Update app.py
Browse files
app.py
CHANGED
@@ -40,18 +40,35 @@ def extract_text_from_pdf(pdf_path):
|
|
40 |
text += page.get_text()
|
41 |
return text
|
42 |
|
43 |
-
def summarize_text_with_chat(text, max_length=10000):
|
44 |
-
|
45 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
46 |
trimmed_text = text[:max_length]
|
47 |
-
response =
|
48 |
-
model="gpt-4o",
|
49 |
messages=[
|
50 |
{"role": "system", "content": "次の文書を要約してください。必ず'## タイトル', '## 要約', '## 専門用語解説'を記載してください。"},
|
51 |
{"role": "user", "content": trimmed_text}
|
52 |
],
|
53 |
-
|
54 |
-
max_tokens=2000
|
55 |
)
|
56 |
summary_text = response.choices[0].message.content
|
57 |
total_token = response.usage.total_tokens
|
|
|
40 |
text += page.get_text()
|
41 |
return text
|
42 |
|
43 |
+
# def summarize_text_with_chat(text, max_length=10000):
|
44 |
+
# """OpenAIのChat APIを使ってテキストを要約。"""
|
45 |
+
# openai.api_key = os.getenv('OPEN_AI_API_KEYS')
|
46 |
+
# trimmed_text = text[:max_length]
|
47 |
+
# response = openai.chat.completions.create(
|
48 |
+
# model="gpt-4o",
|
49 |
+
# messages=[
|
50 |
+
# {"role": "system", "content": "次の文書を要約してください。必ず'## タイトル', '## 要約', '## 専門用語解説'を記載してください。"},
|
51 |
+
# {"role": "user", "content": trimmed_text}
|
52 |
+
# ],
|
53 |
+
# temperature=0.7,
|
54 |
+
# max_tokens=2000
|
55 |
+
# )
|
56 |
+
# summary_text = response.choices[0].message.content
|
57 |
+
# total_token = response.usage.total_tokens
|
58 |
+
# return summary_text, total_token
|
59 |
+
|
60 |
+
def summarize_text_with_chat(text, max_length=5000):
|
61 |
+
"""GroqのLlama3 70Bを使ってテキストを要約。"""
|
62 |
+
client = Groq(
|
63 |
+
api_key=os.environ.get("GROQ_API_KEY"),
|
64 |
+
)
|
65 |
trimmed_text = text[:max_length]
|
66 |
+
response = client.chat.completions.create(
|
|
|
67 |
messages=[
|
68 |
{"role": "system", "content": "次の文書を要約してください。必ず'## タイトル', '## 要約', '## 専門用語解説'を記載してください。"},
|
69 |
{"role": "user", "content": trimmed_text}
|
70 |
],
|
71 |
+
model="llama3-70b-8192",
|
|
|
72 |
)
|
73 |
summary_text = response.choices[0].message.content
|
74 |
total_token = response.usage.total_tokens
|