Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -3,9 +3,9 @@ import time
|
|
| 3 |
import gradio as gr
|
| 4 |
import pandas as pd
|
| 5 |
import matplotlib.pyplot as plt
|
| 6 |
-
from io import BytesIO
|
| 7 |
import base64
|
| 8 |
-
from
|
|
|
|
| 9 |
|
| 10 |
from core.dashboard import ErohaDashboard
|
| 11 |
from core.intelligence import update_memory, summarize_context
|
|
@@ -19,46 +19,59 @@ from core.model_selector import choose_model
|
|
| 19 |
HF_TOKEN = os.getenv("HF_TOKEN", "")
|
| 20 |
LOG_FILE = "logs/history.csv"
|
| 21 |
os.makedirs("logs", exist_ok=True)
|
| 22 |
-
|
| 23 |
dashboard = ErohaDashboard()
|
| 24 |
|
| 25 |
# ==============================
|
| 26 |
-
#
|
| 27 |
# ==============================
|
| 28 |
-
def
|
| 29 |
"""
|
| 30 |
-
|
| 31 |
"""
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
|
| 43 |
-
|
| 44 |
-
|
| 45 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 46 |
|
| 47 |
# ==============================
|
| 48 |
-
# 🧠
|
| 49 |
# ==============================
|
| 50 |
def generate_response(user_input):
|
| 51 |
try:
|
| 52 |
start = time.time()
|
| 53 |
prefs = analyze_user_input(user_input)
|
| 54 |
model_id = choose_model(user_input)
|
| 55 |
-
client = create_inference_client(model_id)
|
| 56 |
|
| 57 |
-
result =
|
| 58 |
-
user_input,
|
| 59 |
-
max_new_tokens=600,
|
| 60 |
-
temperature=0.7,
|
| 61 |
-
)
|
| 62 |
|
| 63 |
check = evaluate_answer(result)
|
| 64 |
improved = improve_answer(result)
|
|
@@ -82,7 +95,7 @@ def generate_response(user_input):
|
|
| 82 |
summary = (
|
| 83 |
f"🧠 **Модель:** `{model_id}`\n"
|
| 84 |
f"⚡ **Время отклика:** {response_time} сек\n"
|
| 85 |
-
f"🔍
|
| 86 |
f"{'; '.join(check['feedback']) if check['feedback'] else '✅ Всё отлично'}\n\n"
|
| 87 |
f"{context}"
|
| 88 |
)
|
|
@@ -97,11 +110,10 @@ def generate_response(user_input):
|
|
| 97 |
# ==============================
|
| 98 |
def generate_chart():
|
| 99 |
if not os.path.exists(LOG_FILE):
|
| 100 |
-
return "⚠️ Нет данных для
|
| 101 |
df = pd.read_csv(LOG_FILE)
|
| 102 |
if df.empty:
|
| 103 |
return "⚠️ История пуста."
|
| 104 |
-
|
| 105 |
df["time"] = pd.to_datetime(df["time"])
|
| 106 |
df = df.tail(50)
|
| 107 |
|
|
@@ -109,9 +121,8 @@ def generate_chart():
|
|
| 109 |
plt.plot(df["time"], df["response_time"], marker="o", linewidth=1.5)
|
| 110 |
plt.title("⚡ Скорость отклика моделей (последние 50 запросов)")
|
| 111 |
plt.xlabel("Время")
|
| 112 |
-
plt.ylabel("
|
| 113 |
plt.grid(True)
|
| 114 |
-
|
| 115 |
buffer = BytesIO()
|
| 116 |
plt.savefig(buffer, format="png", bbox_inches="tight")
|
| 117 |
buffer.seek(0)
|
|
@@ -125,14 +136,14 @@ def show_dashboard():
|
|
| 125 |
return metrics_text, df, chart_html
|
| 126 |
|
| 127 |
# ==============================
|
| 128 |
-
# 🎨 Gradio
|
| 129 |
# ==============================
|
| 130 |
-
with gr.Blocks(title="Eroha AgentAPI v5.
|
| 131 |
-
gr.Markdown("# 🤖 Eroha AgentAPI v5.
|
| 132 |
-
gr.Markdown("
|
| 133 |
|
| 134 |
with gr.Tab("💬 Agent Chat"):
|
| 135 |
-
user_input = gr.Textbox(label="Введите запрос", placeholder="Например:
|
| 136 |
output_box = gr.Textbox(label="Ответ", lines=15)
|
| 137 |
submit_btn = gr.Button("🚀 Отправить")
|
| 138 |
submit_btn.click(fn=generate_response, inputs=user_input, outputs=output_box)
|
|
|
|
| 3 |
import gradio as gr
|
| 4 |
import pandas as pd
|
| 5 |
import matplotlib.pyplot as plt
|
|
|
|
| 6 |
import base64
|
| 7 |
+
from io import BytesIO
|
| 8 |
+
import requests
|
| 9 |
|
| 10 |
from core.dashboard import ErohaDashboard
|
| 11 |
from core.intelligence import update_memory, summarize_context
|
|
|
|
| 19 |
HF_TOKEN = os.getenv("HF_TOKEN", "")
|
| 20 |
LOG_FILE = "logs/history.csv"
|
| 21 |
os.makedirs("logs", exist_ok=True)
|
|
|
|
| 22 |
dashboard = ErohaDashboard()
|
| 23 |
|
| 24 |
# ==============================
|
| 25 |
+
# 🔁 Автоматический Router с Failover
|
| 26 |
# ==============================
|
| 27 |
+
def query_huggingface_router(prompt, model_id):
|
| 28 |
"""
|
| 29 |
+
Универсальный запрос к Router API с fallback на прямую модель.
