@property
def has_tool_calls(self) -> bool:
"""Whether there are tool calls."""
return len(self.tool_calls) > 0
def get_tool_calls(self) -> list[ToolCall]:
"""Get ToolCall objects for execution."""
return [tc.to_tool_call() for tc in self.tool_calls]
def parse_openai_response(response: Any) -> LLMResponse:
"""
Parse an OpenAI API response into LLMResponse.
Supports both the new format (tool_calls) and legacy (function_call).
Args:
response: Response from the OpenAI ChatCompletion API.
Returns:
LLMResponse with parsed data.
"""
message = response.choices[0].message
tool_calls = []
# New format: tool_calls
if hasattr(message, "tool_calls") and message.tool_calls:
for tc in message.tool_calls:
try:
args = json.loads(tc.function.arguments) if tc.function.arguments else {}
except json.JSONDecodeError:
args = {}
tool_calls.append(
LLMToolCall(
id=tc.id,
name=tc.function.name,
arguments=args,
)
)
# Legacy format: function_call
elif hasattr(message, "function_call") and message.function_call:
fc = message.function_call
try:
args = json.loads(fc.arguments) if fc.arguments else {}
except json.JSONDecodeError:
args = {}
tool_calls.append(
LLMToolCall(
id="legacy_call",
name=fc.name,
arguments=args,
)
)
return LLMResponse(
content=message.content or "",
tool_calls=tool_calls,
raw_response=response,
)
def parse_anthropic_response(response: Any) -> LLMResponse:
"""
Parse an Anthropic API response into LLMResponse.
Args:
response: Response from the Anthropic Messages API.
Returns:
LLMResponse with parsed data.
"""
tool_calls = []
content_parts = []
for block in response.content:
if block.type == "text":
content_parts.append(block.text)
elif block.type == "tool_use":
tool_calls.append(
LLMToolCall(
id=block.id,
name=block.name,
arguments=block.input if isinstance(block.input, dict) else {},
)
)
return LLMResponse(
content="\n".join(content_parts),
tool_calls=tool_calls,
raw_response=response,
)
class OpenAICaller:
"""
SINGLE LLM caller for OpenAI — works both with and without tools.
This is the RECOMMENDED way to create callers for agents.
- Without tools: returns str (like a regular caller)
- With tools: returns LLMResponse with tool_calls
Example:
from openai import OpenAI
client = OpenAI(api_key="...")
caller = OpenAICaller(client, model="gpt-4")
# Without tools — regular text response
response = caller("Hello!") # -> str
# With tools — LLMResponse with tool_calls
response = caller("Calculate fib(15)", tools=[...]) # -> LLMResponse
if response.has_tool_calls:
for tc in response.tool_calls:
print(f"Call {tc.name} with {tc.arguments}")
"""
def __init__(
self,
client: Any, # OpenAI client
model: str = "gpt-4",
temperature: float = 0.1, # Low temperature for determinism
max_tokens: int = 2048,
system_prompt: str | None = None,
tool_choice: str = "required", # "required" = mandatory, "auto" = optional
):
"""
Create a universal OpenAI caller.
Args:
client: OpenAI client instance.
model: Model name.
temperature: Generation temperature (default 0.1 for determinism).
max_tokens: Maximum tokens in the response.
system_prompt: System prompt (optional).
tool_choice: Tool usage policy:
- "required": LLM MUST call a tool (default)
- "auto": LLM decides whether to use tools
"""
self.client = client
self.model = model
self.temperature = temperature
self.max_tokens = max_tokens
self.system_prompt = system_prompt
self.tool_choice = tool_choice
def __call__(
self,
prompt: str,
tools: list[dict[str, Any]] | None = None,
) -> str | LLMResponse:
"""
Call the OpenAI API.
Args:
prompt: User prompt.
tools: Tools in OpenAI format (optional).
Returns:
- str: if tools are not passed
- LLMResponse: if tools are passed
"""
messages = []
if self.system_prompt:
messages.append({"role": "system", "content": self.system_prompt})
messages.append({"role": "user", "content": prompt})
kwargs: dict[str, Any] = {
"model": self.model,
"messages": messages,
"temperature": self.temperature,
"max_tokens": self.max_tokens,
}
if tools:
kwargs["tools"] = tools
kwargs["tool_choice"] = self.tool_choice
response = self.client.chat.completions.create(**kwargs)
# If tools were passed — return LLMResponse
if tools:
return parse_openai_response(response)
# If without tools — return just a string
return response.choices[0].message.content or ""
# Alias for backward compatibility
OpenAIToolsCaller = OpenAICaller
def create_openai_caller(
api_key: str | None = None,
base_url: str | None = None,
model: str = "gpt-4",
temperature: float = 0.1, # Low temperature by default
max_tokens: int = 2048,
system_prompt: str | None = None,
tool_choice: str = "required",
) -> OpenAICaller:
"""
Create a universal OpenAI caller.
This is the RECOMMENDED way to create callers for agents.
Works both with and without tools.
Args:
api_key: OpenAI API key (or from an environment variable).
base_url: Base URL (for compatible APIs).
model: Model name.
temperature: Generation temperature (default 0.1 for determinism).
max_tokens: Maximum tokens.
system_prompt: System prompt.
tool_choice: Tool usage policy:
- "required": LLM MUST call a tool (default)
- "auto": LLM decides whether to use tools
Returns:
Ready-to-use OpenAICaller.
Example:
# One caller for all agents
caller = create_openai_caller(
api_key="sk-...",
model="gpt-4",
)
# Without tools — plain text
response = caller("Hello!") # -> str
# With tools — LLMResponse
response = caller("Calculate fib(10)", tools=[...])
if response.has_tool_calls:
...
"""
try:
from openai import OpenAI
except ImportError as e:
msg = "openai package required: pip install openai"
raise ImportError(msg) from e
kwargs: dict[str, Any] = {}
if api_key:
kwargs["api_key"] = api_key
if base_url:
kwargs["base_url"] = base_url
client = OpenAI(**kwargs)
return OpenAICaller(
client=client,
model=model,
temperature=temperature,
max_tokens=max_tokens,
system_prompt=system_prompt,
tool_choice=tool_choice,
)
# Alias for backward compatibility
create_openai_tools_caller = create_openai_caller
|