Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -9,6 +9,7 @@ import json
|
|
9 |
_ = load_dotenv(find_dotenv())
|
10 |
st.set_page_config(page_icon="💬", layout="wide", page_title="Groq Chat Bot...")
|
11 |
|
|
|
12 |
def icon(emoji: str):
|
13 |
"""Shows an emoji as a Notion-style page icon."""
|
14 |
st.write(
|
@@ -16,6 +17,7 @@ def icon(emoji: str):
|
|
16 |
unsafe_allow_html=True,
|
17 |
)
|
18 |
|
|
|
19 |
icon("📣")
|
20 |
|
21 |
st.subheader("Groq Chat Streamlit App", divider="rainbow", anchor=False)
|
@@ -24,6 +26,73 @@ client = Groq(
|
|
24 |
api_key=os.environ['GROQ_API_KEY'],
|
25 |
)
|
26 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
27 |
models = {
|
28 |
"mixtral-8x7b-32768": {
|
29 |
"name": "Mixtral-8x7b-Instruct-v0.1",
|
@@ -34,6 +103,7 @@ models = {
|
|
34 |
"gemma-7b-it": {"name": "Gemma-7b-it", "tokens": 8192, "developer": "Google"},
|
35 |
}
|
36 |
|
|
|
37 |
col1, col2 = st.columns(2)
|
38 |
|
39 |
with col1:
|
@@ -41,15 +111,10 @@ with col1:
|
|
41 |
"Choose a model:",
|
42 |
options=list(models.keys()),
|
43 |
format_func=lambda x: models[x]["name"],
|
44 |
-
index=0,
|
45 |
)
|
46 |
|
47 |
-
if
|
48 |
-
st.session_state.messages = []
|
49 |
-
|
50 |
-
if "selected_model" not in st.session_state:
|
51 |
-
st.session_state.selected_model = None
|
52 |
-
|
53 |
if st.session_state.selected_model != model_option:
|
54 |
st.session_state.messages = []
|
55 |
st.session_state.selected_model = model_option
|
@@ -57,110 +122,55 @@ if st.session_state.selected_model != model_option:
|
|
57 |
max_tokens_range = models[model_option]["tokens"]
|
58 |
|
59 |
with col2:
|
|
|
60 |
max_tokens = st.slider(
|
61 |
"Max Tokens:",
|
62 |
-
min_value=512,
|
63 |
max_value=max_tokens_range,
|
|
|
64 |
value=min(32768, max_tokens_range),
|
65 |
step=512,
|
66 |
help=f"Adjust the maximum number of tokens (words) for the model's response. Max for selected model: {max_tokens_range}",
|
67 |
)
|
68 |
|
|
|
69 |
for message in st.session_state.messages:
|
70 |
avatar = "🤖" if message["role"] == "assistant" else "🕺"
|
71 |
with st.chat_message(message["role"], avatar=avatar):
|
72 |
st.markdown(message["content"])
|
73 |
|
74 |
-
def generate_chat_responses(user_prompt):
|
75 |
-
"""Fetches response from the Groq API using the run_conversation function."""
|
76 |
-
response = run_conversation(user_prompt)
|
77 |
-
yield response # Yield the response content
|
78 |
-
|
79 |
-
def run_conversation(user_prompt):
|
80 |
-
messages=[
|
81 |
-
{
|
82 |
-
"role": "system",
|
83 |
-
"content": "You are a helpful assistant named ChattyBot."