|
| 30 |
"""
|
| 31 |
+
headers = {"Authorization": f"Bearer {HF_TOKEN}"}
|
| 32 |
+
payload = {
|
| 33 |
+
"inputs": prompt,
|
| 34 |
+
"parameters": {"max_new_tokens": 600, "temperature": 0.7},
|
| 35 |
+
}
|
| 36 |
+
|
| 37 |
+
try:
|
| 38 |
+
# 1️⃣ Основной Router API
|
| 39 |
+
response = requests.post(
|
| 40 |
+
"https://router.huggingface.co/models/" + model_id,
|
| 41 |
+
headers=headers,
|
| 42 |
+
json=payload,
|
| 43 |
+
timeout=30,
|
| 44 |
+
)
|
| 45 |
+
if response.status_code == 200:
|
| 46 |
+
return response.json()[0]["generated_text"]
|
| 47 |
+
elif response.status_code == 410:
|
| 48 |
+
raise RuntimeError("Old endpoint rejected (410)")
|
| 49 |
+
else:
|
| 50 |
+
raise RuntimeError(f"Router error: {response.status_code}")
|
| 51 |
+
|
| 52 |
+
except Exception as e:
|
| 53 |
+
print("⚠️ Router недоступен, выполняется fallback:", str(e))
|
| 54 |
+
# 2️⃣ Fallback: прямое обращение к конкретной модели
|
| 55 |
+
fallback_url = f"https://huggingface.co/api/models/{model_id}"
|
| 56 |
+
try:
|
| 57 |
+
response = requests.post(fallback_url, headers=headers, json=payload, timeout=30)
|
| 58 |
+
if response.status_code == 200:
|
| 59 |
+
return response.json()[0]["generated_text"]
|
| 60 |
+
else:
|
| 61 |
+
raise RuntimeError(f"Fallback error: {response.status_code}")
|
| 62 |
+
except Exception as e2:
|
| 63 |
+
return f"❌ Ошибка при обращении к API: {str(e2)}"
|
| 64 |
|
| 65 |
# ==============================
|
| 66 |
+
# 🧠 Основная логика
|
| 67 |
# ==============================
|
| 68 |
def generate_response(user_input):
|
| 69 |
try:
|
| 70 |
start = time.time()
|
| 71 |
prefs = analyze_user_input(user_input)
|
| 72 |
model_id = choose_model(user_input)
|
|
|
|
| 73 |
|
| 74 |
+
result = query_huggingface_router(user_input, model_id)
|
|
|
|
|
|
|
|
|
|
|
|
|
| 75 |
|
| 76 |
check = evaluate_answer(result)
|
| 77 |
improved = improve_answer(result)
|
|
|
|
| 95 |
summary = (
|
| 96 |
f"🧠 **Модель:** `{model_id}`\n"
|
| 97 |
f"⚡ **Время отклика:** {response_time} сек\n"
|
| 98 |
+
f"🔍 **Проверка качества:** {check['result']}\n"
|
| 99 |
f"{'; '.join(check['feedback']) if check['feedback'] else '✅ Всё отлично'}\n\n"
|
| 100 |
f"{context}"
|
| 101 |
)
|
|
|
|
| 110 |
# ==============================
|
| 111 |
def generate_chart():
|
| 112 |
if not os.path.exists(LOG_FILE):
|
| 113 |
+
return "⚠️ Нет данных для графика."
|
| 114 |
df = pd.read_csv(LOG_FILE)
|
| 115 |
if df.empty:
|
| 116 |
return "⚠️ История пуста."
|
|
|
|
| 117 |
df["time"] = pd.to_datetime(df["time"])
|
| 118 |
df = df.tail(50)
|
| 119 |
|
|
|
|
| 121 |
plt.plot(df["time"], df["response_time"], marker="o", linewidth=1.5)
|
| 122 |
plt.title("⚡ Скорость отклика моделей (последние 50 запросов)")
|
| 123 |
plt.xlabel("Время")
|
| 124 |
+
plt.ylabel("Секунды")
|
| 125 |
plt.grid(True)
|
|
|
|
| 126 |
buffer = BytesIO()
|
| 127 |
plt.savefig(buffer, format="png", bbox_inches="tight")
|
| 128 |
buffer.seek(0)
|
|
|
|
| 136 |
return metrics_text, df, chart_html
|
| 137 |
|
| 138 |
# ==============================
|
| 139 |
+
# 🎨 Интерфейс Gradio
|
| 140 |
# ==============================
|
| 141 |
+
with gr.Blocks(title="Eroha AgentAPI v5.6 — AutoRouter Failover Core", theme="soft") as app:
|
| 142 |
+
gr.Markdown("# 🤖 Eroha AgentAPI v5.6 — Guru Edition (AutoRouter + Failover Core)")
|
| 143 |
+
gr.Markdown("**Автоматический Router + fallback + визуальная аналитика** ⚙️")
|
| 144 |
|
| 145 |
with gr.Tab("💬 Agent Chat"):
|
| 146 |
+
user_input = gr.Textbox(label="Введите запрос", placeholder="Например: придумай философскую сказку об ИИ.")
|
| 147 |
output_box = gr.Textbox(label="Ответ", lines=15)
|
| 148 |
submit_btn = gr.Button("🚀 Отправить")
|
| 149 |
submit_btn.click(fn=generate_response, inputs=user_input, outputs=output_box)
|