Spaces:
Sleeping
Sleeping
aelin commited on
Commit ·
a1074ab
1
Parent(s): 8ebfdcd
Adds answer caching and user score types
Browse files
_types.py
CHANGED
|
@@ -1,10 +1,23 @@
|
|
| 1 |
-
from typing import TypedDict, Optional, List
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
|
| 3 |
class Question(TypedDict):
|
| 4 |
task_id: str
|
| 5 |
question: str
|
| 6 |
file_name: Optional[str] = None
|
| 7 |
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from typing import TypedDict, Optional, List, Dict
|
| 2 |
+
|
| 3 |
+
class AnswerCacheEntry(TypedDict):
|
| 4 |
+
answer: str
|
| 5 |
+
isCorrect: bool
|
| 6 |
+
|
| 7 |
+
AnswerCache = Dict[str, AnswerCacheEntry]
|
| 8 |
|
| 9 |
class Question(TypedDict):
|
| 10 |
task_id: str
|
| 11 |
question: str
|
| 12 |
file_name: Optional[str] = None
|
| 13 |
|
| 14 |
+
class UserScore(TypedDict):
|
| 15 |
+
username: str
|
| 16 |
+
score: int
|
| 17 |
+
correct_count: int
|
| 18 |
+
total_attempted: int
|
| 19 |
+
message: str
|
| 20 |
+
timestamp: str
|
| 21 |
+
|
| 22 |
+
Questions = List[Question]
|
| 23 |
+
|
app.py
CHANGED
|
@@ -1,12 +1,14 @@
|
|
|
|
|
| 1 |
import os
|
| 2 |
import gradio as gr
|
| 3 |
import requests
|
| 4 |
import pandas as pd
|
| 5 |
-
from _types import Questions, Question
|
| 6 |
from llama_index.llms.huggingface_api import HuggingFaceInferenceAPI
|
| 7 |
from llama_index.core.agent.workflow import AgentWorkflow
|
| 8 |
from _tools import tools
|
| 9 |
import asyncio
|
|
|
|
| 10 |
|
| 11 |
# (Keep Constants as is)
|
| 12 |
# --- Constants ---
|
|
@@ -54,11 +56,19 @@ class BasicAgent:
|
|
| 54 |
as you want. The file name is: {file_name}.\n
|
| 55 |
"""
|
| 56 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 57 |
answer = await self.agent.run(prompt)
|
| 58 |
|
| 59 |
print(f"Agent returning answer: {answer}")
|
| 60 |
|
| 61 |
-
return answer
|
| 62 |
|
| 63 |
def instantiate_agent():
|
| 64 |
try:
|
|
@@ -124,13 +134,15 @@ async def fetch_file(question: Question) -> str | None:
|
|
| 124 |
async def run_agent_on_questions(agent: BasicAgent, questions_data: Questions):
|
| 125 |
results_log = []
|
| 126 |
answers_payload = []
|
| 127 |
-
|
| 128 |
print(f"Running agent on {len(questions_data)} questions...")
|
| 129 |
-
|
|
|
|
|
|
|
|
|
|
| 130 |
for item in questions_data:
|
| 131 |
task_id = item.get("task_id")
|
| 132 |
question_text = item.get("question")
|
| 133 |
-
|
| 134 |
if not task_id or question_text is None:
|
| 135 |
print(f"Skipping item with missing task_id or question: {item}")
|
| 136 |
continue
|
|
@@ -139,11 +151,15 @@ async def run_agent_on_questions(agent: BasicAgent, questions_data: Questions):
|
|
| 139 |
submitted_answer = await agent.run(item)
|
| 140 |
answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
|
| 141 |
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
|
|
|
|
|
|
|
| 142 |
|
| 143 |
except Exception as e:
|
| 144 |
print(f"Error running agent on task {task_id}: {e}")
|
| 145 |
-
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})
|
| 146 |
|
|
|
|
|
|
|
|
|
|
| 147 |
return answers_payload, results_log
|
| 148 |
|
| 149 |
def submit_answers(submit_url, submission_data, results_log):
|
|
@@ -152,7 +168,7 @@ def submit_answers(submit_url, submission_data, results_log):
|
|
| 152 |
try:
|
| 153 |
response = requests.post(submit_url, json=submission_data, timeout=60)
|
| 154 |
response.raise_for_status()
|
| 155 |
-
result_data = response.json()
|
| 156 |
|
| 157 |
final_status = (
|
| 158 |
f"Submission Successful!\n"
|
|
@@ -245,7 +261,6 @@ async def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
| 245 |
return submit_answers(submit_url, submission_data, results_log)
|
| 246 |
|
| 247 |
|
| 248 |
-
|
| 249 |
async def main():
|
| 250 |
await run_and_submit_all(profile=None)
|
| 251 |
|
|
|
|
| 1 |
+
|
| 2 |
import os
|
| 3 |
import gradio as gr
|
| 4 |
import requests
|
| 5 |
import pandas as pd
|
| 6 |
+
from _types import Questions, Question, UserScore
|
| 7 |
from llama_index.llms.huggingface_api import HuggingFaceInferenceAPI
|
| 8 |
from llama_index.core.agent.workflow import AgentWorkflow
|
| 9 |
from _tools import tools
|
| 10 |
import asyncio
|
| 11 |
+
from utils import cache_answers, update_cache_answer, get_cached_answer
|
| 12 |
|
| 13 |
# (Keep Constants as is)
|
| 14 |
# --- Constants ---
|
|
|
|
| 56 |
as you want. The file name is: {file_name}.\n
|
| 57 |
"""
|
| 58 |
|
| 59 |
+
|
| 60 |
+
# Retrieve the answer from the cache if it exists
|
| 61 |
+
cached = get_cached_answer(task_id)
|
| 62 |
+
if cached and cached.get("isCorrect") and cached.get("answer"):
|
| 63 |
+
print(f"Returning cached correct answer for task_id {task_id}: {cached['answer']}")
|
| 64 |
+
|
| 65 |
+
return str(cached["answer"])
|
| 66 |
+
|
| 67 |
answer = await self.agent.run(prompt)
|
| 68 |
|
| 69 |
print(f"Agent returning answer: {answer}")
|
| 70 |
|
| 71 |
+
return str(answer)
|
| 72 |
|
| 73 |
def instantiate_agent():
|
| 74 |
try:
|
|
|
|
| 134 |
async def run_agent_on_questions(agent: BasicAgent, questions_data: Questions):
|
| 135 |
results_log = []
|
| 136 |
answers_payload = []
|
|
|
|
| 137 |
print(f"Running agent on {len(questions_data)} questions...")
