Spaces:
Sleeping
Sleeping
| import os | |
| import tempfile | |
| import time | |
| from collections import deque | |
| import gradio as gr | |
| import requests | |
| import inspect | |
| import pandas as pd | |
| import spaces | |
| from huggingface_hub import hf_hub_download | |
| from smolagents import ( | |
| ActionStep, | |
| LiteLLMModel, | |
| PythonInterpreterTool, | |
| ToolCallingAgent, | |
| WebSearchTool, | |
| VisitWebpageTool, | |
| WikipediaSearchTool, | |
| ) | |
| # (Keep Constants as is) | |
| # --- Constants --- | |
| DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space" | |
| # The scoring server's own /files/{task_id} endpoint is broken (confirmed via | |
| # a direct request and a matching open PR on agents-course/Unit4_scoring: | |
| # it stores GAIA attachment paths as HF-Hub-repo-relative paths but checks | |
| # them as local filesystem paths, so every lookup 404s: "No file path | |
| # associated with task_id ..."). Attachments live directly in the gated | |
| # gaia-benchmark/GAIA dataset instead; fetch them from there using the | |
| # logged-in user's own OAuth token (requires that account to have requested | |
| # access to the dataset, and the `gated-repos` OAuth scope in README.md). | |
| GAIA_DATASET_REPO = "gaia-benchmark/GAIA" | |
| GAIA_DATASET_SUBDIR = "2023/validation" | |
| def _zerogpu_startup_check(): | |
| # This Space runs on ZeroGPU hardware but the agent below only makes | |
| # network calls (Groq API, web search) and never touches CUDA. | |
| # ZeroGPU requires at least one @spaces.GPU function to be declared, | |
| # so this no-op satisfies that check without spending any GPU quota | |
| # (it is never actually invoked). | |
| return None | |
| # GAIA benchmark expects a terse, exact-match final answer. | |
| GAIA_ANSWER_FORMAT_INSTRUCTIONS = """You are a general AI assistant. I will ask you a question. | |
| Report your thoughts, and finish your work by calling final_answer() with your answer. | |
| Your final answer should be a number OR as few words as possible OR a comma separated list of numbers and/or strings. | |
| If you are asked for a number, don't use commas to write it, and don't use units such as $ or % unless specified otherwise. | |
| If you are asked for a string, don't use articles or abbreviations (e.g. for cities), and write digits in plain text unless specified otherwise. | |
| If you are asked for a comma separated list, apply the above rules to each element depending on whether it's a number or a string. | |
| Work efficiently: if a search or lookup doesn't find what you need after 1-2 tries, try a meaningfully different approach rather than repeating similar queries, and give your best-guess final_answer rather than exhausting all steps. | |
| Never call visit_webpage on a wikipedia.org URL - it always returns 403 Forbidden. Use the wikipedia_search tool for Wikipedia content instead. | |
| """ | |
| class TokenPacer: | |
| """ | |
| Step_callback that tracks actual token usage per step in a trailing 60s | |
| window and sleeps as needed to stay under a tokens-per-minute budget. | |
| Necessary because a single call's fixed overhead (system prompt + tool | |
| schemas + question, before any tool output) already runs 2,400+ tokens | |
| on this agent and grows with context - smolagents' native | |
| requests_per_minute throttle can't account for that since it only paces | |
| call count, not size. This is a secondary safeguard: Cerebras' free-tier | |
| TPM budget (30,000) is generous enough that this should rarely trigger. | |
| """ | |
| def __init__(self, tokens_per_minute_budget: int = 5000): | |
| self.tokens_per_minute_budget = tokens_per_minute_budget | |
| self._usage_window: deque[tuple[float, int]] = deque() | |
| def __call__(self, memory_step: ActionStep, agent: ToolCallingAgent) -> None: | |
| now = time.monotonic() | |
| usage = getattr(memory_step, "token_usage", None) | |
| tokens = (usage.input_tokens + usage.output_tokens) if usage else 0 | |
| self._usage_window.append((now, tokens)) | |
| cutoff = now - 60 | |
| while self._usage_window and self._usage_window[0][0] < cutoff: | |
| self._usage_window.popleft() | |
| window_tokens = sum(t for _, t in self._usage_window) | |
| if window_tokens > self.tokens_per_minute_budget and self._usage_window: | |
| wait = 60 - (now - self._usage_window[0][0]) + 0.5 | |
| if wait > 0: | |
| print(f"Token pacer: {window_tokens} tokens in the last 60s (budget {self.tokens_per_minute_budget}), sleeping {wait:.1f}s") | |
| time.sleep(wait) | |
