File size: 10,699 Bytes
b30f068 9e42edc b30f068 9deb751 b30f068 9deb751 b30f068 9deb751 9e42edc 9deb751 b30f068 | 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 | """
Tool Executor with LLM Response Generation
This is the main execution pipeline that coordinates tools and LLM response generation
"""
import sys
from pathlib import Path
# Add project root to path
project_root = Path(__file__).parent.parent.parent
sys.path.insert(0, str(project_root))
from typing import Dict
from src.tools.weather_tool import execute_weather_tool
from src.tools.llm_response_generator import LLMResponseGenerator
# Soil tool imported lazily in __init__ to avoid circular dependencies
class ToolExecutor:
"""
Executes tools and generates natural language responses using LLM
This is the complete pipeline:
User Question → Tool Execution → LLM Response Generation → User
"""
def __init__(self):
"""Initialize tool executor with LLM response generator"""
self.llm_generator = LLMResponseGenerator()
# Import tools (lazy import to avoid circular dependencies)
from src.tools.soil_tool import execute_soil_tool
from src.tools.cdms_label_tool import execute_cdms_label_tool
from src.tools.agriculture_web_tool import execute_agriculture_web_tool
# Map of tool names to execution functions
# Note: RAG tool removed - CDMS is now the RAG tool for pesticide labels
self.tools = {
"weather": execute_weather_tool,
"soil": execute_soil_tool,
"rag": execute_cdms_label_tool, # Redirect old RAG to CDMS
"documentation": execute_cdms_label_tool, # Redirect to CDMS
"cdms_label": execute_cdms_label_tool,
"cdms": execute_cdms_label_tool, # Alias for cdms_label
"pesticide_label": execute_cdms_label_tool, # Alias
"agriculture_web": execute_agriculture_web_tool,
"ag_web": execute_agriculture_web_tool, # Alias
}
def execute(self, tool_name: str, user_question: str, conversation_context: list = None, offline: bool = None, on_step=None) -> Dict:
"""
Execute a tool and generate LLM response
Args:
tool_name: Name of the tool to execute
user_question: Original user question
conversation_context: Optional list of previous messages for context
Format: [{"role": "user/assistant", "content": "..."}, ...]
Returns:
Dict with:
{
"success": True/False,
"tool_used": "weather",
"raw_data": {...},
"llm_response": "Natural language response from LLM",
"error": "error message if failed"
}
"""
# Check if tool exists
if tool_name not in self.tools:
return {
"success": False,
"tool_used": tool_name,
"error": f"Unknown tool: {tool_name}"
}
try:
# Step 1: Execute the tool (pass context if tool supports it)
tool_function = self.tools[tool_name]
# Pass optional params only to tools whose signature accepts them.
import inspect
sig = inspect.signature(tool_function)
kwargs = {}
if 'conversation_context' in sig.parameters:
kwargs['conversation_context'] = conversation_context
if offline is not None and 'offline' in sig.parameters:
kwargs['offline'] = offline # optional force-index-only override (CDMS tool)
if on_step is not None and 'on_step' in sig.parameters:
kwargs['on_step'] = on_step # live pipeline step callback (CDMS tool)
tool_result = tool_function(user_question, **kwargs)
# Special handling: If CDMS fails or finds no results, try agriculture_web as fallback
if tool_name in ["cdms_label", "cdms", "pesticide_label", "rag", "documentation"]:
cdms_data = tool_result.get("data", {})
rag_chunks = cdms_data.get("rag_chunks", [])
total_chunks = cdms_data.get("total_chunks_found", 0)
should_fallback = tool_result.get("should_fallback", False)
# Debug logging
print(f"🔍 CDMS Tool Result Debug:")
print(f" success: {tool_result.get('success')}")
print(f" total_chunks: {total_chunks}")
print(f" should_fallback: {should_fallback}")
print(f" has_rag_chunks: {len(rag_chunks) if rag_chunks else 0}")
# PHASE 2 FIX: Only fallback if explicitly requested
# Don't fallback just because no chunks were found - CDMS might still have Tavily results
# or be processing new PDFs. Only fallback if explicitly requested.
if should_fallback:
# Explicitly requested fallback - try agriculture_web
print(f"⚠️ CDMS explicitly requested fallback, trying agriculture_web...")
