Spaces:
Sleeping
Sleeping
File size: 11,588 Bytes
9bf2b3a a19cb61 |
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 |
import os
from pathlib import Path
import traceback
from typing import Annotated, List, Union
from typing_extensions import TypedDict
import pandas as pd
import gradio as gr
from langchain_core.messages import AIMessage, HumanMessage, ToolMessage
from langchain_core.tools import tool
from langchain_google_genai import ChatGoogleGenerativeAI
from langgraph.graph import StateGraph
from langgraph.prebuilt import ToolNode
from rag_pipeline import load_or_create_vector_store
from google import genai
from google.genai import types
from google.api_core import retry
# ---------------------------
# --- Setup Google API ---
# ---------------------------
GOOGLE_API_KEY = os.environ.get("GOOGLE_API_KEY")
if not GOOGLE_API_KEY:
raise ValueError("GOOGLE_API_KEY not found in environment variables.")
client = genai.Client(api_key=GOOGLE_API_KEY)
# Retry policy
is_retriable = lambda e: (isinstance(e, genai.errors.APIError) and e.code in {429, 503})
if not hasattr(genai.models.Models.generate_content, '__wrapped__'):
genai.models.Models.generate_content = retry.Retry(predicate=is_retriable)(genai.models.Models.generate_content)
# ---------------------------
# --- Config ---
# ---------------------------
model_name = "gemini-2.5-flash-lite" #"gemini-2.0-flash-lite"#"gemini-2.0-flash"
base_dir = Path(__file__).resolve().parent.parent
doc2_path = str( base_dir / "sport_tool_docs/toughestsport.csv")
search_kwargs_k = 5
search_kwargs_fetch_k = 10
# ---------------------------
# --- RAG Setup ---
# ---------------------------
vector_store = load_or_create_vector_store()
retriever = vector_store.as_retriever(
search_type="mmr",
search_kwargs={
"k": search_kwargs_k,
"fetch_k": search_kwargs_fetch_k
}
)
# ---------------------------
# --- Sports Data Setup ---
# ---------------------------
sports_by_skills = pd.read_csv(doc2_path)
sports_by_skills.columns = sports_by_skills.columns.str.lower()
sports_by_skills['sport'] = sports_by_skills['sport'].str.lower()
# ---------------------------
# --- LangGraph Tools ---
# ---------------------------
@tool
def get_sports() -> str:
"""Return a structured list of sports from the dataset."""
prompt = """Parse the provided sports into a structured list where each line has indentation and starts with a category, followed by a colon,
and then a comma-separated list of sports within that category.
If a sport has no obvious category, group it under "General"
EXAMPLE:
Provide me sport options
Answer:
- Ball games:
- Football, Baskettball
- Skiing:
- Alpine, Nordic
- General:
- Boxing, Water polo
"""
sports = sports_by_skills["sport"].tolist()
response = client.models.generate_content(model=model_name, contents=[prompt, sports])
return response.candidates[0].content.parts[0].text.strip()
@tool
def get_document_answer(query: str) -> str:
"""Retrieve an answer from documents with a grounded paraphrase."""
try:
results = retriever.invoke(query)
if not results:
return "I could not find any relevant information in the documents."
# Combine the retrieved chunks for context
combined_text = "\n---\n".join([r.page_content for r in results])
prompt = f"""
Answer the question based on the following documents.
If the information is not available, state that you cannot find the answer in the provided documents.
Chunks:
{combined_text}
Question: {query}
Answer:
"""
# Call the LLM
response = client.models.generate_content(
model=model_name,
contents=[prompt]
)
if response.candidates:
answer_text = response.candidates[0].content.parts[0].text.strip()
return answer_text
else:
return "I could not generate an answer."
except Exception as e:
return f"RAG error: {e}"
@tool
def get_skills_by_sport(sport: str) -> str:
"""Get the sport name. Return: The top 3 highest skill rates."""
sport = sport.lower().strip()
skill_rates = sports_by_skills.loc[sports_by_skills['sport'] == sport]
if skill_rates.empty:
return f"No data found for sport '{sport}'. Please check the spelling or try another sport."
skills_only = skill_rates.drop(columns=['sport', 'total', 'rank'])
transposed = skills_only.T
col = transposed.columns[0]
top_3_skills = transposed.nlargest(3, col)
top_3_skill_names = "\n".join(f"{skill}" for skill, value in top_3_skills[col].items())
return f"Top 3 skills for {sport.capitalize()}:\n{top_3_skill_names}"
@tool
def get_equipment_by_sport(sport: str) -> str:
"""Get the equipment list for a sport using a google search grounded prompt."""
sport = sport.lower()
prompt = """Parse a customer's sport equipment question to the list:
EXAMPLE: What are the necessary equipment for boxing?
Response:
- Mandatory: 1 gloves, 3 socks
- Recommended: 1 towel
- Fun: resistance bands
"""
config_with_search = types.GenerateContentConfig(
tools=[types.Tool(google_search=types.GoogleSearch())],
temperature=0.0,
)
contents_text = "What are the necessary equipment for this " + sport + "?"
response = client.models.generate_content(
model=model_name,
contents=[prompt, contents_text],
config=config_with_search,
)
return response.candidates[0].content.parts[0].text if response.candidates else "No information found."