| class MemoryTrimmer: | |
| """ | |
| Step_callback that collapses old tool outputs in the agent's memory. | |
| The agent re-sends the *entire* step history on every call, so input | |
| tokens grow every single step (2k -> 5k -> 8k -> ... -> 20k+ by step 6), | |
| which both wrecks the TPM budget and slows every later step far more | |
| than it needs to. Keeps the most recent `keep_recent` steps' tool | |
| outputs intact (the agent still needs that detail) and truncates older | |
| ones to a short placeholder, keeping per-call token cost roughly flat | |
| across a run instead of growing unbounded. | |
| """ | |
| def __init__(self, keep_recent: int = 2, max_old_observation_chars: int = 300): | |
| self.keep_recent = keep_recent | |
| self.max_old_observation_chars = max_old_observation_chars | |
| def __call__(self, memory_step: ActionStep, agent: ToolCallingAgent) -> None: | |
| action_steps = [step for step in agent.memory.steps if isinstance(step, ActionStep)] | |
| for step in action_steps[:-self.keep_recent]: | |
| observations = getattr(step, "observations", None) | |
| if observations and len(observations) > self.max_old_observation_chars: | |
| step.observations = observations[:self.max_old_observation_chars] + " [...older output truncated to save context]" | |
| # --- Basic Agent Definition --- | |
| # ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------ | |
| class BasicAgent: | |
| def __init__(self): | |
| # gpt-oss-120b's daily token quota (1M, separate per model on | |
| # Cerebras) got fully used up mid-session. zai-glm-4.7 has its own | |
| # untouched 1M/day quota, so switch there rather than waiting ~24h | |
| # for gpt-oss-120b's to reset. | |
| model_id = os.getenv("AGENT_MODEL_ID", "cerebras/zai-glm-4.7") | |
| api_key = os.getenv("CEREBRAS_API_KEY") | |
| if not api_key: | |
| print("Warning: CEREBRAS_API_KEY is not set - the agent will fail to call the model.") | |
| # Cerebras' free tier: 5 RPM, 30,000 TPM, 1,000,000 TPD - far more | |
| # TPM headroom than Groq's free tier (6000-12000), which was the | |
| # actual bottleneck there. requests_per_minute paces call frequency | |
| # natively; retry=False avoids smolagents' internal retry-on-error | |
| # firing extra invisible HTTP calls on a rate-limited response. | |
| self.model = LiteLLMModel( | |
| model_id=model_id, | |
| api_key=api_key, | |
| temperature=0, | |
| requests_per_minute=float(os.getenv("RATE_LIMIT_RPM", "4.5")), | |
| retry=False, | |
| ) | |
| # ToolCallingAgent (not CodeAgent): gpt-oss-120b kept mixing reasoning | |
| # text into its code instead of cleanly wrapping it in <code> tags, | |
| # so CodeAgent's regex-based code-block parser failed on a large | |
| # fraction of steps - wasted steps that burned tokens/rate-limit | |
| # budget without making progress, and led to hallucinated final | |
| # answers. ToolCallingAgent uses the provider's native structured | |
| # tool-calling instead of parsing free-form text, which sidesteps | |
| # this failure mode entirely. PythonInterpreterTool replaces the | |
| # code-execution capability CodeAgent had built in. | |
| self.agent = ToolCallingAgent( | |
| model=self.model, | |
| tools=[ | |
| WebSearchTool(), | |
| VisitWebpageTool(max_output_length=3000), # default 40000 chars blows the TPM budget in one call | |
| WikipediaSearchTool(content_type="summary"), # "text" (default) returns the full article | |
| PythonInterpreterTool(authorized_imports=[ | |
| "pandas", "numpy", "math", "re", "json", "itertools", | |
| "collections", "statistics", "datetime", "io", "openpyxl", "PIL", | |
| ]), | |
| ], | |
| # Reverted 12 -> 7: raising it caused several questions to run | |
| # 12-13 steps each with growing context, and the cumulative | |
| # token usage blew Cerebras' *daily* quota partway through the | |
| # run (confirmed: "Tokens per day limit exceeded" starting | |
| # around question 13) - every question after that auto-failed, | |
| # dropping the score from 25% to 10%. 7 steps was empirically | |
| # better: fewer wasted tokens per question, more questions | |
| # actually get a real shot before the daily budget runs out. | |
| max_steps=7, | |
| step_callbacks=[ | |
| MemoryTrimmer(), | |
| TokenPacer(tokens_per_minute_budget=int(os.getenv("RATE_LIMIT_TOKENS_PER_MINUTE", "25000"))), | |
| ], | |
| ) | |
| print("BasicAgent initialized.") | |
| def __call__(self, question: str, file_path: str | None = None) -> str: | |
| print(f"Agent received question (first 50 chars): {question[:50]}...") | |
