errchh commited on
Commit
5682fff
·
1 Parent(s): 2cb2810

fix stop iternation

Browse files
Files changed (3) hide show
  1. app.py +11 -2
  2. backup/agent.py +198 -0
  3. backup/app.py +202 -0
app.py CHANGED
@@ -22,8 +22,17 @@ class BasicAgent:
22
  print(f"Agent received question (first 50 chars): {question[:50]}...")
23
  messages = [HumanMessage(content=question)]
24
  result = self.graph.invoke({"messages": messages})
25
- answer = result['messages'][-1].content
26
- return answer
 
 
 
 
 
 
 
 
 
27
 
28
  def run_and_submit_all( profile: gr.OAuthProfile | None):
29
  """
 
22
  print(f"Agent received question (first 50 chars): {question[:50]}...")
23
  messages = [HumanMessage(content=question)]
24
  result = self.graph.invoke({"messages": messages})
25
+ # Extract the final answer message from the state
26
+ # The final message should be the content of the last AIMessage
27
+ submitted_answer = "Could not get answer from agent output." # Default if extraction fails
28
+ if result and "messages" in result:
29
+ for msg in reversed(result["messages"]): # Look for the last AI message
30
+ # Check if the message is an AIMessage and has content
31
+ if isinstance(msg, AIMessage) and hasattr(msg, 'content') and msg.content:
32
+ submitted_answer = msg.content
33
+ break # Found the last AI message content
34
+
35
+ return submitted_answer
36
 