|
84 |
-
},
|
85 |
-
{
|
86 |
-
"role": "user",
|
87 |
-
"content": user_prompt,
|
88 |
-
}
|
89 |
-
]
|
90 |
-
tools = [
|
91 |
-
{
|
92 |
-
"type": "function",
|
93 |
-
"function": {
|
94 |
-
"name": "time_date",
|
95 |
-
"description": "The tool will return information about the time and date to the AI.",
|
96 |
-
"parameters": {},
|
97 |
-
},
|
98 |
-
}
|
99 |
-
]
|
100 |
-
try:
|
101 |
-
response = client.chat.completions.create(
|
102 |
-
model=model_option,
|
103 |
-
messages=messages,
|
104 |
-
tools=tools,
|
105 |
-
tool_choice="auto",
|
106 |
-
max_tokens=4096
|
107 |
-
)
|
108 |
-
|
109 |
-
response_message = response.choices[0].delta
|
110 |
-
tool_calls = response_message.tool_calls
|
111 |
-
|
112 |
-
if tool_calls:
|
113 |
-
available_functions = {
|
114 |
-
"time_date": get_tool_owner_info
|
115 |
-
}
|
116 |
-
|
117 |
-
messages.append(response_message)
|
118 |
-
|
119 |
-
for tool_call in tool_calls:
|
120 |
-
function_name = tool_call.function.name
|
121 |
-
function_to_call = available_functions[function_name]
|
122 |
-
function_args = json.loads(tool_call.function.arguments)
|
123 |
-
function_response = function_to_call(**function_args)
|
124 |
-
messages.append(
|
125 |
-
{
|
126 |
-
"tool_call_id": tool_call.id,
|
127 |
-
"role": "tool",
|
128 |
-
"name": function_name,
|
129 |
-
"content": function_response,
|
130 |
-
}
|
131 |
-
)
|
132 |
-
|
133 |
-
second_response = client.chat.completions.create(
|
134 |
-
model=model_option,
|
135 |
-
messages=messages
|
136 |
-
)
|
137 |
|
138 |
-
|
139 |
-
|
140 |
-
|
141 |
-
|
142 |
-
|
143 |
-
return None
|
144 |
|
145 |
-
def get_tool_owner_info():
|
146 |
-
owner_info = {
|
147 |
-
"date_time": datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")
|
148 |
-
}
|
149 |
-
return json.dumps(owner_info)
|
150 |
|
151 |
if prompt := st.chat_input("Enter your prompt here..."):
|
152 |
st.session_state.messages.append({"role": "user", "content": prompt})
|
153 |
|
154 |
-
with st.chat_message("user", avatar=""):
|
155 |
st.markdown(prompt)
|
156 |
|
|
|
157 |
try:
|
158 |
-
|
159 |
-
|
160 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
161 |
full_response = st.write_stream(chat_responses_generator)
|
162 |
except Exception as e:
|
163 |
-
st.error(e, icon="")
|
164 |
|
165 |
# Append the full response to session_state.messages
|
166 |
if isinstance(full_response, str):
|
@@ -172,4 +182,5 @@ if prompt := st.chat_input("Enter your prompt here..."):
|
|
172 |
combined_response = "\n".join(str(item) for item in full_response)
|
173 |
st.session_state.messages.append(
|
174 |
{"role": "assistant", "content": combined_response}
|
175 |
-
)
|
|
|
|
9 |
_ = load_dotenv(find_dotenv())
|
10 |
st.set_page_config(page_icon="💬", layout="wide", page_title="Groq Chat Bot...")
|
11 |
|
12 |
+
|
13 |
def icon(emoji: str):
|
14 |
"""Shows an emoji as a Notion-style page icon."""
|
15 |
st.write(
|
|
|
17 |
unsafe_allow_html=True,
|
18 |
)
|
19 |
|
20 |
+
|
21 |
icon("📣")
|
22 |
|
23 |
st.subheader("Groq Chat Streamlit App", divider="rainbow", anchor=False)
|
|
|
26 |
api_key=os.environ['GROQ_API_KEY'],
|
27 |
)
|
28 |
|
29 |
+
def get_tool_owner_info():
|
30 |
+
owner_info = {
|
31 |
+
"date_time": datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")
|
32 |
+
}
|
33 |
+
return json.dumps(owner_info)
|
34 |
+
|
35 |
+
def run_conversation(user_prompt, messages):
|
36 |
+
tools = [
|
37 |
+
{
|
38 |
+
"type": "function",
|
39 |
+
"function": {
|
40 |
+
"name": "time_date",
|
41 |
+
"description": "The tool will return information about the time and date to the AI.",
|
42 |
+
"parameters": {},
|
43 |
+
},
|
44 |
+
}
|
45 |
+
]
|
46 |
+
response = client.chat.completions.create(
|
47 |
+
model=model_option,
|
48 |
+
messages=messages,
|
49 |
+
tools=tools,
|
50 |
+
tool_choice="auto",
|
51 |
+
max_tokens=max_tokens
|
52 |
+
)
|
53 |
+
|
54 |
+
response_message = response.choices[0].message
|
55 |
+
tool_calls = response_message.tool_calls
|
56 |
+
|
57 |
+
if tool_calls:
|
58 |
+
available_functions = {
|
59 |
+
"time_date": get_tool_owner_info
|
60 |
+
}
|
61 |
+
|
62 |
+
messages.append(response_message)
|
63 |
+
|
64 |
+
for tool_call in tool_calls:
|
65 |
+
function_name = tool_call.function.name