| task = GAIA_ANSWER_FORMAT_INSTRUCTIONS + f"\nQuestion: {question}" | |
| if file_path: | |
| task += ( | |
| f"\n\nA file for this question was downloaded locally to: {file_path}\n" | |
| "Open/read it with Python (pandas, openpyxl, PIL, etc. as appropriate) to answer the question." | |
| ) | |
| try: | |
| answer = self.agent.run(task) | |
| except Exception as e: | |
| print(f"Agent run failed: {e}") | |
| return f"AGENT ERROR: {e}" | |
| answer = str(answer).strip() | |
| print(f"Agent returning answer: {answer}") | |
| return answer | |
| def run_and_submit_all( profile: gr.OAuthProfile | None, oauth_token: gr.OAuthToken | None): | |
| """ | |
| Fetches all questions, runs the BasicAgent on them, submits all answers, | |
| and displays the results. | |
| """ | |
| # --- Determine HF Space Runtime URL and Repo URL --- | |
| space_id = os.getenv("SPACE_ID") # Get the SPACE_ID for sending link to the code | |
| if profile: | |
| username= f"{profile.username}" | |
| print(f"User logged in: {username}") | |
| else: | |
| print("User not logged in.") | |
| return "Please Login to Hugging Face with the button.", None | |
| api_url = DEFAULT_API_URL | |
| questions_url = f"{api_url}/questions" | |
| submit_url = f"{api_url}/submit" | |
| # 1. Instantiate Agent ( modify this part to create your agent) | |
| try: | |
| agent = BasicAgent() | |
| except Exception as e: | |
| print(f"Error instantiating agent: {e}") | |
| return f"Error initializing agent: {e}", None | |
| # 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) | |
| agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main" | |
| print(agent_code) | |
| # 2. Fetch Questions | |
| print(f"Fetching questions from: {questions_url}") | |
| try: | |
| response = requests.get(questions_url, timeout=15) | |
| response.raise_for_status() | |
| questions_data = response.json() | |
| if not questions_data: | |
| print("Fetched questions list is empty.") | |
| return "Fetched questions list is empty or invalid format.", None | |
| print(f"Fetched {len(questions_data)} questions.") | |
| except requests.exceptions.RequestException as e: | |
| print(f"Error fetching questions: {e}") | |
| return f"Error fetching questions: {e}", None | |
| except requests.exceptions.JSONDecodeError as e: | |
| print(f"Error decoding JSON response from questions endpoint: {e}") | |
| print(f"Response text: {response.text[:500]}") | |
| return f"Error decoding server response for questions: {e}", None | |
| except Exception as e: | |
| print(f"An unexpected error occurred fetching questions: {e}") | |
| return f"An unexpected error occurred fetching questions: {e}", None | |
| # 3. Run your Agent | |
| results_log = [] | |
| answers_payload = [] | |
| print(f"Running agent on {len(questions_data)} questions...") | |
| with tempfile.TemporaryDirectory() as tmp_dir: | |
| for item in questions_data: | |
| task_id = item.get("task_id") | |
| question_text = item.get("question") | |
| file_name = item.get("file_name") | |
| if not task_id or question_text is None: | |
| print(f"Skipping item with missing task_id or question: {item}") | |
| continue | |
| file_path = None | |
| if file_name: | |
| try: | |
| file_path = hf_hub_download( | |
| repo_id=GAIA_DATASET_REPO, | |
| repo_type="dataset", | |
| filename=f"{GAIA_DATASET_SUBDIR}/{file_name}", | |
| token=oauth_token.token if oauth_token else None, | |
| ) | |
| except Exception as e: | |
| print(f"Could not download attached file for task {task_id} from {GAIA_DATASET_REPO}: {e}") | |
| # Fall back to the scoring server's own endpoint, in case | |
| # it's since been fixed (see GAIA_DATASET_REPO comment above). | |
| try: | |
| file_response = requests.get(f"{api_url}/files/{task_id}", timeout=30) | |
| file_response.raise_for_status() | |
| file_path = os.path.join(tmp_dir, file_name) | |
| with open(file_path, "wb") as f: | |
| f.write(file_response.content) | |
| except requests.exceptions.RequestException as e2: | |
| print(f"Fallback download also failed for task {task_id}: {e2}") | |
| file_path = None | |
| try: | |
| submitted_answer = agent(question_text, file_path=file_path) | |
| answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer}) | |
| results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer}) | |
| except Exception as e: | |
| print(f"Error running agent on task {task_id}: {e}") | |
| results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"}) | |
| if not answers_payload: | |
| print("Agent did not produce any answers to submit.") | |
| return "Agent did not produce any answers to submit.", pd.DataFrame(results_log) | |