37
  def run_and_submit_all( profile: gr.OAuthProfile | None):
38
  """
backup/agent.py ADDED
@@ -0,0 +1,198 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # import libraries for langgraph, huggingface
2
+ import os
3
+ from dotenv import load_dotenv
4
+ from typing import TypedDict, List, Dict, Any, Optional, Annotated
5
+
6
+ from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint, HuggingFaceEmbeddings
7
+
8
+ from langgraph.graph import StateGraph, MessagesState, START, END
9
+ from langgraph.graph.message import add_messages
10
+ from langchain_core.messages import SystemMessage, HumanMessage, AnyMessage, AIMessage
11
+ from langchain_core.messages.ai import subtract_usage
12
+
13
+ from langchain.tools import Tool
14
+ from langchain_core.tools import tool
15
+ from langchain_community.tools import WikipediaQueryRun
16
+ from langchain_community.utilities import WikipediaAPIWrapper
17
+ from langchain_community.utilities import SerpAPIWrapper
18
+ from langchain_community.utilities import ArxivAPIWrapper
19
+ from langchain_community.retrievers import BM25Retriever
20
+
21
+ from langgraph.prebuilt import ToolNode, tools_condition
22
+
23
+
24
+ # load environment variables
25
+ load_dotenv()
26
+ HUGGINGFACEHUB_API_TOKEN = os.getenv("HUGGINGFACEHUB_API_TOKEN")
27
+
28
+
29
+ # maths tool
30
+ @tool
31
+ def add(a:int, b:int) -> int:
32
+ """add two numbers.
33
+ args:
34
+ a: first int
35
+ b: second int
36
+ """
37
+ return a + b
38
+
39
+
40
+ @tool
41
+ def subtract(a:int, b:int) -> int:
42
+ """subtract two numbers.
43
+ args:
44
+ a: first int
45
+ b: second int
46
+ """
47
+ return a - b
48
+
49
+
50
+ @tool
51
+ def multiply(a:int, b:int) -> int:
52
+ """multiply two numbers.
53
+ args:
54
+ a: first int
55
+ b: second int
56
+ """
57
+ return a * b
58
+
59
+
60
+ @tool
61
+ def divide(a:int, b:int) -> float:
62
+ """divide two numbers.
63
+ args:
64
+ a: first int
65
+ b: second int
66
+ """
67
+ try:
68
+ # Attempt the division
69
+ result = a / b
70
+ return result
71
+ except ZeroDivisionError:
72
+ # Handle the case where b is zero
73
+ raise ValueError("Cannot divide by zero.")
74
+
75
+
76
+ @tool
77
+ def modulus(a:int, b:int) -> int:
78
+ """modulus remainder of two numbers.
79
+ args:
80
+ a: first int
81
+ b: second int
82
+ """
83
+ return a % b
84
+
85
+
86
+ # wikipedia search tool
87
+ @tool
88
+ def search_wiki(query: str) -> Dict[str, str]:
89
+ """search wikipedia with a query
90
+ args:
91
+ query: a search query
92
+ """
93
+ docs = WikipediaQueryRun(api_wrapper=WikipediaAPIWrapper())
94
+ docs.run(query)
95
+ formatted_result = f'<Document source="{docs.metadata["source"]}" page="{docs.metadata.get("page", "")}"/>\n{docs.page_content}\n</Document>'
96
+ return {"wiki_results": formatted_result}
97
+
98
+
99
+ # internet search tool
100
+ @tool
101
+ def search_web(query: str) -> Dict[str, str]:
102
+ """search internet with a query
103
+ args:
104
+ query: a search query
105
+ """
106
+ docs = SerpAPIWrapper()
107
+ docs.run(query)
108
+ formatted_result = f'<Document source="{docs.metadata["source"]}" page="{docs.metadata.get("page", "")}"/>\n{docs.page_content}\n</Document>'
109
+ return {"wiki_results": formatted_result}
110
+
111
+
112
+ # ArXiv search tool
113
+ @tool
114
+ def search_arxiv(query: str) -> Dict[str, str]:
115
+ """search ArXiv for the paper with the given identifier
116
+ args:
117
+ query: a search identifier
118
+ """
119
+ arxiv = ArxivAPIWrapper()
120
+ docs = arxiv.run(query)
121
+ formatted_result = f'<Document source="{docs.metadata["source"]}" page="{docs.metadata.get("page", "")}"/>\n{docs.page_content}\n</Document>'
122
+ return {"wiki_results": formatted_result}
123
+
124
+
125
+ # build retriever
126
+ # bm25_retriever = BM25Retriever.from_documents(docs)
127
+
128
+
129
+ # load system prompt from file
130
+ with open("system_prompt.txt", "r", encoding="utf-8") as f:
131
+ system_prompt = f.read()
132
+
133
+
134
+ # init system message
135
+ sys_msg = SystemMessage(content=system_prompt)
136
+
137
+
138
+ tools = [
139
+ add,
140
+ subtract,
141
+ multiply,
142
+ divide,
143
+ modulus,
144
+ search_wiki,
145
+ search_web,
146
+ search_arxiv
147
+ ]
148
+
149
+
150
+ # build graph function
151
+ def build_graph():
152
+ # llm
153
+ llm = HuggingFaceEndpoint(
154
+ repo_id = "microsoft/Phi-4-reasoning-plus",
155
+ huggingfacehub_api_token=HUGGINGFACEHUB_API_TOKEN,
156
+ )
157
+
158
+ chat = ChatHuggingFace(llm=llm, verbose=False)
159
+
160
+ # bind tools to llm
161
+ chat_with_tools = chat.bind_tools(tools)
162
+
163
+ # generate AgentState and Agent graph
164
+ class AgentState(TypedDict):
165
+ messages: Annotated[list[AnyMessage], add_messages]
166
+
167
+ def assistant(state: AgentState):
168
+ return {
169
