zaf786 commited on
Commit
379be03
Β·
verified Β·
1 Parent(s): 3bf1c7e

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +172 -107
app.py CHANGED
@@ -1,142 +1,207 @@
 
1
  import os
 
2
  import gradio as gr
3
  import requests
4
  import pandas as pd
 
 
5
 
6
- # GAIA evaluation API
7
- DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
8
-
9
- # --- Basic Agent Definition (HTTP to Hugging Face Inference API) ---
10
- class BasicAgent:
11
- def __init__(self):
12
- self.api_token = os.getenv("HUGGINGFACEHUB_API_TOKEN")
13
- if not self.api_token:
14
- raise ValueError("Missing HUGGINGFACEHUB_API_TOKEN in Secrets")
15
-
16
- # Mistral-7B-Instruct inference endpoint
17
- self.model_url = "https://api-inference.huggingface.co/models/mistralai/Mistral-7B-Instruct-v0.1"
18
- self.headers = {
19
- "Authorization": f"Bearer {self.api_token}",
20
- "Content-Type": "application/json"
21
- }
22
- print("βœ… BasicAgent initialized with Mistral-7B-Instruct-v0.1")
23
-
24
- def __call__(self, question: str) -> str:
25
- # Wrap for instruct format (Mistral expects <s>[INST] … [/INST])
26
- prompt = f"<s>[INST] {question.strip()} [/INST]"
27
- print(f"πŸ“₯ Prompting: {prompt[:60]}…")
28
-
29
- payload = {
30
- "inputs": prompt,
31
- "options": {"wait_for_model": True},
32
- "parameters": {"max_new_tokens": 200}
33
- }
34
 
35
- try:
36
- resp = requests.post(self.model_url, headers=self.headers, json=payload, timeout=60)
37
- resp.raise_for_status()
38
- result = resp.json()
39
-
40
- # The API returns a list of {generated_text: …}
41
- if isinstance(result, list) and "generated_text" in result[0]:
42
- answer = result[0]["generated_text"].strip()
43
- elif isinstance(result, dict) and "generated_text" in result:
44
- answer = result["generated_text"].strip()
45
- else:
46
- print("⚠️ Unexpected format:", result)
47
- answer = "error: unexpected format"
48
-
49
- print(f"πŸ“€ Answer: {answer}")
50
- return answer
51
-
52
- except Exception as e:
53
- print(f"❌ Inference error: {e}")
54
- return f"error: {e}"
55
 
 
 
 
56
 
57
- # --- Fetch GAIA questions, run agent, submit answers ---
58
- def run_and_submit_all(profile: gr.OAuthProfile | None):
59
- if not profile:
60
- return "Please log in to Hugging Face.", None
61
 
62
- username = profile.username
63
- print(f"πŸ” Logged in as {username}")
64
 
65
- # Endpoints
66
- questions_url = f"{DEFAULT_API_URL}/questions"
67
- submit_url = f"{DEFAULT_API_URL}/submit"
68
- space_id = os.getenv("SPACE_ID")
69
- agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
70
 
71
- # 1. Initialize agent
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
72
  try:
73
  agent = BasicAgent()
74
  except Exception as e:
75
- return f"Agent init error: {e}", None
76
-
77
- # 2. Fetch questions
 
 
 
 
 
78
  try:
79
- r = requests.get(questions_url, timeout=15)
80
- r.raise_for_status()
81
- questions = r.json()
82
- except Exception as e:
 
 
 
 
 
83
  return f"Error fetching questions: {e}", None
 
 
 
 
 
 
 
84
 
85
- # 3. Run agent
86
- results = []
87
  answers_payload = []
88
- for item in questions:
 
89
  task_id = item.get("task_id")
90
- q_text = item.get("question")
91
- if not task_id or not q_text:
 
92
  continue
93
-
94
- ans = agent(q_text)
95
- answers_payload.append({"task_id": task_id, "submitted_answer": ans})
96
- results.append({"Task ID": task_id, "Question": q_text, "Answer": ans})
 
 
 
97
 
98
  if not answers_payload:
99
- return "No answers generated.", pd.DataFrame(results)
100
-
101
- # 4. Submit
102
- submission = {
103
- "username": username,
104
- "agent_code": agent_code,
105
- "answers": answers_payload
106
- }
 
 
107
  try:
108
- r = requests.post(submit_url, json=submission, timeout=60)
109
- r.raise_for_status()
110
- data = r.json()
111
- status = (
112
- f"πŸŽ‰ Submission Successful!\n"
113
- f"User: {data.get('username')}\n"
114
- f"Score: {data.get('score')}% "
115
- f"({data.get('correct_count')}/{data.get('total_attempted')} correct)\n"
116
- f"{data.get('message')}"
117
  )
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
118
  except Exception as e:
119
- status = f"Submission error: {e}"
120
-
121
- return status, pd.DataFrame(results)
 
