Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -15,7 +15,6 @@ intents.messages = True
|
|
15 |
# hf_client = InferenceClient("meta-llama/Meta-Llama-3-70B-Instruct", token=os.getenv("HF_TOKEN"))
|
16 |
hf_client = InferenceClient("CohereForAI/c4ai-command-r-plus", token=os.getenv("HF_TOKEN"))
|
17 |
|
18 |
-
|
19 |
# λν νμ€ν 리λ₯Ό μ μ₯ν λ³μ
|
20 |
conversation_history = []
|
21 |
|
@@ -31,13 +30,14 @@ class MyClient(discord.Client):
|
|
31 |
logging.info('μμ μ λ©μμ§λ 무μν©λλ€.')
|
32 |
return
|
33 |
|
34 |
-
logging.debug(f'Receiving message: {message.content}')
|
35 |
response = await generate_response(message.content)
|
36 |
await message.channel.send(response)
|
37 |
|
38 |
async def generate_response(user_input):
|
39 |
-
system_message = "DISCORDμμ μ¬μ©μλ€μ μ§λ¬Έμ
|
40 |
system_prefix = """
|
|
|
41 |
λ°λμ νκΈλ‘ λ΅λ³νμμμ€. μΆλ ₯μ λμμ°κΈ°λ₯Ό νκ³ markdownμΌλ‘ μΆλ ₯νλΌ.
|
42 |
μ§λ¬Έμ μ ν©ν λ΅λ³μ μ 곡νλ©°, κ°λ₯ν ν ꡬ체μ μ΄κ³ λμμ΄ λλ λ΅λ³μ μ 곡νμμμ€.
|
43 |
λͺ¨λ λ΅λ³μ νκΈλ‘ νκ³ , λν λ΄μ©μ κΈ°μ΅νμμμ€.
|
@@ -45,25 +45,29 @@ async def generate_response(user_input):
|
|
45 |
λ°λμ νκΈλ‘ λ΅λ³νμμμ€.
|
46 |
"""
|
47 |
|
|
|
48 |
# λν νμ€ν 리 κ΄λ¦¬
|
49 |
global conversation_history
|
50 |
conversation_history.append({"role": "user", "content": user_input})
|
|
|
51 |
|
52 |
messages = [{"role": "system", "content": f"{system_prefix} {system_message}"}] + conversation_history
|
|
|
53 |
|
54 |
# λκΈ° ν¨μλ₯Ό λΉλκΈ°λ‘ μ²λ¦¬νκΈ° μν λνΌ μ¬μ©, stream=trueλ‘ λ³κ²½
|
55 |
loop = asyncio.get_event_loop()
|
56 |
response = await loop.run_in_executor(None, lambda: hf_client.chat_completion(
|
57 |
-
messages, max_tokens=
|
58 |
|
59 |
# μ€νΈλ¦¬λ° μλ΅μ μ²λ¦¬νλ λ‘μ§ μΆκ°
|
60 |
full_response = ""
|
61 |
for part in response:
|
62 |
-
|
|
|
63 |
|
64 |
conversation_history.append({"role": "assistant", "content": full_response})
|
|
|
65 |
|
66 |
-
logging.debug(f'Model response: {full_response}')
|
67 |
return full_response
|
68 |
|
69 |
# λμ€μ½λ λ΄ μΈμ€ν΄μ€ μμ± λ° μ€ν
|
|
|
15 |
# hf_client = InferenceClient("meta-llama/Meta-Llama-3-70B-Instruct", token=os.getenv("HF_TOKEN"))
|
16 |
hf_client = InferenceClient("CohereForAI/c4ai-command-r-plus", token=os.getenv("HF_TOKEN"))
|
17 |
|
|
|
18 |
# λν νμ€ν 리λ₯Ό μ μ₯ν λ³μ
|
19 |
conversation_history = []
|
20 |
|
|
|
30 |
logging.info('μμ μ λ©μμ§λ 무μν©λλ€.')
|
31 |
return
|
32 |
|
33 |
+
logging.debug(f'Receiving message: {message.content}') # μ
λ ₯ λ©μμ§ λ‘κΉ
|
34 |
response = await generate_response(message.content)
|
35 |
await message.channel.send(response)
|
36 |
|
37 |
async def generate_response(user_input):
|
38 |
+
system_message = "DISCORDμμ μ¬μ©μλ€μ μ§λ¬Έμ λ΅νλ μ λ¬Έ AI μ΄μμ€ν΄νΈμ
λλ€. λνλ₯Ό κ³μ μ΄μ΄κ°κ³ , μ΄μ μλ΅μ μ°Έκ³ νμμμ€."
|
39 |
system_prefix = """
|
40 |
+
|
41 |
λ°λμ νκΈλ‘ λ΅λ³νμμμ€. μΆλ ₯μ λμμ°κΈ°λ₯Ό νκ³ markdownμΌλ‘ μΆλ ₯νλΌ.
|
42 |
μ§λ¬Έμ μ ν©ν λ΅λ³μ μ 곡νλ©°, κ°λ₯ν ν ꡬ체μ μ΄κ³ λμμ΄ λλ λ΅λ³μ μ 곡νμμμ€.
|
43 |
λͺ¨λ λ΅λ³μ νκΈλ‘ νκ³ , λν λ΄μ©μ κΈ°μ΅νμμμ€.
|
|
|
45 |
λ°λμ νκΈλ‘ λ΅λ³νμμμ€.
|
46 |
"""
|
47 |
|
48 |
+
|
49 |
# λν νμ€ν 리 κ΄λ¦¬
|
50 |
global conversation_history
|
51 |
conversation_history.append({"role": "user", "content": user_input})
|
52 |
+
logging.debug(f'Conversation history updated: {conversation_history}') # λν νμ€ν 리 λ‘κΉ
|
53 |
|
54 |
messages = [{"role": "system", "content": f"{system_prefix} {system_message}"}] + conversation_history
|
55 |
+
logging.debug(f'Messages to be sent to the model: {messages}') # λͺ¨λΈλ‘ μ μ‘λ λ©μμ§ λ‘κΉ
|
56 |
|
57 |
# λκΈ° ν¨μλ₯Ό λΉλκΈ°λ‘ μ²λ¦¬νκΈ° μν λνΌ μ¬μ©, stream=trueλ‘ λ³κ²½
|
58 |
loop = asyncio.get_event_loop()
|
59 |
response = await loop.run_in_executor(None, lambda: hf_client.chat_completion(
|
60 |
+
messages, max_tokens=1000, stream=True, temperature=0.7, top_p=0.85))
|
61 |
|
62 |
# μ€νΈλ¦¬λ° μλ΅μ μ²λ¦¬νλ λ‘μ§ μΆκ°
|
63 |
full_response = ""
|
64 |
for part in response:
|
65 |
+
if part.choices and part.choices[0].delta.content: # λΈνκ° μλμ§ νμΈ
|
66 |
+
full_response += part.choices[0].delta.content.strip()
|
67 |
|
68 |
conversation_history.append({"role": "assistant", "content": full_response})
|
69 |
+
logging.debug(f'Model response: {full_response}') # μλ΅ λ‘κΉ
|
70 |
|
|
|
71 |
return full_response
|
72 |
|
73 |
# λμ€μ½λ λ΄ μΈμ€ν΄μ€ μμ± λ° μ€ν
|