|
| 138 |
+
|
| 139 |
+
# Inicializa o cache com todas as respostas erradas (se ainda não existir)
|
| 140 |
+
cache_answers(questions_data)
|
| 141 |
+
|
| 142 |
for item in questions_data:
|
| 143 |
task_id = item.get("task_id")
|
| 144 |
question_text = item.get("question")
|
| 145 |
+
|
| 146 |
if not task_id or question_text is None:
|
| 147 |
print(f"Skipping item with missing task_id or question: {item}")
|
| 148 |
continue
|
|
|
|
| 151 |
submitted_answer = await agent.run(item)
|
| 152 |
answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
|
| 153 |
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
|
| 154 |
+
# Update the cache with the answer
|
| 155 |
+
update_cache_answer(task_id, submitted_answer, is_correct=False)
|
| 156 |
|
| 157 |
except Exception as e:
|
| 158 |
print(f"Error running agent on task {task_id}: {e}")
|
|
|
|
| 159 |
|
| 160 |
+
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})
|
| 161 |
+
update_cache_answer(task_id, f"AGENT ERROR: {e}", is_correct=False)
|
| 162 |
+
|
| 163 |
return answers_payload, results_log
|
| 164 |
|
| 165 |
def submit_answers(submit_url, submission_data, results_log):
|
|
|
|
| 168 |
try:
|
| 169 |
response = requests.post(submit_url, json=submission_data, timeout=60)
|
| 170 |
response.raise_for_status()
|
| 171 |
+
result_data: UserScore = response.json()
|
| 172 |
|
| 173 |
final_status = (
|
| 174 |
f"Submission Successful!\n"
|
|
|
|
| 261 |
return submit_answers(submit_url, submission_data, results_log)
|
| 262 |
|
| 263 |
|
|
|
|
| 264 |
async def main():
|
| 265 |
await run_and_submit_all(profile=None)
|
| 266 |
|
utils.py
ADDED
|
@@ -0,0 +1,43 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
|
| 2 |
+
import json
|
| 3 |
+
import os
|
| 4 |
+
from typing import Optional
|
| 5 |
+
from _types import AnswerCacheEntry, AnswerCache
|
| 6 |
+
|
| 7 |
+
CACHE_FILE = "answers_cache.json"
|
| 8 |
+
|
| 9 |
+
def load_cache(cache_path: str = CACHE_FILE) -> AnswerCache:
|
| 10 |
+
if not os.path.exists(cache_path):
|
| 11 |
+
return {}
|
| 12 |
+
with open(cache_path, "r", encoding="utf-8") as f:
|
| 13 |
+
try:
|
| 14 |
+
return json.load(f)
|
| 15 |
+
except json.JSONDecodeError:
|
| 16 |
+
return {}
|
| 17 |
+
|
| 18 |
+
def save_cache(cache: AnswerCache, cache_path: str = CACHE_FILE) -> None:
|
| 19 |
+
with open(cache_path, "w", encoding="utf-8") as f:
|
| 20 |
+
json.dump(cache, f, ensure_ascii=False, indent=2)
|
| 21 |
+
|
| 22 |
+
def cache_answers(questions, cache_path: str = CACHE_FILE) -> AnswerCache:
|
| 23 |
+
"""
|
| 24 |
+
Initializes the cache with all answers as incorrect.
|
| 25 |
+
Structure: { task_id: { answer: '', isCorrect: False } }
|
| 26 |
+
"""
|
| 27 |
+
cache: AnswerCache = {}
|
| 28 |
+
for q in questions:
|
| 29 |
+
task_id = q.get("task_id")
|
| 30 |
+
if task_id:
|
| 31 |
+
cache[task_id] = {"answer": "", "isCorrect": False}
|
| 32 |
+
save_cache(cache, cache_path)
|
| 33 |
+
return cache
|
| 34 |
+
|
| 35 |
+
def update_cache_answer(task_id: str, answer: str, is_correct: bool = False, cache_path: str = CACHE_FILE):
|
| 36 |
+
cache = load_cache(cache_path)
|
| 37 |
+
cache[task_id] = {"answer": answer, "isCorrect": is_correct}
|
| 38 |
+
save_cache(cache, cache_path)
|
| 39 |
+
|
| 40 |
+
|
| 41 |
+
def get_cached_answer(task_id: str, cache_path: str = CACHE_FILE) -> Optional[AnswerCacheEntry]:
|
| 42 |
+
cache = load_cache(cache_path)
|
| 43 |
+
return cache.get(task_id)
|