Spaces:
Build error
Build error
Update app.py
Browse files
app.py
CHANGED
@@ -1,77 +1,45 @@
|
|
1 |
-
import gradio as gr
|
2 |
-
import csv
|
3 |
-
import re
|
4 |
-
import tempfile
|
5 |
-
import os
|
6 |
-
import requests
|
7 |
-
|
8 |
-
# 1. Load system prompt from a file
|
9 |
-
with open("system_instructions.txt", "r", encoding="utf-8") as f:
|
10 |
-
ECO_PROMPT = f.read()
|
11 |
-
|
12 |
-
# DeepSeek API configuration
|
13 |
-
DEEPSEEK_API_KEY = os.environ.get("DEEPSEEK_API_KEY")
|
14 |
-
DEEPSEEK_API_URL = "https://api.deepseek.com/v1/chat/completions" # Verify actual API endpoint
|
15 |
-
|
16 |
-
def score_qa(question, answer):
|
17 |
-
"""Query DeepSeek API to get a score for Q&A pair"""
|
18 |
-
try:
|
19 |
-
# Format the prompt using our template
|
20 |
-
prompt = ECO_PROMPT.format(question=question, answer=answer)
|
21 |
-
|
22 |
-
headers = {
|
23 |
-
"Authorization": f"Bearer {DEEPSEEK_API_KEY}",
|
24 |
-
"Content-Type": "application/json"
|
25 |
-
}
|
26 |
-
|
27 |
-
payload = {
|
28 |
-
"model": "deepseek-chat", # Verify correct model name
|
29 |
-
"messages": [{
|
30 |
-
"role": "user",
|
31 |
-
"content": prompt
|
32 |
-
}],
|
33 |
-
"temperature": 0.1, # More deterministic output for scoring
|
34 |
-
"max_tokens": 5
|
35 |
-
}
|
36 |
-
|
37 |
-
response = requests.post(DEEPSEEK_API_URL, json=payload, headers=headers)
|
38 |
-
response.raise_for_status()
|
39 |
-
|
40 |
-
# Parse response (adjust based on actual API response structure)
|
41 |
-
output = response.json()['choices'][0]['message']['content']
|
42 |
-
|
43 |
-
# Extract score using same logic as before
|
44 |
-
match = re.search(r"\d+", output)
|
45 |
-
return int(match.group(0)) if match else 1
|
46 |
-
|
47 |
-
except Exception as e:
|
48 |
-
print(f"API Error: {str(e)}")
|
49 |
-
return 1 # Fallback score on error
|
50 |
-
|
51 |
def judge_ecolinguistics_from_csv(csv_file):
|
52 |
-
"""
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
#
|
57 |
-
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
-
|
74 |
-
)
|
75 |
-
|
76 |
-
|
77 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
def judge_ecolinguistics_from_csv(csv_file):
|
2 |
+
"""
|
3 |
+
Reads CSV of Q&A pairs, scores each,
|
4 |
+
returns a new CSV and a percentage score.
|
5 |
+
"""
|
6 |
+
# Read the CSV
|
7 |
+
rows = []
|
8 |
+
with open(csv_file.name, "r", encoding="utf-8") as f:
|
9 |
+
reader = csv.DictReader(f)
|
10 |
+
for r in reader:
|
11 |
+
rows.append(r)
|
12 |
+
|
13 |
+
results = []
|
14 |
+
total_score = 0
|
15 |
+
|
16 |
+
# Score each Q&A
|
17 |
+
for r in rows:
|
18 |
+
q_num = r.get("question_number", "")
|
19 |
+
question = r.get("question", "")
|
20 |
+
answer = r.get("answer", "")
|
21 |
+
|
22 |
+
sc = score_qa(question, answer)
|
23 |
+
total_score += sc
|
24 |
+
results.append({"question_number": q_num, "score": sc})
|
25 |
+
|
26 |
+
# Write results to a new CSV in a temp file
|
27 |
+
with tempfile.NamedTemporaryFile(mode="w", delete=False, suffix=".csv", encoding="utf-8") as out_file:
|
28 |
+
fieldnames = ["question_number", "score"]
|
29 |
+
writer = csv.DictWriter(out_file, fieldnames=fieldnames)
|
30 |
+
writer.writeheader()
|
31 |
+
for row in results:
|
32 |
+
writer.writerow(row)
|
33 |
+
writer.writerow({"question_number": "Total", "score": total_score})
|
34 |
+
out_path = out_file.name # This is the critical line
|
35 |
+
|
36 |
+
# Compute % score: total / (num_questions * 5) * 100
|
37 |
+
num_questions = len(rows)
|
38 |
+
if num_questions > 0:
|
39 |
+
percentage = (total_score / (num_questions * 5)) * 100
|
40 |
+
else:
|
41 |
+
percentage = 0.0
|
42 |
+
|
43 |
+
percentage_display = f"<h2 style='font-size:2em; color:blue;'>Overall Score: {percentage:.1f}%</h2>"
|
44 |
+
|
45 |
+
return out_path, percentage_display # Now out_path is defined
|