Spaces:
Running
Running
File size: 9,605 Bytes
676582c |
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 |
"""
Agent Runner
Core orchestrator for AI agent execution with tool calling support.
Manages the full request cycle: LLM generation → tool execution → final response.
"""
import logging
from typing import List, Dict, Any, Optional
import asyncio
from .agent_config import AgentConfiguration
from .providers.base import LLMProvider
from .providers.gemini import GeminiProvider
from .providers.openrouter import OpenRouterProvider
from .providers.cohere import CohereProvider
from ..mcp.tool_registry import MCPToolRegistry, ToolExecutionResult
logger = logging.getLogger(__name__)
class AgentRunner:
"""
Agent execution orchestrator with tool calling support.
This class manages the full agent request cycle:
1. Generate LLM response with tool definitions
2. If tool calls requested, execute tools with user context injection
3. Generate final response with tool results
4. Handle rate limiting with fallback providers
"""
def __init__(self, config: AgentConfiguration, tool_registry: MCPToolRegistry):
"""
Initialize the agent runner.
Args:
config: Agent configuration
tool_registry: MCP tool registry
"""
self.config = config
self.tool_registry = tool_registry
self.primary_provider = self._create_provider(config.provider)
self.fallback_provider = None
if config.fallback_provider:
self.fallback_provider = self._create_provider(config.fallback_provider)
logger.info(f"Initialized AgentRunner with provider: {config.provider}")
def _create_provider(self, provider_name: str) -> LLMProvider:
"""
Create an LLM provider instance.
Args:
provider_name: Provider name (gemini, openrouter, cohere)
Returns:
LLMProvider instance
Raises:
ValueError: If provider is not supported or API key is missing
"""
api_key = self.config.get_provider_api_key(provider_name)
if not api_key:
raise ValueError(f"API key not configured for provider: {provider_name}")
model = self.config.get_provider_model(provider_name)
if provider_name == "gemini":
return GeminiProvider(
api_key=api_key,
model=model,
temperature=self.config.temperature,
max_tokens=self.config.max_tokens
)
elif provider_name == "openrouter":
return OpenRouterProvider(
api_key=api_key,
model=model,
temperature=self.config.temperature,
max_tokens=self.config.max_tokens
)
elif provider_name == "cohere":
return CohereProvider(
api_key=api_key,
model=model,
temperature=self.config.temperature,
max_tokens=self.config.max_tokens
)
else:
raise ValueError(f"Unsupported provider: {provider_name}")
async def execute(
self,
messages: List[Dict[str, str]],
user_id: int,
system_prompt: Optional[str] = None
) -> Dict[str, Any]:
"""
Execute agent request with tool calling support.
SECURITY: user_id is injected by backend, never from LLM output.
Args:
messages: Conversation history [{"role": "user", "content": "..."}]
user_id: User ID (injected by backend for security)
system_prompt: Optional system prompt (uses config default if not provided)
Returns:
Dict with response content and metadata
"""
prompt = system_prompt or self.config.system_prompt
provider = self.primary_provider
try:
# Get tool definitions
tool_definitions = self.tool_registry.get_tool_definitions()
logger.info(f"Executing agent for user {user_id} with {len(tool_definitions)} tools")
# Generate initial response with tool definitions
response = await provider.generate_response_with_tools(
messages=messages,
system_prompt=prompt,
tools=tool_definitions
)
# Check if tool calls were requested
if response.tool_calls:
logger.info(f"Agent requested {len(response.tool_calls)} tool calls")
# Execute all tool calls
tool_results = []
for tool_call in response.tool_calls:
result = await self.tool_registry.execute_tool(
tool_name=tool_call["name"],
arguments=tool_call["arguments"],
user_id=user_id # Inject user context for security
)
tool_results.append(result)
# Generate final response with tool results
final_response = await provider.generate_response_with_tool_results(
messages=messages,
tool_calls=response.tool_calls,
tool_results=tool_results
)
return {
"content": final_response.content,
"tool_calls": response.tool_calls,
"tool_results": tool_results,
"provider": provider.get_provider_name()
}
# No tool calls, return direct response
logger.info("Agent generated direct response (no tool calls)")
return {
"content": response.content,
"tool_calls": None,
"tool_results": None,
"provider": provider.get_provider_name()
}
except Exception as e:
logger.error(f"Agent execution failed with primary provider: {str(e)}")
# Try fallback provider if configured
if self.fallback_provider:
logger.info("Attempting fallback provider")
try:
return await self._execute_with_provider(
provider=self.fallback_provider,
messages=messages,
user_id=user_id,
system_prompt=prompt
)
except Exception as fallback_error:
logger.error(f"Fallback provider also failed: {str(fallback_error)}")
raise
raise
async def _execute_with_provider(
self,
provider: LLMProvider,
messages: List[Dict[str, str]],
user_id: int,
system_prompt: str
) -> Dict[str, Any]:
"""
Execute agent request with a specific provider.
Args:
provider: LLM provider to use
messages: Conversation history
user_id: User ID
system_prompt: System prompt
Returns:
Dict with response content and metadata
"""
tool_definitions = self.tool_registry.get_tool_definitions()
# Generate initial response
response = await provider.generate_response_with_tools(
messages=messages,
system_prompt=system_prompt,
tools=tool_definitions
)
# Handle tool calls
if response.tool_calls:
tool_results = []
for tool_call in response.tool_calls:
result = await self.tool_registry.execute_tool(
tool_name=tool_call["name"],
arguments=tool_call["arguments"],
user_id=user_id
)
tool_results.append(result)
final_response = await provider.generate_response_with_tool_results(
messages=messages,
tool_calls=response.tool_calls,
tool_results=tool_results
)
return {
"content": final_response.content,
"tool_calls": response.tool_calls,
"tool_results": tool_results,
"provider": provider.get_provider_name()
}
return {
"content": response.content,
"tool_calls": None,
"tool_results": None,
"provider": provider.get_provider_name()
}
async def execute_simple(
self,
messages: List[Dict[str, str]],
system_prompt: Optional[str] = None
) -> str:
"""
Execute a simple agent request without tool calling.
Args:
messages: Conversation history
system_prompt: Optional system prompt
Returns:
Response content as string
"""
prompt = system_prompt or self.config.system_prompt
provider = self.primary_provider
try:
response = await provider.generate_simple_response(
messages=messages,
system_prompt=prompt
)
return response.content or ""
except Exception as e:
logger.error(f"Simple execution failed: {str(e)}")
# Try fallback provider
if self.fallback_provider:
logger.info("Attempting fallback provider for simple execution")
response = await self.fallback_provider.generate_simple_response(
messages=messages,
system_prompt=prompt
)
return response.content or ""
raise
|