122
 
123
 
124
- # --- Gradio UI ---
125
  with gr.Blocks() as demo:
126
- gr.Markdown("# πŸ€– GAIA Level 1 Agent (Mistral-7B API)")
127
  gr.Markdown(
128
- "1. Log in with Hugging Face \n"
129
- "2. Click **Run Evaluation & Submit All Answers** \n"
130
- "3. View your score on the leaderboard"
 
 
 
 
 
 
 
131
  )
 
132
  gr.LoginButton()
133
- run_btn = gr.Button("Run Evaluation & Submit All Answers")
134
- status_box = gr.Textbox(label="Result", lines=5, interactive=False)
135
- results_table = gr.DataFrame(label="QA Log")
136
 
137
- run_btn.click(fn=run_and_submit_all, outputs=[status_box, results_table])
 
 
 
 
138
 
 
 
 
 
139
 
140
  if __name__ == "__main__":
141
- print("πŸš€ Launching GAIA Agent App")
142
- demo.launch(debug=True, share=False)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """ Basic Agent Evaluation Runner"""
2
  import os
3
+ import inspect
4
  import gradio as gr
5
  import requests
6
  import pandas as pd
7
+ from langchain_core.messages import HumanMessage
8
+ from agent import build_graph
9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
10
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
11
 
12
+ # (Keep Constants as is)
13
+ # --- Constants ---
14
+ DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
15
 
16
+ # --- Basic Agent Definition ---
17
+ # ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
 
 
18
 
 
 
19
 
20
+ class BasicAgent:
21
+ """A langgraph agent."""
22
+ def __init__(self):
23
+ print("BasicAgent initialized.")
24
+ self.graph = build_graph()
25
 
26
+ def __call__(self, question: str) -> str:
27
+ print(f"Agent received question (first 50 chars): {question[:50]}...")
28
+ messages = [HumanMessage(content=question)]
29
+ result = self.graph.invoke({"messages": messages})
30
+ answer = result['messages'][-1].content
31
+ return answer # kein [14:] mehr nΓΆtig!
32
+
33
+
34
+
35
+ def run_and_submit_all( profile: gr.OAuthProfile | None):
36
+ """
37
+ Fetches all questions, runs the BasicAgent on them, submits all answers,
38
+ and displays the results.
39
+ """
40
+ # --- Determine HF Space Runtime URL and Repo URL ---
41
+ space_id = os.getenv("SPACE_ID") # Get the SPACE_ID for sending link to the code
42
+
43
+ if profile:
44
+ username= f"{profile.username}"
45
+ print(f"User logged in: {username}")
46
+ else:
47
+ print("User not logged in.")
48
+ return "Please Login to Hugging Face with the button.", None
49
+
50
+ api_url = DEFAULT_API_URL
51
+ questions_url = f"{api_url}/questions"
52
+ submit_url = f"{api_url}/submit"
53
+
54
+ # 1. Instantiate Agent ( modify this part to create your agent)
55
  try:
56
  agent = BasicAgent()
57
  except Exception as e:
58
+ print(f"Error instantiating agent: {e}")
59
+ return f"Error initializing agent: {e}", None
60
+ # 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)
61
+ agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
62
+ print(agent_code)
63
+
64
+ # 2. Fetch Questions
65
+ print(f"Fetching questions from: {questions_url}")
66
  try:
67
+ response = requests.get(questions_url, timeout=15)
68
+ response.raise_for_status()
69
+ questions_data = response.json()
70
+ if not questions_data:
71
+ print("Fetched questions list is empty.")
72
+ return "Fetched questions list is empty or invalid format.", None
73
+ print(f"Fetched {len(questions_data)} questions.")
74
+ except requests.exceptions.RequestException as e:
75
+ print(f"Error fetching questions: {e}")
76
  return f"Error fetching questions: {e}", None
77
+ except requests.exceptions.JSONDecodeError as e:
78
+ print(f"Error decoding JSON response from questions endpoint: {e}")
79
+ print(f"Response text: {response.text[:500]}")
80
+ return f"Error decoding server response for questions: {e}", None
81
+ except Exception as e:
82
+ print(f"An unexpected error occurred fetching questions: {e}")
83
+ return f"An unexpected error occurred fetching questions: {e}", None
84
 
85
+ # 3. Run your Agent
86
+ results_log = []
87
  answers_payload = []
88
+ print(f"Running agent on {len(questions_data)} questions...")
89
+ for item in questions_data:
90
  task_id = item.get("task_id")
91
+ question_text = item.get("question")
92
+ if not task_id or question_text is None:
93
+ print(f"Skipping item with missing task_id or question: {item}")
94
  continue
95
+ try:
96
+ submitted_answer = agent(question_text)
97
+ answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
98
+ results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
99
+ except Exception as e:
100
+ print(f"Error running agent on task {task_id}: {e}")
101
+ results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})
102
 