| # 4. Prepare Submission | |
| submission_data = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload} | |
| status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..." | |
| print(status_update) | |
| # 5. Submit | |
| print(f"Submitting {len(answers_payload)} answers to: {submit_url}") | |
| try: | |
| response = requests.post(submit_url, json=submission_data, timeout=60) | |
| response.raise_for_status() | |
| result_data = response.json() | |
| final_status = ( | |
| f"Submission Successful!\n" | |
| f"User: {result_data.get('username')}\n" | |
| f"Overall Score: {result_data.get('score', 'N/A')}% " | |
| f"({result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')} correct)\n" | |
| f"Message: {result_data.get('message', 'No message received.')}" | |
| ) | |
| print("Submission successful.") | |
| results_df = pd.DataFrame(results_log) | |
| return final_status, results_df | |
| except requests.exceptions.HTTPError as e: | |
| error_detail = f"Server responded with status {e.response.status_code}." | |
| try: | |
| error_json = e.response.json() | |
| error_detail += f" Detail: {error_json.get('detail', e.response.text)}" | |
| except requests.exceptions.JSONDecodeError: | |
| error_detail += f" Response: {e.response.text[:500]}" | |
| status_message = f"Submission Failed: {error_detail}" | |
| print(status_message) | |
| results_df = pd.DataFrame(results_log) | |
| return status_message, results_df | |
| except requests.exceptions.Timeout: | |
| status_message = "Submission Failed: The request timed out." | |
| print(status_message) | |
| results_df = pd.DataFrame(results_log) | |
| return status_message, results_df | |
| except requests.exceptions.RequestException as e: | |
| status_message = f"Submission Failed: Network error - {e}" | |
| print(status_message) | |
| results_df = pd.DataFrame(results_log) | |
| return status_message, results_df | |
| except Exception as e: | |
| status_message = f"An unexpected error occurred during submission: {e}" | |
| print(status_message) | |
| results_df = pd.DataFrame(results_log) | |
| return status_message, results_df | |
| # --- Build Gradio Interface using Blocks --- | |
| with gr.Blocks() as demo: | |
| gr.Markdown("# Basic Agent Evaluation Runner") | |
| gr.Markdown( | |
| """ | |
| **Instructions:** | |
| 1. Please clone this space, then modify the code to define your agent's logic, the tools, the necessary packages, etc ... | |
| 2. Log in to your Hugging Face account using the button below. This uses your HF username for submission. | |
| 3. Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see the score. | |
| --- | |
| **Disclaimers:** | |
| 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). | |
| 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. | |
| **Setup:** This agent calls Cerebras (zai-glm-4.7) via `smolagents`. Get a free key at https://cloud.cerebras.ai and set it as the `CEREBRAS_API_KEY` secret in this Space's settings before running. | |
| """ | |
| ) | |
| gr.LoginButton() | |
| run_button = gr.Button("Run Evaluation & Submit All Answers") | |
| status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, interactive=False) | |
| # Removed max_rows=10 from DataFrame constructor | |
| results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True) | |
| run_button.click( | |
| fn=run_and_submit_all, | |
| outputs=[status_output, results_table] | |
| ) | |
| if __name__ == "__main__": | |
| print("\n" + "-"*30 + " App Starting " + "-"*30) | |
| # Check for SPACE_HOST and SPACE_ID at startup for information | |
| space_host_startup = os.getenv("SPACE_HOST") | |
| space_id_startup = os.getenv("SPACE_ID") # Get SPACE_ID at startup | |
| if space_host_startup: | |
| print(f"✅ SPACE_HOST found: {space_host_startup}") | |
| print(f" Runtime URL should be: https://{space_host_startup}.hf.space") | |
| else: | |
| print("ℹ️ SPACE_HOST environment variable not found (running locally?).") | |
| if space_id_startup: # Print repo URLs if SPACE_ID is found | |
| print(f"✅ SPACE_ID found: {space_id_startup}") | |
| print(f" Repo URL: https://huggingface.co/spaces/{space_id_startup}") | |
| print(f" Repo Tree URL: https://huggingface.co/spaces/{space_id_startup}/tree/main") | |
| else: | |
| print("ℹ️ SPACE_ID environment variable not found (running locally?). Repo URL cannot be determined.") | |
| print("-"*(60 + len(" App Starting ")) + "\n") | |
| print("Launching Gradio Interface for Basic Agent Evaluation...") | |
| demo.launch(debug=True, share=False) |