Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,8 +1,10 @@
|
|
| 1 |
import os
|
| 2 |
-
import
|
| 3 |
-
import json
|
| 4 |
import requests
|
|
|
|
| 5 |
import pandas as pd
|
|
|
|
|
|
|
| 6 |
from pathlib import Path
|
| 7 |
from typing import Optional
|
| 8 |
from dotenv import load_dotenv
|
|
@@ -11,16 +13,16 @@ from langgraph.prebuilt import create_react_agent
|
|
| 11 |
from langchain_core.messages import HumanMessage
|
| 12 |
from langchain_core.tools import tool
|
| 13 |
from langchain_openai import ChatOpenAI
|
| 14 |
-
import inspect
|
| 15 |
|
| 16 |
load_dotenv()
|
| 17 |
|
|
|
|
| 18 |
|
| 19 |
class OpenRouterLLM(ChatOpenAI):
|
| 20 |
"""Custom OpenRouter LLM wrapper for LangGraph"""
|
| 21 |
|
| 22 |
def __init__(self, model: str = "deepseek/deepseek-v3.1-terminus", **kwargs):
|
| 23 |
-
api_key = os.getenv("OPENROUTER_API_KEY") or os.getenv("
|
| 24 |
super().__init__(
|
| 25 |
model=model,
|
| 26 |
openai_api_key=api_key,
|
|
@@ -28,12 +30,10 @@ class OpenRouterLLM(ChatOpenAI):
|
|
| 28 |
**kwargs
|
| 29 |
)
|
| 30 |
|
| 31 |
-
|
| 32 |
# ------------------ TOOLS ------------------
|
| 33 |
|
| 34 |
@tool
|
| 35 |
def search_web(query: str) -> str:
|
| 36 |
-
"""Search the web using DuckDuckGo for current information."""
|
| 37 |
try:
|
| 38 |
search_url = f"https://api.duckduckgo.com/?q={query}&format=json&no_html=1&skip_disambig=1"
|
| 39 |
response = requests.get(search_url, timeout=10)
|
|
@@ -51,10 +51,8 @@ def search_web(query: str) -> str:
|
|
| 51 |
except Exception as e:
|
| 52 |
return f"Search error: {str(e)}"
|
| 53 |
|
| 54 |
-
|
| 55 |
@tool
|
| 56 |
def search_wikipedia(query: str) -> str:
|
| 57 |
-
"""Search Wikipedia for factual information."""
|
| 58 |
try:
|
| 59 |
search_url = "https://en.wikipedia.org/api/rest_v1/page/summary/" + query.replace(" ", "_")
|
| 60 |
response = requests.get(search_url, timeout=10)
|
|
@@ -66,10 +64,8 @@ def search_wikipedia(query: str) -> str:
|
|
| 66 |
except Exception as e:
|
| 67 |
return f"Wikipedia search error: {str(e)}"
|
| 68 |
|
| 69 |
-
|
| 70 |
@tool
|
| 71 |
def execute_python(code: str) -> str:
|
| 72 |
-
"""Execute Python code and return the result."""
|
| 73 |
try:
|
| 74 |
safe_globals = {
|
| 75 |
'__builtins__': {
|
|
@@ -95,10 +91,8 @@ def execute_python(code: str) -> str:
|
|
| 95 |
except Exception as e:
|
| 96 |
return f"Python execution error: {str(e)}"
|
| 97 |
|
| 98 |
-
|
| 99 |
@tool
|
| 100 |
def read_excel_file(file_path: str, sheet_name: Optional[str] = None) -> str:
|
| 101 |
-
"""Read an Excel file and return its contents as a formatted string."""
|
| 102 |
try:
|
| 103 |
file_path_obj = Path(file_path)
|
| 104 |
if not file_path_obj.exists():
|
|
@@ -120,10 +114,8 @@ def read_excel_file(file_path: str, sheet_name: Optional[str] = None) -> str:
|
|
| 120 |
except Exception as e:
|
| 121 |
return f"Error reading Excel file: {str(e)}"
|
| 122 |
|
| 123 |
-
|
| 124 |
@tool
|
| 125 |
def read_text_file(file_path: str) -> str:
|
| 126 |
-
"""Read a text file and return its contents."""
