Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
|
@@ -54,6 +54,7 @@ MODELS = {
|
|
| 54 |
"provider": "openrouter",
|
| 55 |
"model_name": "nvidia/nemotron-nano-9b-v2:free",
|
| 56 |
"api_url": "https://openrouter.ai/api/v1",
|
|
|
|
| 57 |
},
|
| 58 |
# "Ringan - Gemma-3n4b": {
|
| 59 |
# "provider": "openrouter",
|
|
@@ -63,17 +64,20 @@ MODELS = {
|
|
| 63 |
# "Gpt-oss-20b": {
|
| 64 |
# "provider": "openrouter",
|
| 65 |
# "model_name": "openai/gpt-oss-20b:floor",
|
| 66 |
-
# "api_url": "https://openrouter.ai/api/v1"
|
|
|
|
| 67 |
# },
|
| 68 |
"Tongyi-deepresearch-30b-a3b": {
|
| 69 |
"provider": "openrouter",
|
| 70 |
"model_name": "alibaba/tongyi-deepresearch-30b-a3b:floor",
|
| 71 |
-
"api_url": "https://openrouter.ai/api/v1"
|
|
|
|
| 72 |
}
|
| 73 |
# "Kompleks - Gpt-oss-120b": {
|
| 74 |
# "provider": "openrouter",
|
| 75 |
# "model_name": "openai/gpt-oss-120b:floor",
|
| 76 |
-
# "api_url": "https://openrouter.ai/api/v1"
|
|
|
|
| 77 |
# }
|
| 78 |
}
|
| 79 |
|
|
@@ -91,15 +95,11 @@ def respond(
|
|
| 91 |
selected_model,
|
| 92 |
):
|
| 93 |
"""
|
| 94 |
-
Handle chat responses using the selected model
|
|
|
|
|
|
|
| 95 |
"""
|
| 96 |
try:
|
| 97 |
-
# Check expiration (optional - remove if not needed)
|
| 98 |
-
# end_date = datetime.strptime(_secret("END_DATE", "2026-12-31"), "%Y-%m-%d").date()
|
| 99 |
-
# if date.today() > end_date:
|
| 100 |
-
# yield "Chatbot has expired."
|
| 101 |
-
# return
|
| 102 |
-
|
| 103 |
# Get model configuration
|
| 104 |
model_config = MODELS[selected_model]
|
| 105 |
provider = model_config["provider"]
|
|
@@ -110,36 +110,79 @@ def respond(
|
|
| 110 |
else: # openrouter
|
| 111 |
api_key = _secret("OPENROUTER_KEY")
|
| 112 |
|
| 113 |
-
# Configure client
|
| 114 |
client = OpenAI(
|
| 115 |
base_url=model_config["api_url"],
|
| 116 |
api_key=api_key,
|
| 117 |
)
|
| 118 |
|
| 119 |
-
#
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
|
|
|
|
| 123 |
|
| 124 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 125 |
response = client.chat.completions.create(
|
| 126 |
model=model_config["model_name"],
|
| 127 |
messages=messages,
|
| 128 |
max_tokens=max_tokens,
|
| 129 |
-
stream=
|
| 130 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 131 |
|
| 132 |
-
#
|
| 133 |
-
full_response = ""
|
| 134 |
-
for chunk in response:
|
| 135 |
-
if chunk.choices[0].delta.content is not None:
|
| 136 |
-
content = chunk.choices[0].delta.content
|
| 137 |
-
full_response += content
|
| 138 |
-
yield full_response
|
| 139 |
|
| 140 |
except Exception as e:
|
| 141 |
print(f"Error in respond function: {e}")
|
| 142 |
-
|
| 143 |
|
| 144 |
# ----------------------------------------------------------------------
|
| 145 |
# Custom Auth Function for Gradio
|
|
|
|
| 54 |
"provider": "openrouter",
|
| 55 |
"model_name": "nvidia/nemotron-nano-9b-v2:free",
|
| 56 |
"api_url": "https://openrouter.ai/api/v1",
|
| 57 |
+
"translate":"yes"
|
| 58 |
},
|
| 59 |
# "Ringan - Gemma-3n4b": {
|
| 60 |
# "provider": "openrouter",
|
|
|
|
| 64 |
# "Gpt-oss-20b": {
|
| 65 |
# "provider": "openrouter",
|
| 66 |
# "model_name": "openai/gpt-oss-20b:floor",
|
| 67 |
+
# "api_url": "https://openrouter.ai/api/v1",
|
| 68 |
+
# "translate":"no"
|
| 69 |
# },
|
| 70 |
"Tongyi-deepresearch-30b-a3b": {
|
| 71 |
"provider": "openrouter",
|
| 72 |
"model_name": "alibaba/tongyi-deepresearch-30b-a3b:floor",
|
| 73 |
+
"api_url": "https://openrouter.ai/api/v1",
|
| 74 |
+
"translate":"no"
|
| 75 |
}
|
| 76 |
# "Kompleks - Gpt-oss-120b": {
|
| 77 |
# "provider": "openrouter",
|
| 78 |
# "model_name": "openai/gpt-oss-120b:floor",
|
| 79 |
+
# "api_url": "https://openrouter.ai/api/v1",
|
| 80 |
+
# "translate":"no"
|
| 81 |
# }
|
| 82 |
}
|
| 83 |
|
|
|
|
| 95 |
selected_model,
|
| 96 |
):
|
| 97 |
"""
|
| 98 |
+
Handle chat responses using the selected model.