103
  if not answers_payload:
104
+ print("Agent did not produce any answers to submit.")
105
+ return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
106
+
107
+ # 4. Prepare Submission
108
+ submission_data = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload}
109
+ status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
110
+ print(status_update)
111
+
112
+ # 5. Submit
113
+ print(f"Submitting {len(answers_payload)} answers to: {submit_url}")
114
  try:
115
+ response = requests.post(submit_url, json=submission_data, timeout=60)
116
+ response.raise_for_status()
117
+ result_data = response.json()
118
+ final_status = (
119
+ f"Submission Successful!\n"
120
+ f"User: {result_data.get('username')}\n"
121
+ f"Overall Score: {result_data.get('score', 'N/A')}% "
122
+ f"({result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')} correct)\n"
123
+ f"Message: {result_data.get('message', 'No message received.')}"
124
  )
125
+ print("Submission successful.")
126
+ results_df = pd.DataFrame(results_log)
127
+ return final_status, results_df
128
+ except requests.exceptions.HTTPError as e:
129
+ error_detail = f"Server responded with status {e.response.status_code}."
130
+ try:
131
+ error_json = e.response.json()
132
+ error_detail += f" Detail: {error_json.get('detail', e.response.text)}"
133
+ except requests.exceptions.JSONDecodeError:
134
+ error_detail += f" Response: {e.response.text[:500]}"
135
+ status_message = f"Submission Failed: {error_detail}"
136
+ print(status_message)
137
+ results_df = pd.DataFrame(results_log)
138
+ return status_message, results_df
139
+ except requests.exceptions.Timeout:
140
+ status_message = "Submission Failed: The request timed out."
141
+ print(status_message)
142
+ results_df = pd.DataFrame(results_log)
143
+ return status_message, results_df
144
+ except requests.exceptions.RequestException as e:
145
+ status_message = f"Submission Failed: Network error - {e}"
146
+ print(status_message)
147
+ results_df = pd.DataFrame(results_log)
148
+ return status_message, results_df
149
  except Exception as e:
150
+ status_message = f"An unexpected error occurred during submission: {e}"
151
+ print(status_message)
152
+ results_df = pd.DataFrame(results_log)
153
+ return status_message, results_df
154
 
155
 
156
+ # --- Build Gradio Interface using Blocks ---
157
  with gr.Blocks() as demo:
158
+ gr.Markdown("# Basic Agent Evaluation Runner")
159
  gr.Markdown(
160
+ """
161
+ **Instructions:**
162
+ 1. Please clone this space, then modify the code to define your agent's logic, the tools, the necessary packages, etc ...
163
+ 2. Log in to your Hugging Face account using the button below. This uses your HF username for submission.
164
+ 3. Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see the score.
165
+ ---
166
+ **Disclaimers:**
167
+ 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).
168
+ 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.
169
+ """
170
  )
171
+
172
  gr.LoginButton()
 
 
 
173
 
174
+ run_button = gr.Button("Run Evaluation & Submit All Answers")
175
+
176
+ status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, interactive=False)
177
+ # Removed max_rows=10 from DataFrame constructor
178
+ results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
179
 
180
+ run_button.click(
181
+ fn=run_and_submit_all,
182
+ outputs=[status_output, results_table]
183
+ )
184
 
185
  if __name__ == "__main__":
186
+ print("\n" + "-"*30 + " App Starting " + "-"*30)
187
+ # Check for SPACE_HOST and SPACE_ID at startup for information
188
+ space_host_startup = os.getenv("SPACE_HOST")
189
+ space_id_startup = os.getenv("SPACE_ID") # Get SPACE_ID at startup
190
+
191
+ if space_host_startup:
192
+ print(f"βœ… SPACE_HOST found: {space_host_startup}")
193
+ print(f" Runtime URL should be: https://{space_host_startup}.hf.space")
194
+ else:
195
+ print("ℹ️ SPACE_HOST environment variable not found (running locally?).")
196
+
197
+ if space_id_startup: # Print repo URLs if SPACE_ID is found
198
+ print(f"βœ… SPACE_ID found: {space_id_startup}")
199
+ print(f" Repo URL: https://huggingface.co/spaces/{space_id_startup}")
200
+ print(f" Repo Tree URL: https://huggingface.co/spaces/{space_id_startup}/tree/main")
201
+ else:
202
+ print("ℹ️ SPACE_ID environment variable not found (running locally?). Repo URL cannot be determined.")
203
+
204
+ print("-"*(60 + len(" App Starting ")) + "\n")
205
+
206
+ print("Launching Gradio Interface for Basic Agent Evaluation...")
207
+ demo.launch(debug=True, share=False)