|
| 127 |
try:
|
| 128 |
file_path_obj = Path(file_path)
|
| 129 |
if not file_path_obj.exists():
|
|
@@ -139,14 +131,13 @@ def read_text_file(file_path: str) -> str:
|
|
| 139 |
except Exception as e:
|
| 140 |
return f"Error reading file: {str(e)}"
|
| 141 |
|
| 142 |
-
|
| 143 |
# ------------------ GAIA AGENT ------------------
|
| 144 |
|
| 145 |
class GaiaAgent:
|
| 146 |
-
"""LangGraph-based agent
|
| 147 |
|
| 148 |
def __init__(self):
|
| 149 |
-
print("Initializing GaiaAgent
|
| 150 |
self.llm = OpenRouterLLM(
|
| 151 |
model="deepseek/deepseek-v3.1-terminus",
|
| 152 |
temperature=0.1,
|
|
@@ -154,8 +145,6 @@ class GaiaAgent:
|
|
| 154 |
)
|
| 155 |
self.tools = [search_web, search_wikipedia, execute_python, read_excel_file, read_text_file]
|
| 156 |
prompt_modifier = self._get_system_prompt()
|
| 157 |
-
|
| 158 |
-
# Detect correct kwarg for your LangGraph version
|
| 159 |
sig = inspect.signature(create_react_agent)
|
| 160 |
accepted = sig.parameters.keys()
|
| 161 |
kwargs = {}
|
|
@@ -165,55 +154,76 @@ class GaiaAgent:
|
|
| 165 |
kwargs["state_modifier"] = prompt_modifier
|
| 166 |
elif "prompt" in accepted:
|
| 167 |
kwargs["prompt"] = prompt_modifier
|
| 168 |
-
|
| 169 |
self.agent = create_react_agent(self.llm, self.tools, **kwargs)
|
| 170 |
print("GaiaAgent initialized successfully!")
|
| 171 |
|
| 172 |
def _get_system_prompt(self) -> str:
|
| 173 |
-
return """You are an advanced AI agent designed to answer complex questions
|
| 174 |
-
|
| 175 |
|
| 176 |
-
def __call__(self,
|
| 177 |
try:
|
| 178 |
-
print(f"Processing task {task_id}: {question[:100]}...")
|
| 179 |
messages = [HumanMessage(content=question)]
|
| 180 |
result = self.agent.invoke({"messages": messages})
|
| 181 |
-
|
| 182 |
-
answer = final_message.content
|
| 183 |
return self._clean_answer(answer)
|
| 184 |
except Exception as e:
|
| 185 |
return f"Agent error: {e}"
|
| 186 |
|
| 187 |
def _clean_answer(self, answer: str) -> str:
|
| 188 |
-
# same cleaning code as before
|
| 189 |
answer = answer.strip()
|
| 190 |
if "final answer:" in answer.lower():
|
| 191 |
parts = re.split(r'final answer:', answer, flags=re.IGNORECASE)
|
| 192 |
if len(parts) > 1:
|
| 193 |
answer = parts[-1].strip()
|
| 194 |
-
prefixes = ["The answer is", "Answer:", "Result:", "Solution:"
|
| 195 |
-
"Based on", "Therefore", "In conclusion", "So the answer is"]
|
| 196 |
for prefix in prefixes:
|
| 197 |
if answer.lower().startswith(prefix.lower()):
|
| 198 |
answer = answer[len(prefix):].strip()
|
| 199 |
if answer.startswith(':'):
|
| 200 |
answer = answer[1:].strip()
|
| 201 |
break
|
| 202 |
-
if len(answer.split()) <= 3:
|
| 203 |
-
answer = answer.strip('"\'.')
