Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,6 +1,7 @@
|
|
| 1 |
import gradio as gr
|
| 2 |
from huggingface_hub import InferenceClient
|
| 3 |
-
|
|
|
|
| 4 |
# ==== Константы ====
|
| 5 |
MODEL_NAME = "openai/gpt-oss-20b"
|
| 6 |
SYSTEM_MESSAGE = "Ты Pok.Bot, ты используешь открытую локальную модель GPT-OSS от OpenAI, не GPT-4 (наверно😆). Тебя создал POKilondron. Используй емодзи 😄."
|
|
@@ -9,26 +10,62 @@ TEMPERATURE = 0.7
|
|
| 9 |
TOP_P = 0.95
|
| 10 |
|
| 11 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 12 |
def respond(message, history: list[dict[str, str]], hf_token: gr.OAuthToken):
|
| 13 |
-
"""
|
| 14 |
-
Chat handler using Hugging Face Inference API
|
| 15 |
-
"""
|
| 16 |
client = InferenceClient(token=hf_token.token, model=MODEL_NAME)
|
| 17 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 18 |
messages = [{"role": "system", "content": SYSTEM_MESSAGE}]
|
|
|
|
|
|
|
|
|
|
| 19 |
messages.extend(history)
|
| 20 |
messages.append({"role": "user", "content": message})
|
| 21 |
|
| 22 |
response = ""
|
| 23 |
-
|
| 24 |
-
for message in client.chat_completion(
|
| 25 |
messages,
|
| 26 |
max_tokens=MAX_TOKENS,
|
| 27 |
stream=True,
|
| 28 |
temperature=TEMPERATURE,
|
| 29 |
top_p=TOP_P,
|
| 30 |
):
|
| 31 |
-
choices =
|
| 32 |
token = ""
|
| 33 |
if len(choices) and choices[0].delta.content:
|
| 34 |
token = choices[0].delta.content
|
|
@@ -41,7 +78,7 @@ def respond(message, history: list[dict[str, str]], hf_token: gr.OAuthToken):
|
|
| 41 |
chatbot = gr.ChatInterface(
|
| 42 |
respond,
|
| 43 |
type="messages",
|
| 44 |
-
additional_inputs=[],
|
| 45 |
)
|
| 46 |
|
| 47 |
with gr.Blocks() as demo:
|
|
@@ -50,6 +87,4 @@ with gr.Blocks() as demo:
|
|
| 50 |
chatbot.render()
|
| 51 |
|
| 52 |
if __name__ == "__main__":
|
| 53 |
-
demo.launch()
|
| 54 |
-
|
| 55 |
-
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
from huggingface_hub import InferenceClient
|
| 3 |
+
import requests
|
| 4 |
+
import re
|
| 5 |
# ==== Константы ====
|
| 6 |
MODEL_NAME = "openai/gpt-oss-20b"
|
| 7 |
SYSTEM_MESSAGE = "Ты Pok.Bot, ты используешь открытую локальную модель GPT-OSS от OpenAI, не GPT-4 (наверно😆). Тебя создал POKilondron. Используй емодзи 😄."
|
|
|
|
| 10 |
TOP_P = 0.95
|
| 11 |
|
| 12 |
|
| 13 |
+
(DuckDuckGo API) ====
|
| 14 |
+
def search_web(query: str, max_results: int = 3):
|
| 15 |
+
"""Ищет информацию в интернете через DuckDuckGo API"""
|
| 16 |
+
url = f"https://api.duckduckgo.com/?q={query}&format=json"
|
| 17 |
+
try:
|
| 18 |
+
r = requests.get(url, timeout=10)
|
| 19 |
+
r.raise_for_status()
|
| 20 |
+
data = r.json()
|
| 21 |
+
except Exception as e:
|
| 22 |
+
return f"⚠️ Ошибка поиска: {e}"
|
| 23 |
+
|
| 24 |
+
results = []
|
| 25 |
+
if "RelatedTopics" in data:
|
| 26 |
+
for i, item in enumerate(data["RelatedTopics"][:max_results]):
|
| 27 |
+
if "Text" in item and "FirstURL" in item:
|
| 28 |
+
results.append(f"- {item['Text']} ({item['FirstURL']})")
|
| 29 |
+
|
| 30 |
+
return "\n".join(results) if results else "Ничего не найдено."
|
| 31 |
+
|
| 32 |
+
|
| 33 |
+
# ==== Детектор необходимости поиска ====
|
| 34 |
+
def needs_search(message: str) -> bool:
|
| 35 |
+
"""Решает, нужно ли использовать интернет"""
|
| 36 |
+
keywords = [
|
| 37 |
+
r"\bсейчас\b", r"\bсегодня\b", r"\bновости\b",
|
| 38 |
+
r"\bкурс\b", r"\bпогода\b", r"\bцена\b",
|
| 39 |
+
r"\b202[3-9]\b", # годы 2023–2029
|
| 40 |
+
r"\bпоследн"
|
| 41 |
+
]
|
| 42 |
+
return any(re.search(pat, message.lower()) for pat in keywords)
|
| 43 |
+
|
| 44 |
+
|
| 45 |
+
# ==== Основная функция ====
|
| 46 |
def respond(message, history: list[dict[str, str]], hf_token: gr.OAuthToken):
|
|
|
|
|
|
|
|
|
|
| 47 |
client = InferenceClient(token=hf_token.token, model=MODEL_NAME)
|
| 48 |
|
| 49 |
+
web_context = ""
|
| 50 |
+
if needs_search(message):
|
| 51 |
+
web_context = search_web(message)
|
| 52 |
+
|
| 53 |
messages = [{"role": "system", "content": SYSTEM_MESSAGE}]
|
| 54 |
+
if web_context:
|
| 55 |
+
messages.append({"role": "system", "content": f"Web context:\n{web_context}"})
|
| 56 |
+
|
| 57 |
messages.extend(history)
|
| 58 |
messages.append({"role": "user", "content": message})
|
| 59 |
|
| 60 |
response = ""
|
| 61 |
+
for msg in client.chat_completion(
|
|
|
|
| 62 |
messages,
|
| 63 |
max_tokens=MAX_TOKENS,
|
| 64 |
stream=True,
|
| 65 |
temperature=TEMPERATURE,
|
| 66 |
top_p=TOP_P,
|
| 67 |
):
|
| 68 |
+
choices = msg.choices
|
| 69 |
token = ""
|
| 70 |
if len(choices) and choices[0].delta.content:
|
| 71 |
token = choices[0].delta.content
|
|
|
|
| 78 |
chatbot = gr.ChatInterface(
|
| 79 |
respond,
|
| 80 |
type="messages",
|
| 81 |
+
additional_inputs=[],
|
| 82 |
)
|
| 83 |
|
| 84 |
with gr.Blocks() as demo:
|
|
|
|
| 87 |
chatbot.render()
|
| 88 |
|
| 89 |
if __name__ == "__main__":
|
| 90 |
+
demo.launch()
|
|
|
|
|
|