|
66 |
+
function_to_call = available_functions[function_name]
|
67 |
+
function_args = json.loads(tool_call.function.arguments)
|
68 |
+
function_response = function_to_call(**function_args)
|
69 |
+
messages.append(
|
70 |
+
{
|
71 |
+
"tool_call_id": tool_call.id,
|
72 |
+
"role": "tool",
|
73 |
+
"name": function_name,
|
74 |
+
"content": function_response,
|
75 |
+
}
|
76 |
+
)
|
77 |
+
|
78 |
+
second_response = client.chat.completions.create(
|
79 |
+
model=model_option,
|
80 |
+
messages=messages
|
81 |
+
)
|
82 |
+
|
83 |
+
return second_response.choices[0].message.content
|
84 |
+
else:
|
85 |
+
return response_message
|
86 |
+
|
87 |
+
|
88 |
+
# Initialize chat history and selected model
|
89 |
+
if "messages" not in st.session_state:
|
90 |
+
st.session_state.messages = []
|
91 |
+
|
92 |
+
if "selected_model" not in st.session_state:
|
93 |
+
st.session_state.selected_model = None
|
94 |
+
|
95 |
+
# Define model details
|
96 |
models = {
|
97 |
"mixtral-8x7b-32768": {
|
98 |
"name": "Mixtral-8x7b-Instruct-v0.1",
|
|
|
103 |
"gemma-7b-it": {"name": "Gemma-7b-it", "tokens": 8192, "developer": "Google"},
|
104 |
}
|
105 |
|
106 |
+
# Layout for model selection and max_tokens slider
|
107 |
col1, col2 = st.columns(2)
|
108 |
|
109 |
with col1:
|
|
|
111 |
"Choose a model:",
|
112 |
options=list(models.keys()),
|
113 |
format_func=lambda x: models[x]["name"],
|
114 |
+
index=0, # Default to the first model in the list
|
115 |
)
|
116 |
|
117 |
+
# Detect model change and clear chat history if model has changed
|
|
|
|
|
|
|
|
|
|
|
118 |
if st.session_state.selected_model != model_option:
|
119 |
st.session_state.messages = []
|
120 |
st.session_state.selected_model = model_option
|
|
|
122 |
max_tokens_range = models[model_option]["tokens"]
|
123 |
|
124 |
with col2:
|
125 |
+
# Adjust max_tokens slider dynamically based on the selected model
|
126 |
max_tokens = st.slider(
|
127 |
"Max Tokens:",
|
128 |
+
min_value=512, # Minimum value to allow some flexibility
|
129 |
max_value=max_tokens_range,
|
130 |
+
# Default value or max allowed if less
|
131 |
value=min(32768, max_tokens_range),
|
132 |
step=512,
|
133 |
help=f"Adjust the maximum number of tokens (words) for the model's response. Max for selected model: {max_tokens_range}",
|
134 |
)
|
135 |
|
136 |
+
# Display chat messages from history on app rerun
|
137 |
for message in st.session_state.messages:
|
138 |
avatar = "🤖" if message["role"] == "assistant" else "🕺"
|
139 |
with st.chat_message(message["role"], avatar=avatar):
|
140 |
st.markdown(message["content"])
|
141 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
142 |
|
143 |
+
def generate_chat_responses(chat_completion) -> Generator[str, None, None]:
|
144 |
+
"""Yield chat response content from the Groq API response."""
|
145 |
+
for chunk in chat_completion:
|
146 |
+
if chunk.choices[0].delta.content:
|
147 |
+
yield chunk.choices[0].delta.content
|
|
|
148 |
|
|
|
|
|
|
|
|
|
|
|
149 |
|
150 |
if prompt := st.chat_input("Enter your prompt here..."):
|
151 |
st.session_state.messages.append({"role": "user", "content": prompt})
|
152 |
|
153 |
+
with st.chat_message("user", avatar="🕺"):
|
154 |
st.markdown(prompt)
|
155 |
|
156 |
+
# Fetch response from Groq API
|
157 |
try:
|
158 |
+
chat_completion = client.chat.completions.create(
|
159 |
+
model=model_option,
|
160 |
+
messages=[
|
161 |
+
{"role": m["role"], "content": m["content"]}
|
162 |
+
for m in st.session_state.messages
|
163 |
+
],
|
164 |
+
max_tokens=max_tokens,
|
165 |
+
stream=True,
|
166 |
+
)
|
167 |
+
|
168 |
+
# Use the generator function with st.write_stream
|
169 |
+
with st.chat_message("assistant", avatar="🤖"):
|
170 |
+
chat_responses_generator = generate_chat_responses(chat_completion)
|
171 |
full_response = st.write_stream(chat_responses_generator)
|
172 |
except Exception as e:
|
173 |
+
st.error(e, icon="🚨")
|
174 |
|
175 |
# Append the full response to session_state.messages
|
176 |
if isinstance(full_response, str):
|
|
|
182 |
combined_response = "\n".join(str(item) for item in full_response)
|
183 |
st.session_state.messages.append(
|
184 |
{"role": "assistant", "content": combined_response}
|
185 |
+
)
|
186 |
+
|