Update app.py
Browse files
app.py
CHANGED
@@ -1,83 +1,21 @@
|
|
1 |
-
# app.py
|
2 |
|
3 |
import os
|
4 |
-
import
|
5 |
import pandas as pd
|
|
|
6 |
import gradio as gr
|
7 |
-
from
|
8 |
-
|
9 |
-
from langchain.agents import initialize_agent, AgentType, Tool
|
10 |
-
from langchain_community.tools import DuckDuckGoSearchRun
|
11 |
-
from langchain_community.tools.wikipedia.tool import WikipediaQueryRun
|
12 |
-
from langchain_experimental.tools.python.tool import PythonREPLTool
|
13 |
-
from langchain_community.tools.youtube.search import YouTubeSearchTool
|
14 |
-
from langchain_community.document_loaders import YoutubeLoader
|
15 |
-
from langchain_openai import ChatOpenAI
|
16 |
-
from langchain.tools import tool
|
17 |
-
|
18 |
-
# --- LangChain LLM and Tools Setup --- #
|
19 |
-
llm = ChatOpenAI(model="gpt-4o", temperature=0)
|
20 |
-
|
21 |
-
@tool
|
22 |
-
def get_yt_transcript(url: str) -> str:
|
23 |
-
loader = YoutubeLoader.from_youtube_url(url, add_video_info=False)
|
24 |
-
docs = loader.load()
|
25 |
-
return " ".join(doc.page_content for doc in docs)
|
26 |
-
|
27 |
-
@tool
|
28 |
-
def reverse_sentence_logic(sentence: str) -> str:
|
29 |
-
return sentence[::-1]
|
30 |
-
|
31 |
-
@tool
|
32 |
-
def commutativity_counterexample(_: str) -> str:
|
33 |
-
return "a, b, c"
|
34 |
-
|
35 |
-
@tool
|
36 |
-
def malko_winner(_: str) -> str:
|
37 |
-
return "Uroš"
|
38 |
|
39 |
-
@tool
|
40 |
-
def ray_actor_answer(_: str) -> str:
|
41 |
-
return "Filip"
|
42 |
-
|
43 |
-
@tool
|
44 |
-
def chess_position_hint(_: str) -> str:
|
45 |
-
return "Qd1+"
|
46 |
-
|
47 |
-
@tool
|
48 |
-
def default_award_number(_: str) -> str:
|
49 |
-
return "80NSSC21K1030"
|
50 |
-
|
51 |
-
# Add your LangChain tools here
|
52 |
-
langchain_tools: List[Tool] = [
|
53 |
-
DuckDuckGoSearchRun(),
|
54 |
-
WikipediaQueryRun(api_wrapper=None),
|
55 |
-
YouTubeSearchTool(),
|
56 |
-
Tool(name="youtube_transcript", func=get_yt_transcript, description="Transcribe YouTube video from URL"),
|
57 |
-
PythonREPLTool(),
|
58 |
-
reverse_sentence_logic,
|
59 |
-
commutativity_counterexample,
|
60 |
-
malko_winner,
|
61 |
-
ray_actor_answer,
|
62 |
-
chess_position_hint,
|
63 |
-
default_award_number,
|
64 |
-
]
|
65 |
-
|
66 |
-
agent = initialize_agent(tools=langchain_tools, llm=llm, agent=AgentType.OPENAI_MULTI_FUNCTIONS, verbose=False)
|
67 |
-
|
68 |
-
# --- Hugging Face Evaluation Integration --- #
|
69 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
70 |
|
71 |
-
class
|
72 |
def __init__(self):
|
73 |
-
print("
|
74 |
|
75 |
def __call__(self, question: str) -> str:
|
76 |
-
print(f"
|
77 |
-
|
78 |
-
return agent.run(question)
|
79 |
-
except Exception as e:
|
80 |
-
return f"[ERROR] {str(e)}"
|
81 |
|
82 |
def run_and_submit_all(profile: gr.OAuthProfile | None):
|
83 |
space_id = os.getenv("SPACE_ID")
|
@@ -88,7 +26,6 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
88 |
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main" if space_id else ""
|
89 |
api_url = DEFAULT_API_URL
|
90 |
|
91 |
-
# Fetch questions
|
92 |
try:
|
93 |
response = requests.get(f"{api_url}/questions", timeout=15)
|
94 |
response.raise_for_status()
|
@@ -98,44 +35,45 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
98 |
|
99 |
answers_payload = []
|
100 |
results_log = []
|
101 |
-
|
102 |
|
103 |
for item in questions_data:
|
104 |
q = item.get("question")
|
105 |
task_id = item.get("task_id")
|
106 |
try:
|
107 |
-
a =
|
108 |
except Exception as e:
|
109 |
a = f"ERROR: {e}"
|
110 |
answers_payload.append({"task_id": task_id, "submitted_answer": a})
|
111 |
results_log.append({"Task ID": task_id, "Question": q, "Submitted Answer": a})
|
112 |
|
113 |
-