# ---------------------------
# --- Tool Node ---
# ---------------------------
tools_list = [get_sports, get_document_answer, get_skills_by_sport, get_equipment_by_sport]
tool_node = ToolNode(tools_list)
# ---------------------------
# --- LangGraph LLM ---
# ---------------------------
llm = ChatGoogleGenerativeAI(model=model_name)
llm_with_tools = llm.bind_tools(tools_list, return_direct=True)
# --- Graph State ---
class SportAdvicerState(TypedDict):
messages: Annotated[List[Union[AIMessage, HumanMessage, ToolMessage]], list.__add__]
def detect_intent(query: str) -> str:
"""Classify user query into one of: equipment, skills, document, sports_list, general."""
classification_prompt = f"""
You are a classifier.
Categorize the following user query into exactly ONE of these categories:
- equipment β if asking about gear, equipment, things needed for a sport
- skills β if asking about skills, abilities, rankings, requirements for a sport
- document β if asking about information that may be inside books, PDFs, or retrieved documents
- sports_list β if asking for a list of sports, categories of sports, or groupings of sports
- general β if it's a general sports question not fitting the above
Query: "{query}"
Answer with one word: equipment, skills, document, sports_list, or general.
"""
response = client.models.generate_content(
model=model_name,
contents=[classification_prompt],
config=types.GenerateContentConfig(
temperature=0.0 # deterministic
)
)
if response.candidates:
return response.candidates[0].content.parts[0].text.strip().lower()
else:
return "general"
def extract_sport_name(query: str) -> str:
"""Extract the sport name from a user query."""
extraction_prompt = f"""
Extract the single sport name from the following query.
If multiple sports are mentioned, return the first one.
If no sport is mentioned, return an empty string.
Query: "{query}"
Extracted sport name:
"""
response = client.models.generate_content(
model=model_name,
contents=[extraction_prompt],
config=types.GenerateContentConfig(
temperature=0.0
)
)
if response.candidates:
return response.candidates[0].content.parts[0].text.strip()
return ""
def chatbot_node(state: SportAdvicerState) -> SportAdvicerState:
user_message = state["messages"][-1]
query = user_message.content
intent = detect_intent(query)
sport_name = extract_sport_name(query)
if intent == "equipment":
response_text = get_equipment_by_sport.invoke({"sport": sport_name})
return {"messages": [AIMessage(content=response_text)]}
elif intent == "skills":
response_text = get_skills_by_sport.invoke({"sport": sport_name})
return {"messages": [AIMessage(content=response_text)]}
elif intent == "document":
response_text = get_document_answer.invoke({"query": query})
return {"messages": [AIMessage(content=response_text)]}
elif intent == "sports_list":
response_text = get_sports.invoke({})
return {"messages": [AIMessage(content=response_text)]}
else: # general
messages_with_instruction = [
HumanMessage(content="""You are a sports advisor chatbot.
You can answer general sports questions.
For equipment, skills, document, or sports list queries, tools are used automatically.""")
] + state["messages"]
response = llm_with_tools.invoke(messages_with_instruction)
return {"messages": [response]}
# Routing: always go to the single tool node if any tool call exists
def should_route_to_tools(state: SportAdvicerState):
last_msg = state["messages"][-1]
if hasattr(last_msg, "tool_calls") and last_msg.tool_calls:
return "tools"
return "__end__"
# ---------------------------
# --- Graph Definition ---
# ---------------------------
graph_builder = StateGraph(SportAdvicerState)
graph_builder.add_node("chatbot", chatbot_node)
graph_builder.add_node("tools", tool_node)
graph_builder.add_conditional_edges("chatbot", should_route_to_tools)
graph_builder.add_edge("tools", "chatbot")
graph_builder.set_entry_point("chatbot")
graph_with_rag = graph_builder.compile()
# ---------------------------
# --- Gradio Interface ---
# ---------------------------
def chatbot_interface(message, history):
langchain_messages = []
for chat_entry in history:
if isinstance(chat_entry, list) and len(chat_entry) == 2:
if chat_entry[0]: langchain_messages.append(HumanMessage(content=chat_entry[0]))
if chat_entry[1]: langchain_messages.append(AIMessage(content=chat_entry[1]))
elif isinstance(chat_entry, dict):
if chat_entry["role"] == "user": langchain_messages.append(HumanMessage(content=chat_entry["content"]))
elif chat_entry["role"] == "assistant": langchain_messages.append(AIMessage(content=chat_entry["content"]))
langchain_messages.append(HumanMessage(content=message))
current_state = {"messages": langchain_messages}
try:
response_state = graph_with_rag.invoke(current_state)
bot_response = response_state["messages"][-1].content
return bot_response
except Exception as e:
traceback.print_exc()
return f"Internal error: {e}"
iface = gr.ChatInterface(
fn=chatbot_interface,
chatbot=gr.Chatbot(height=500, type="messages",
value=[{"role": "assistant", "content": "Hello! I am your AI Sport Advisor. Ask me anything."}]),
title="Agentic RAG Sport Advisor Chatbot",
description="LangGraph chatbot integrated with RAG document retrieval and sports tools.",
type="messages",
)
if __name__ == "__main__":
iface.launch(share=True) |