File size: 6,226 Bytes
bee32cd 9bd34a5 bee32cd 64b6717 bee32cd d888f9f bee32cd d888f9f bee32cd 9bd34a5 46d5b0d bee32cd 64b6717 bee32cd d888f9f bee32cd d888f9f bee32cd 9bd34a5 bee32cd 9bd34a5 bee32cd 9bd34a5 bee32cd 9bd34a5 bee32cd 9bd34a5 bee32cd e7c10d7 2139c78 fef2a27 bee32cd eb15d69 5abc589 eb15d69 | 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 | import re
from typing import List
from typing_extensions import TypedDict, Annotated
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.messages import AnyMessage, HumanMessage
from langchain_core.tools import tool
from langgraph.graph.message import add_messages
from langgraph.graph import StateGraph, START, END
from langgraph.prebuilt import ToolNode, tools_condition
import sys
# Increase Python's recursion limit
sys.setrecursionlimit(10000) # Default is typically 1000
from prompts import prompt, solve_prompt
from llm import model
from tools import \
web_search, \
wiki_search, \
arxiv_search, \
multiply, \
add, \
subtract, \
divide, \
square_root, \
save_file, \
read_file, \
extract_text_from_image, \
get_youtube_transcript, \
execute_code_multilang, \
download_file_from_url, \
analyze_csv_file, \
analyze_excel_file, \
analyze_image
# Define regex pattern at the module level
regex_pattern = r"Plan:\s*(.+)\s*(#E\d+)\s*=\s*(\w+)\s*\[([^\]]+)\]"
class ReWOO(TypedDict):
messages: Annotated[list[AnyMessage], add_messages]
task: str
plan_string: str
steps: List
results: dict
result: str
tool_calls: List
tool_results: List
current_step: int
tools = [
web_search,
wiki_search,
arxiv_search,
multiply,
add,
subtract,
divide,
square_root,
save_file,
read_file,
extract_text_from_image,
get_youtube_transcript,
execute_code_multilang,
download_file_from_url,
analyze_csv_file,
analyze_excel_file,
analyze_image
]
try:
llm_with_tools = model.bind_tools(tools)
except ValueError as e:
print(f"Warning: Failed to bind tools: {e}")
print("Continuing without binding tools...")
llm_with_tools = model
# Define planner at module level
prompt_template = ChatPromptTemplate.from_messages([("user", prompt)])
planner = prompt_template | model
def get_plan(state: ReWOO):
task = state["task"]
result = planner.invoke({"task": task})
matches = re.findall(regex_pattern, result.content)
# Safety check - if no steps found, add a simple LLM step
if not matches:
print("No steps found in planner output, creating fallback step")
return {
"steps": [("Answer the question directly", "#E1", "LLM", task)],
"plan_string": f"Plan: Answer the question directly\n#E1 = LLM[{task}]"
}
return {"steps": matches, "plan_string": result.content}
def _get_current_task(state: ReWOO):
if "results" not in state or state["results"] is None:
return 1
# Safety check - ensure steps exist
if "steps" not in state or not state["steps"]:
return None
if len(state["results"]) >= len(state["steps"]):
return None
else:
return len(state["results"]) + 1
def tool_execution(state: ReWOO):
_step = _get_current_task(state)
# Safety checks
if _step is None:
return {"results": state.get("results", {})}
if "steps" not in state or not state["steps"] or _step > len(state["steps"]):
return {"results": state.get("results", {})}
try:
_, step_name, tool, tool_input = state["steps"][_step - 1]
_results = (state["results"] or {}) if "results" in state else {}
for k, v in _results.items():
if isinstance(v, str) and isinstance(tool_input, str):
tool_input = tool_input.replace(k, v)
if tool == "Google":
result = web_search.invoke(tool_input)
elif tool == "LLM":
result = model.invoke(tool_input)
else:
# Handle unsupported tools gracefully
result = f"Tool '{tool}' not supported. Using LLM instead."
result = model.invoke(tool_input)
_results[step_name] = str(result)
return {"results": _results}
except Exception as e:
print(f"Error in tool execution: {e}")
# Return current results without modification in case of error
return {"results": state.get("results", {})}
def solve(state: ReWOO):
plan = ""
# Safety check
if "steps" not in state or not state["steps"]:
return {"result": "I couldn't generate a plan to solve this task. Here's my best attempt: " + state["task"]}
try:
for _plan, step_name, tool, tool_input in state["steps"]:
_results = (state["results"] or {}) if "results" in state else {}
for k, v in _results.items():
if isinstance(v, str) and isinstance(tool_input, str):
tool_input = tool_input.replace(k, v)
if isinstance(v, str) and isinstance(step_name, str):
step_name = step_name.replace(k, v)
plan += f"Plan: {_plan}\n{step_name} = {tool}[{tool_input}]\n\n"
# Fallback if plan is empty
if not plan.strip():
plan = f"Direct answer: {state['task']}"
prompt_text = solve_prompt.format(plan=plan, task=state["task"])
result = model.invoke(prompt_text)
return {"result": result.content}
except Exception as e:
print(f"Error in solve: {e}")
# Provide a fallback answer in case of errors
return {"result": f"I tried to solve this but encountered a problem. Here's what I know: {state.get('results', {})}"}
def _route(state):
_step = _get_current_task(state)
if _step is None:
return "solve"
else:
return "tool"
def rewoo_graph():
prompt_template = ChatPromptTemplate.from_messages([("user", prompt)])
global planner # Define planner at the global scope so get_plan can use it
planner = prompt_template | model
graph = StateGraph(ReWOO)
graph.add_node("plan", get_plan)
graph.add_node("tool", tool_execution)
graph.add_node("solve", solve)
graph.add_edge("plan", "tool")
graph.add_edge("solve", END)
graph.add_conditional_edges("tool", _route)
graph.add_edge(START, "plan")
compiled_graph = graph.compile()
return compiled_graph |