+ "messages": [chat_with_tools.invoke(state["messages"])],
170
+ }
171
+
172
+ # build graph
173
+ builder = StateGraph(AgentState)
174
+
175
+ # define nodes
176
+ builder.add_node("assistant", assistant)
177
+ builder.add_node("tools", ToolNode(tools))
178
+
179
+ # define edges
180
+ builder.add_edge(START, "assistant")
181
+ builder.add_conditional_edges(
182
+ "assistant",
183
+ # If the latest message requires a tool, route to tools
184
+ # Otherwise, provide a direct response
185
+ tools_condition,
186
+ )
187
+ builder.add_edge("tools", "assistant")
188
+
189
+ return builder.compile()
190
+
191
+
192
+ if __name__ == "__main__":
193
+ question = "When was a picture of St. Thomas Aquinas first added to the Wikipedia page on the Principle of double effect?"
194
+ graph = build_graph()
195
+ messages = [HumanMessage(content=question)]
196
+ messages = graph.invoke({"messages": messages})
197
+ for m in messages["messages"]:
198
+ m.pretty_print()
backup/app.py ADDED
@@ -0,0 +1,202 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import os
2
+ import gradio as gr
3
+ import requests
4
+ import inspect
5
+ import pandas as pd
6
+
7
+ from langchain_core.messages import HumanMessage
8
+ from agent import build_graph
9
+
10
+ # (Keep Constants as is)
11
+ # --- Constants ---
12
+ DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
13
+
14
+ # --- Basic Agent Definition ---
15
+ # ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
16
+ class BasicAgent:
17
+ def __init__(self):
18
+ print("BasicAgent initialized.")
19
+ self.graph = build_graph()
20
+
21
+ def __call__(self, question: str) -> str:
22
+ print(f"Agent received question (first 50 chars): {question[:50]}...")
23
+ messages = [HumanMessage(content=question)]
24
+ result = self.graph.invoke({"messages": messages})
25
+ answer = result['messages'][-1].content
26
+ return answer
27
+
28
+ def run_and_submit_all( profile: gr.OAuthProfile | None):
29
+ """
30
+ Fetches all questions, runs the BasicAgent on them, submits all answers,
31
+ and displays the results.
32
+ """
33
+ # --- Determine HF Space Runtime URL and Repo URL ---
34
+ space_id = os.getenv("SPACE_ID") # Get the SPACE_ID for sending link to the code
35
+
36
+ if profile:
37
+ username= f"{profile.username}"
38
+ print(f"User logged in: {username}")
39
+ else:
40
+ print("User not logged in.")
41
+ return "Please Login to Hugging Face with the button.", None
42
+
43
+ api_url = DEFAULT_API_URL
44
+ questions_url = f"{api_url}/questions"
45
+ submit_url = f"{api_url}/submit"
46
+
47
+ # 1. Instantiate Agent ( modify this part to create your agent)
48
+ try:
49
+ agent = BasicAgent()
50
+ except Exception as e:
51
+ print(f"Error instantiating agent: {e}")
52
+ return f"Error initializing agent: {e}", None
53
+ # In the case of an app running as a hugging Face space, this link points toward your codebase ( usefull for others so please keep it public)
54
+ agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
55
+ print(agent_code)
56
+
57
+ # 2. Fetch Questions
58
+ print(f"Fetching questions from: {questions_url}")
59
+ try:
60
+ response = requests.get(questions_url, timeout=15)
61
+ response.raise_for_status()
62
+ questions_data = response.json()
63
+ if not questions_data:
64
+ print("Fetched questions list is empty.")
65
+ return "Fetched questions list is empty or invalid format.", None
66
+ print(f"Fetched {len(questions_data)} questions.")
67
+ except requests.exceptions.RequestException as e:
68
+ print(f"Error fetching questions: {e}")
69
+ return f"Error fetching questions: {e}", None
70
+ except requests.exceptions.JSONDecodeError as e:
71
+ print(f"Error decoding JSON response from questions endpoint: {e}")
72
+ print(f"Response text: {response.text[:500]}")
73
+ return f"Error decoding server response for questions: {e}", None
74
+ except Exception as e:
75
+ print(f"An unexpected error occurred fetching questions: {e}")
76
+ return f"An unexpected error occurred fetching questions: {e}", None
77
+
78
+ # 3. Run your Agent
79
+ results_log = []
80
+ answers_payload = []
81
+ print(f"Running agent on {len(questions_data)} questions...")
82
+ for item in questions_data:
83
+ task_id = item.get("task_id")
84
+ question_text = item.get("question")
85
+ if not task_id or question_text is None:
86
+ print(f"Skipping item with missing task_id or question: {item}")
87
+ continue
88
+ try:
89
+ submitted_answer = agent(question_text)
90
+ answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
91
+ results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
92
+ except Exception as e:
93
+ print(f"Error running agent on task {task_id}: {e}")
94
+ results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})
95
+
96
+ if not answers_payload:
97
+ print("Agent did not produce any answers to submit.")