submission_data = {
|
|
|
|
|
|
|
|
|
114 |
|
115 |
-
# Submit answers
|
116 |
try:
|
117 |
response = requests.post(f"{api_url}/submit", json=submission_data, timeout=60)
|
118 |
response.raise_for_status()
|
119 |
result_data = response.json()
|
120 |
-
|
121 |
-
f"Submission Successful!\n"
|
122 |
f"User: {result_data.get('username')}\n"
|
123 |
f"Score: {result_data.get('score')}%\n"
|
124 |
f"Correct: {result_data.get('correct_count')}/{result_data.get('total_attempted')}\n"
|
125 |
f"Message: {result_data.get('message')}"
|
126 |
)
|
127 |
-
return
|
128 |
except Exception as e:
|
129 |
return f"Submission failed: {e}", pd.DataFrame(results_log)
|
130 |
|
131 |
-
# --- Gradio UI --- #
|
132 |
with gr.Blocks() as demo:
|
133 |
-
gr.Markdown("#
|
134 |
gr.LoginButton()
|
135 |
run_btn = gr.Button("Run Evaluation & Submit All Answers")
|
136 |
status_box = gr.Textbox(label="Status", lines=5)
|
137 |
result_table = gr.DataFrame(label="Agent Answers")
|
138 |
-
|
139 |
run_btn.click(fn=run_and_submit_all, outputs=[status_box, result_table])
|
140 |
|
141 |
demo.launch(debug=True)
|
|
|
1 |
+
# app.py
|
2 |
|
3 |
import os
|
4 |
+
import asyncio
|
5 |
import pandas as pd
|
6 |
+
import requests
|
7 |
import gradio as gr
|
8 |
+
from agent import answer_question
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
9 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
11 |
|
12 |
+
class GAIALlamaAgent:
|
13 |
def __init__(self):
|
14 |
+
print("Initialized LlamaIndex Agent")
|
15 |
|
16 |
def __call__(self, question: str) -> str:
|
17 |
+
print(f"Received question: {question[:60]}...")
|
18 |
+
return asyncio.run(answer_question(question))
|
|
|
|
|
|
|
19 |
|
20 |
def run_and_submit_all(profile: gr.OAuthProfile | None):
|
21 |
space_id = os.getenv("SPACE_ID")
|
|
|
26 |
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main" if space_id else ""
|
27 |
api_url = DEFAULT_API_URL
|
28 |
|
|
|
29 |
try:
|
30 |
response = requests.get(f"{api_url}/questions", timeout=15)
|
31 |
response.raise_for_status()
|
|
|
35 |
|
36 |
answers_payload = []
|
37 |
results_log = []
|
38 |
+
agent = GAIALlamaAgent()
|
39 |
|
40 |
for item in questions_data:
|
41 |
q = item.get("question")
|
42 |
task_id = item.get("task_id")
|
43 |
try:
|
44 |
+
a = agent(q)
|
45 |
except Exception as e:
|
46 |
a = f"ERROR: {e}"
|
47 |
answers_payload.append({"task_id": task_id, "submitted_answer": a})
|
48 |
results_log.append({"Task ID": task_id, "Question": q, "Submitted Answer": a})
|
49 |
|
50 |
+
submission_data = {
|
51 |
+
"username": username,
|
52 |
+
"agent_code": agent_code,
|
53 |
+
"answers": answers_payload
|
54 |
+
}
|
55 |
|
|
|
56 |
try:
|
57 |
response = requests.post(f"{api_url}/submit", json=submission_data, timeout=60)
|
58 |
response.raise_for_status()
|
59 |
result_data = response.json()
|
60 |
+
status = (
|
61 |
+
f"✅ Submission Successful!\n"
|
62 |
f"User: {result_data.get('username')}\n"
|
63 |
f"Score: {result_data.get('score')}%\n"
|
64 |
f"Correct: {result_data.get('correct_count')}/{result_data.get('total_attempted')}\n"
|
65 |
f"Message: {result_data.get('message')}"
|
66 |
)
|
67 |
+
return status, pd.DataFrame(results_log)
|
68 |
except Exception as e:
|
69 |
return f"Submission failed: {e}", pd.DataFrame(results_log)
|
70 |
|
|
|
71 |
with gr.Blocks() as demo:
|
72 |
+
gr.Markdown("# LlamaIndex GAIA Agent – Evaluation Portal")
|
73 |
gr.LoginButton()
|
74 |
run_btn = gr.Button("Run Evaluation & Submit All Answers")
|
75 |
status_box = gr.Textbox(label="Status", lines=5)
|
76 |
result_table = gr.DataFrame(label="Agent Answers")
|
|
|
77 |
run_btn.click(fn=run_and_submit_all, outputs=[status_box, result_table])
|
78 |
|
79 |
demo.launch(debug=True)
|