fallback_result = self._try_agriculture_web_fallback(
user_question, conversation_context
)
if fallback_result.get("success"):
print(f"✅ Fallback to agriculture_web successful")
return fallback_result
else:
print(f"⚠️ Fallback to agriculture_web failed, continuing with CDMS")
# If fallback also fails, continue with CDMS error
else:
# No explicit fallback requested - continue with CDMS even if no chunks
# CDMS might be downloading/processing PDFs, or Tavily results might be available
if total_chunks == 0:
print(f"ℹ️ CDMS found 0 chunks, but continuing (may be processing PDFs or have Tavily results)")
# Check if tool execution was successful
if not tool_result.get("success"):
return {
"success": False,
"tool_used": tool_name,
"error": tool_result.get("error", "Tool execution failed"),
"raw_data": tool_result
}
# Step 2: Generate LLM response from tool result (with context)
llm_response = self.llm_generator.generate_response(
user_question=user_question,
tool_name=tool_name,
tool_result=tool_result.get("data", {}),
conversation_context=conversation_context
)
# Step 3: Return complete result
return {
"success": True,
"tool_used": tool_name,
"raw_data": tool_result.get("data", {}),
"llm_response": llm_response
}
except Exception as e:
return {
"success": False,
"tool_used": tool_name,
"error": f"Execution error: {str(e)}"
}
def _try_agriculture_web_fallback(self, user_question: str, conversation_context: list = None) -> Dict:
"""
Fallback to agriculture_web tool when CDMS finds no results
Args:
user_question: User's question
conversation_context: Optional conversation context
Returns:
Dict with tool result or failure
"""
try:
from src.tools.agriculture_web_tool import execute_agriculture_web_tool
# Try agriculture_web tool (with context for follow-ups)
tool_result = execute_agriculture_web_tool(user_question, conversation_context=conversation_context)
if tool_result.get("success"):
# Generate LLM response
llm_response = self.llm_generator.generate_response(
user_question=user_question,
tool_name="agriculture_web",
tool_result=tool_result.get("data", {}),
conversation_context=conversation_context
)
return {
"success": True,
"tool_used": "agriculture_web",
"raw_data": tool_result.get("data", {}),
"llm_response": llm_response,
"fallback_used": True # Indicate this was a fallback
}
else:
return {
"success": False,
"tool_used": "agriculture_web",
"error": tool_result.get("error", "Agriculture web search failed")
}
except Exception as e:
return {
"success": False,
"tool_used": "agriculture_web",
"error": f"Fallback error: {str(e)}"
}
# Test function
if __name__ == "__main__":
print("Testing Tool Executor with LLM Response...")
print("=" * 70)
try:
executor = ToolExecutor()
test_questions = [
("weather", "What's the weather in London?"),
("weather", "Is it hot in Dubai today?"),
("soil", "Show me soil data for Iowa"),
("soil", "What's the soil composition in California?"),
("rag", "How do I use the weather API?"),
("documentation", "What's the weather API documentation?"),
]
for tool_name, question in test_questions:
print(f"\n{'─' * 70}")
print(f"📝 Question: {question}")
print(f"🔧 Tool: {tool_name}")
print("─" * 70)
result = executor.execute(tool_name, question)
if result["success"]:
print("✅ Success!")
print(f"\n🤖 LLM Response:")
print(f" {result['llm_response']}")
print(f"\n📊 Raw Data:")
data = result["raw_data"]
print(f" Location: {data.get('city', 'N/A')}")
print(f" Temperature: {data.get('temperature', 'N/A')}°C")
print(f" Conditions: {data.get('description', 'N/A')}")
else:
print(f"❌ Failed: {result['error']}")
print("\n" + "=" * 70)
print("✅ Testing complete!")
except Exception as e:
print(f"❌ Error: {e}")
|