|
| 204 |
return answer
|
| 205 |
|
|
|
|
| 206 |
|
| 207 |
-
|
| 208 |
|
| 209 |
-
|
| 210 |
-
|
| 211 |
-
|
|
|
|
|
|
|
| 212 |
|
| 213 |
-
|
| 214 |
-
|
|
|
|
|
|
|
| 215 |
|
| 216 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 217 |
|
| 218 |
-
|
| 219 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
import os
|
| 2 |
+
import gradio as gr
|
|
|
|
| 3 |
import requests
|
| 4 |
+
import inspect
|
| 5 |
import pandas as pd
|
| 6 |
+
import re
|
| 7 |
+
import json
|
| 8 |
from pathlib import Path
|
| 9 |
from typing import Optional
|
| 10 |
from dotenv import load_dotenv
|
|
|
|
| 13 |
from langchain_core.messages import HumanMessage
|
| 14 |
from langchain_core.tools import tool
|
| 15 |
from langchain_openai import ChatOpenAI
|
|
|
|
| 16 |
|
| 17 |
load_dotenv()
|
| 18 |
|
| 19 |
+
# ------------------ CUSTOM LLM ------------------
|
| 20 |
|
| 21 |
class OpenRouterLLM(ChatOpenAI):
|
| 22 |
"""Custom OpenRouter LLM wrapper for LangGraph"""
|
| 23 |
|
| 24 |
def __init__(self, model: str = "deepseek/deepseek-v3.1-terminus", **kwargs):
|
| 25 |
+
api_key = os.getenv("OPENROUTER_API_KEY") or os.getenv("my_key")
|
| 26 |
super().__init__(
|
| 27 |
model=model,
|
| 28 |
openai_api_key=api_key,
|
|
|
|
| 30 |
**kwargs
|
| 31 |
)
|
| 32 |
|
|
|
|
| 33 |
# ------------------ TOOLS ------------------
|
| 34 |
|
| 35 |
@tool
|
| 36 |
def search_web(query: str) -> str:
|
|
|
|
| 37 |
try:
|
| 38 |
search_url = f"https://api.duckduckgo.com/?q={query}&format=json&no_html=1&skip_disambig=1"
|
| 39 |
response = requests.get(search_url, timeout=10)
|
|
|
|
| 51 |
except Exception as e:
|
| 52 |
return f"Search error: {str(e)}"
|
| 53 |
|
|
|
|
| 54 |
@tool
|
| 55 |
def search_wikipedia(query: str) -> str:
|
|
|
|
| 56 |
try:
|
| 57 |
search_url = "https://en.wikipedia.org/api/rest_v1/page/summary/" + query.replace(" ", "_")
|
| 58 |
response = requests.get(search_url, timeout=10)
|
|
|
|
| 64 |
except Exception as e:
|
| 65 |
return f"Wikipedia search error: {str(e)}"
|
| 66 |
|
|
|
|
| 67 |
@tool
|
| 68 |
def execute_python(code: str) -> str:
|
|
|
|
| 69 |
try:
|
| 70 |
safe_globals = {
|
| 71 |
'__builtins__': {
|
|
|
|
| 91 |
except Exception as e:
|
| 92 |
return f"Python execution error: {str(e)}"
|
| 93 |
|
|
|
|
| 94 |
@tool
|
| 95 |
def read_excel_file(file_path: str, sheet_name: Optional[str] = None) -> str:
|
|
|
|
| 96 |
try:
|
| 97 |
file_path_obj = Path(file_path)
|
| 98 |
if not file_path_obj.exists():
|
|
|
|
| 114 |
except Exception as e:
|
| 115 |
return f"Error reading Excel file: {str(e)}"
|
| 116 |
|
|
|
|
| 117 |
@tool
|
| 118 |
def read_text_file(file_path: str) -> str:
|
|
|
|
| 119 |
try:
|
| 120 |
file_path_obj = Path(file_path)
|
| 121 |
if not file_path_obj.exists():
|
|
|
|
| 131 |
except Exception as e:
|
| 132 |
return f"Error reading file: {str(e)}"
|
| 133 |
|
|
|
|
| 134 |
# ------------------ GAIA AGENT ------------------
|
| 135 |
|
| 136 |
class GaiaAgent:
|
| 137 |
+
"""LangGraph-based agent using OpenRouter DeepSeek"""
|
| 138 |
|
| 139 |
def __init__(self):
|
| 140 |
+
print("Initializing GaiaAgent...")