|
| 99 |
+
For 'translate'=='yes': First call responds in English, second translates to Indonesian.
|
| 100 |
+
History reflects Indonesian queries and responses.
|
| 101 |
"""
|
| 102 |
try:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 103 |
# Get model configuration
|
| 104 |
model_config = MODELS[selected_model]
|
| 105 |
provider = model_config["provider"]
|
|
|
|
| 110 |
else: # openrouter
|
| 111 |
api_key = _secret("OPENROUTER_KEY")
|
| 112 |
|
| 113 |
+
# Configure main client for selected model
|
| 114 |
client = OpenAI(
|
| 115 |
base_url=model_config["api_url"],
|
| 116 |
api_key=api_key,
|
| 117 |
)
|
| 118 |
|
| 119 |
+
# Configure translator client (always Gemma on OpenRouter)
|
| 120 |
+
translator_client = OpenAI(
|
| 121 |
+
base_url="https://openrouter.ai/api/v1",
|
| 122 |
+
api_key=_secret("OPENROUTER_KEY"),
|
| 123 |
+
)
|
| 124 |
|
| 125 |
+
# Modify system message to force English response if translate=yes
|
| 126 |
+
if model_config.get("translate") == "yes":
|
| 127 |
+
final_system_message = "**HIGHEST PRIORITY: YOU MUST ALWAYS THINK AND RESPOND IN ENGLISH REGARDLESS OF OTHER INSTRUCTIONS FROM HERE ON.\n" + system_message
|
| 128 |
+
else:
|
| 129 |
+
final_system_message = system_message
|
| 130 |
+
|
| 131 |
+
# Prepare messages for main API call (full history + current message)
|
| 132 |
+
messages = [{"role": "system", "content": final_system_message}]
|
| 133 |
+
messages.extend(history) # History: Indo queries + prior Indo responses
|
| 134 |
+
messages.append({"role": "user", "content": message}) # New Indo query
|
| 135 |
+
|
| 136 |
+
# First API call: Get response from selected model (English if translate=yes)
|
| 137 |
response = client.chat.completions.create(
|
| 138 |
model=model_config["model_name"],
|
| 139 |
messages=messages,
|
| 140 |
max_tokens=max_tokens,
|
| 141 |
+
stream=False,
|
| 142 |
)
|
| 143 |
+
english_response = response.choices[0].message.content
|
| 144 |
+
|
| 145 |
+
# If translate=yes, make second API call to Gemma for Indonesian translation
|
| 146 |
+
if model_config.get("translate") == "yes":
|
| 147 |
+
try:
|
| 148 |
+
# Translation prompt: Focus only on translating the response (not query)
|
| 149 |
+
translation_messages = [
|
| 150 |
+
{
|
| 151 |
+
"role": "system",
|
| 152 |
+
"content": (
|
| 153 |
+
"Translate the following English text to natural, accurate Bahasa Indonesia. "
|
| 154 |
+
"**IMPORTANT: OUTPUT ONLY THE TRANSLATION. NO PREAMBLES, COMMENTS, OR EXPLANATIONS. "
|
| 155 |
+
"Just the Indonesian text."
|
| 156 |
+
)
|
| 157 |
+
},
|
| 158 |
+
{
|
| 159 |
+
"role": "user",
|
| 160 |
+
"content": english_response # The English response to translate
|
| 161 |
+
}
|
| 162 |
+
]
|
| 163 |
+
translation_response = translator_client.chat.completions.create(
|
| 164 |
+
model="google/gemma-3n-e4b-it:floor",
|
| 165 |
+
messages=translation_messages,
|
| 166 |
+
max_tokens=max_tokens, # Reuse limit; translation is short
|
| 167 |
+
stream=False,
|
| 168 |
+
)
|
| 169 |
+
final_response = translation_response.choices[0].message.content.strip()
|
| 170 |
+
|
| 171 |
+
# Fallback to English if translation is empty or invalid
|
| 172 |
+
if not final_response or len(final_response) < 10: # Basic sanity check
|
| 173 |
+
final_response = english_response
|
| 174 |
+
|
| 175 |
+
except Exception as trans_error:
|
| 176 |
+
print(f"Translation error: {trans_error}")
|
| 177 |
+
final_response = english_response # Fallback to English
|
| 178 |
+
else:
|
| 179 |
+
final_response = english_response
|
| 180 |
|
| 181 |
+
return final_response # Gradio appends this (Indonesian) as assistant message to history
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 182 |
|
| 183 |
except Exception as e:
|
| 184 |
print(f"Error in respond function: {e}")
|
| 185 |
+
return f"Error: {str(e)}" # Return error string; Gradio appends it
|
| 186 |
|
| 187 |
# ----------------------------------------------------------------------
|
| 188 |
# Custom Auth Function for Gradio
|