98
+ return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
99
+
100
+ # 4. Prepare Submission
101
+ submission_data = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload}
102
+ status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
103
+ print(status_update)
104
+
105
+ # 5. Submit
106
+ print(f"Submitting {len(answers_payload)} answers to: {submit_url}")
107
+ try:
108
+ response = requests.post(submit_url, json=submission_data, timeout=60)
109
+ response.raise_for_status()
110
+ result_data = response.json()
111
+ final_status = (
112
+ f"Submission Successful!\n"
113
+ f"User: {result_data.get('username')}\n"
114
+ f"Overall Score: {result_data.get('score', 'N/A')}% "
115
+ f"({result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')} correct)\n"
116
+ f"Message: {result_data.get('message', 'No message received.')}"
117
+ )
118
+ print("Submission successful.")
119
+ results_df = pd.DataFrame(results_log)
120
+ return final_status, results_df
121
+ except requests.exceptions.HTTPError as e:
122
+ error_detail = f"Server responded with status {e.response.status_code}."
123
+ try:
124
+ error_json = e.response.json()
125
+ error_detail += f" Detail: {error_json.get('detail', e.response.text)}"
126
+ except requests.exceptions.JSONDecodeError:
127
+ error_detail += f" Response: {e.response.text[:500]}"
128
+ status_message = f"Submission Failed: {error_detail}"
129
+ print(status_message)
130
+ results_df = pd.DataFrame(results_log)
131
+ return status_message, results_df
132
+ except requests.exceptions.Timeout:
133
+ status_message = "Submission Failed: The request timed out."
134
+ print(status_message)
135
+ results_df = pd.DataFrame(results_log)
136
+ return status_message, results_df
137
+ except requests.exceptions.RequestException as e:
138
+ status_message = f"Submission Failed: Network error - {e}"
139
+ print(status_message)
140
+ results_df = pd.DataFrame(results_log)
141
+ return status_message, results_df
142
+ except Exception as e:
143
+ status_message = f"An unexpected error occurred during submission: {e}"
144
+ print(status_message)
145
+ results_df = pd.DataFrame(results_log)
146
+ return status_message, results_df
147
+
148
+
149
+ # --- Build Gradio Interface using Blocks ---
150
+ with gr.Blocks() as demo:
151
+ gr.Markdown("# Basic Agent Evaluation Runner")
152
+ gr.Markdown(
153
+ """
154
+ **Instructions:**
155
+
156
+ 1. Please clone this space, then modify the code to define your agent's logic, the tools, the necessary packages, etc ...
157
+ 2. Log in to your Hugging Face account using the button below. This uses your HF username for submission.
158
+ 3. Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see the score.
159
+
160
+ ---
161
+ **Disclaimers:**
162
+ Once clicking on the "submit button, it can take quite some time ( this is the time for the agent to go through all the questions).
163
+ This space provides a basic setup and is intentionally sub-optimal to encourage you to develop your own, more robust solution. For instance for the delay process of the submit button, a solution could be to cache the answers and submit in a seperate action or even to answer the questions in async.
164
+ """
165
+ )
166
+
167
+ gr.LoginButton()
168
+
169
+ run_button = gr.Button("Run Evaluation & Submit All Answers")
170
+
171
+ status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, interactive=False)
172
+ # Removed max_rows=10 from DataFrame constructor
173
+ results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
174
+
175
+ run_button.click(
176
+ fn=run_and_submit_all,
177
+ outputs=[status_output, results_table]
178
+ )
179
+
180
+ if __name__ == "__main__":
181
+ print("\n" + "-"*30 + " App Starting " + "-"*30)
182
+ # Check for SPACE_HOST and SPACE_ID at startup for information
183
+ space_host_startup = os.getenv("SPACE_HOST")
184
+ space_id_startup = os.getenv("SPACE_ID") # Get SPACE_ID at startup
185
+
186
+ if space_host_startup:
187
+ print(f"✅ SPACE_HOST found: {space_host_startup}")
188
+ print(f" Runtime URL should be: https://{space_host_startup}.hf.space")
189
+ else:
190
+ print("ℹ️ SPACE_HOST environment variable not found (running locally?).")
191
+
192
+ if space_id_startup: # Print repo URLs if SPACE_ID is found
193
+ print(f"✅ SPACE_ID found: {space_id_startup}")
194
+ print(f" Repo URL: https://huggingface.co/spaces/{space_id_startup}")
195
+ print(f" Repo Tree URL: https://huggingface.co/spaces/{space_id_startup}/tree/main")
196
+ else:
197
+ print("ℹ️ SPACE_ID environment variable not found (running locally?). Repo URL cannot be determined.")
198
+
199
+ print("-"*(60 + len(" App Starting ")) + "\n")
200
+
201
+ print("Launching Gradio Interface for Basic Agent Evaluation...")
202
+ demo.launch(debug=True, share=False)