|
| 141 |
self.llm = OpenRouterLLM(
|
| 142 |
model="deepseek/deepseek-v3.1-terminus",
|
| 143 |
temperature=0.1,
|
|
|
|
| 145 |
)
|
| 146 |
self.tools = [search_web, search_wikipedia, execute_python, read_excel_file, read_text_file]
|
| 147 |
prompt_modifier = self._get_system_prompt()
|
|
|
|
|
|
|
| 148 |
sig = inspect.signature(create_react_agent)
|
| 149 |
accepted = sig.parameters.keys()
|
| 150 |
kwargs = {}
|
|
|
|
| 154 |
kwargs["state_modifier"] = prompt_modifier
|
| 155 |
elif "prompt" in accepted:
|
| 156 |
kwargs["prompt"] = prompt_modifier
|
|
|
|
| 157 |
self.agent = create_react_agent(self.llm, self.tools, **kwargs)
|
| 158 |
print("GaiaAgent initialized successfully!")
|
| 159 |
|
| 160 |
def _get_system_prompt(self) -> str:
|
| 161 |
+
return """You are an advanced AI agent designed to answer complex questions.
|
| 162 |
+
Keep answers concise and factual where possible."""
|
| 163 |
|
| 164 |
+
def __call__(self, question: str) -> str:
|
| 165 |
try:
|
|
|
|
| 166 |
messages = [HumanMessage(content=question)]
|
| 167 |
result = self.agent.invoke({"messages": messages})
|
| 168 |
+
answer = result["messages"][-1].content
|
|
|
|
| 169 |
return self._clean_answer(answer)
|
| 170 |
except Exception as e:
|
| 171 |
return f"Agent error: {e}"
|
| 172 |
|
| 173 |
def _clean_answer(self, answer: str) -> str:
|
|
|
|
| 174 |
answer = answer.strip()
|
| 175 |
if "final answer:" in answer.lower():
|
| 176 |
parts = re.split(r'final answer:', answer, flags=re.IGNORECASE)
|
| 177 |
if len(parts) > 1:
|
| 178 |
answer = parts[-1].strip()
|
| 179 |
+
prefixes = ["The answer is", "Answer:", "Result:", "Solution:"]
|
|
|
|
| 180 |
for prefix in prefixes:
|
| 181 |
if answer.lower().startswith(prefix.lower()):
|
| 182 |
answer = answer[len(prefix):].strip()
|
| 183 |
if answer.startswith(':'):
|
| 184 |
answer = answer[1:].strip()
|
| 185 |
break
|
|
|
|
|
|
|
| 186 |
return answer
|
| 187 |
|
| 188 |
+
# ------------------ RUN AND SUBMIT ------------------
|
| 189 |
|
| 190 |
+
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
| 191 |
|
| 192 |
+
def run_and_submit_all(profile: gr.OAuthProfile | None):
|
| 193 |
+
if profile:
|
| 194 |
+
username = profile.username
|
| 195 |
+
else:
|
| 196 |
+
return "Please login to Hugging Face first.", None
|
| 197 |
|
| 198 |
+
space_id = os.getenv("SPACE_ID") or "your_space_username/your_space_name"
|
| 199 |
+
api_url = DEFAULT_API_URL
|
| 200 |
+
questions_url = f"{api_url}/questions"
|
| 201 |
+
submit_url = f"{api_url}/submit"
|
| 202 |
|
| 203 |
+
# Instantiate GaiaAgent
|
| 204 |
+
try:
|
| 205 |
+
agent = GaiaAgent()
|
| 206 |
+
except Exception as e:
|
| 207 |
+
return f"Error initializing GaiaAgent: {e}", None
|
| 208 |
|
| 209 |
+
# Fetch questions
|
| 210 |
+
try:
|
| 211 |
+
response = requests.get(questions_url, timeout=15)
|
| 212 |
+
response.raise_for_status()
|
| 213 |
+
questions_data = response.json()
|
| 214 |
+
except Exception as e:
|
| 215 |
+
return f"Error fetching questions: {e}", None
|
| 216 |
+
|
| 217 |
+
# Run agent on questions
|
| 218 |
+
answers_payload = []
|
| 219 |
+
results_log = []
|
| 220 |
+
for item in questions_data:
|
| 221 |
+
task_id = item.get("task_id")
|
| 222 |
+
question_text = item.get("question")
|
| 223 |
+
if not task_id or question_text is None:
|
| 224 |
+
continue
|
| 225 |
+
try:
|
| 226 |
+
submitted_answer = agent(question_text)
|
| 227 |
+
answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
|
| 228 |
+
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
|
| 229 |
+
except Exception as e:
|