Create app_v2.py

app_v2.py

@@ -0,0 +1,185 @@
+import streamlit as st
+from io import BytesIO
+import ibm_watsonx_ai
+import secretsload
+import genparam
+import requests
+import time
+import re
+from ibm_watsonx_ai.foundation_models import ModelInference
+from ibm_watsonx_ai import Credentials, APIClient
+from ibm_watsonx_ai.metanames import GenTextParamsMetaNames as GenParams
+from ibm_watsonx_ai.metanames import GenTextReturnOptMetaNames as RetParams
+from secretsload import load_stsecrets
+
+credentials = load_stsecrets()
+
+st.set_page_config(
+    page_title="Jimmy",
+    page_icon="😒",
+    initial_sidebar_state="collapsed"
+)
+
+# Password protection
+def check_password():
+    def password_entered():
+        if st.session_state["password"] == st.secrets["app_password"]:
+            st.session_state["password_correct"] = True
+            del st.session_state["password"]
+        else:
+            st.session_state["password_correct"] = False
+
+    if "password_correct" not in st.session_state:
+        st.markdown("\n\n")
+        st.text_input("Enter the password", type="password", on_change=password_entered, key="password")
+        st.divider()
+        st.info("Developed by Milan Mrdenovic © IBM Norway 2025")
+        return False
+    elif not st.session_state["password_correct"]:
+        st.markdown("\n\n")
+        st.text_input("Enter the password", type="password", on_change=password_entered, key="password")
+        st.divider()
+        st.info("Developed by Milan Mrdenovic © IBM Norway 2025")
+        st.error("😕 Password incorrect")
+        return False
+    else:
+        return True
+
+if not check_password():
+    st.stop()
+
+
+# Initialize session state
+if 'current_page' not in st.session_state:
+    st.session_state.current_page = 0
+
+def initialize_session_state():
+    if 'chat_history' not in st.session_state:
+        st.session_state.chat_history = []
+
+def setup_client():
+    credentials = Credentials(
+        url=st.secrets["url"],
+        api_key=st.secrets["api_key"]
+    )
+    return APIClient(credentials, project_id=st.secrets["project_id"])
+
+def prepare_prompt(prompt, chat_history):
+    if genparam.TYPE == "chat" and chat_history:
+        chats = "\n".join([f"{message['role']}: \"{message['content']}\"" for message in chat_history])
+        return f"Conversation History:\n{chats}\n\n New User Prompt: {prompt}"
+    return prompt
+
+def apply_prompt_syntax(prompt, system_prompt, prompt_template, bake_in_prompt_syntax):
+    model_family_syntax = {
+        ### Llama Models
+        "llama3_1_instruct - system": """<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\n{system_prompt}<|eot_id|><|start_header_id|>user<|end_header_id|>\n\n{prompt}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n""",
+        "llama3_3_instruct - system": """<|start_header_id|>system<|end_header_id|>\n\n{system_prompt}<|eot_id|><|start_header_id|>user<|end_header_id|>\n\n{prompt}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n""",
+        
+        "llama3_1_instruct - user": """<|begin_of_text|><|start_header_id|>user<|end_header_id|>\n\n{prompt}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n""",
+        "llama3_3_instruct - user": """<|start_header_id|>user<|end_header_id|>\n\n{prompt}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n""",
+        ### Granite Models
+        "granite_3 - system": """<|start_of_role|>system<|end_of_role|>{system_prompt}<|end_of_text|>\n<|start_of_role|>user<|end_of_role|>{prompt}<|end_of_text|>\n<|start_of_role|>assistant<|end_of_role|>""",
+        "granite_3 - user": """<|start_of_role|>user<|end_of_role|>{prompt}<|end_of_text|>\n<|start_of_role|>assistant<|end_of_role|>""",
+        ### Granite Code Only
+        "granite_code - with system": """System:\n{system_prompt}\n\nQuestion:\n{prompt}\n\nAnswer:\n""",
+        "granite_code - instruction only": """Question:\n{prompt}\n\nAnswer:\n""",
+        ### Mistral Models
+        "mistral_large - sys": """[INST] {system_prompt}\n\n{prompt}[/INST]""", ### mistral-large-2407
+        "mistral_large - user": """[INST] {prompt}[/INST]""",                   ### mistral-large-2407
+
+        "mistral_large_2411 - sys": """[SYSTEM_PROMPT] {system_prompt}[/SYSTEM_PROMPT][INST] {prompt}[/INST]""", ### Only deploy on demand on watsonx.ai
+        
+        "mistral_ai_small - sys": """[INST] {system_prompt}\n\n{prompt}[/INST]""", ### mistral-small-24b-2507 seems to have tokenization issues
+        ### No Syntax
+        "no syntax - system": """{system_prompt}\n\n{prompt}""",
+        "no syntax - user": """{prompt}""",
+    }
+
+
+    
+    # model_family_syntax = {
+    #     ### Llama Models
+    #     "llama3_1_instruct - system": """<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\n{system_prompt}<|eot_id|><|start_header_id|>user<|end_header_id|>\n\n{prompt}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n""",
+    #     "llama3_3_instruct - system": """<|start_header_id|>system<|end_header_id|>\n\n{system_prompt}<|eot_id|><|start_header_id|>user<|end_header_id|>\n\n{prompt}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n""",
+    #     "llama3_instruct - user": """<|begin_of_text|><|start_header_id|>user<|end_header_id|>\n\n{prompt}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n""",
+    #     ### Granite Models
+    #     "granite_3 - system": """<|start_of_role|>system<|end_of_role|>{system_prompt}<|end_of_text|>\n<|start_of_role|>user<|end_of_role|>{prompt}<|end_of_text|>\n<|start_of_role|>assistant<|end_of_role|>""",
+    #     "granite_3 - user": """<|start_of_role|>user<|end_of_role|>{prompt}<|end_of_text|>\n<|start_of_role|>assistant<|end_of_role|>""",
+    #     ### Granite Code Only
+    #     "granite_code - with system": """System:\n{system_prompt}\n\nQuestion:\n{prompt}\n\nAnswer:\n""",
+    #     "granite_code - instruction only": """Question:\n{prompt}\n\nAnswer:\n""",
+    #     ### Mistral Models
+    #     "mistral_large_sys": """[INST]{system_prompt} \n\n {prompt}[/INST]""",
+    #     "mistral_large": """[INST]{prompt}[/INST]""",
+        
+    #     "mistral_ai_small_sys": """<s>[SYSTEM_PROMPT]{system_prompt}[/SYSTEM_PROMPT][INST]{prompt}[/INST]""",
+    #     "mistral_ai_small_sys": """<s>[SYSTEM_PROMPT]{system_prompt}[/SYSTEM_PROMPT][INST]{prompt}[/INST]""",
+    #     ### No Syntax
+    #     "no syntax - system": """{system_prompt}\n\n{prompt}""",
+    #     "no syntax - user": """{prompt}""",
+    # }
+    
+    if bake_in_prompt_syntax:
+        template = model_family_syntax[prompt_template]
+        if system_prompt:
+            return template.format(system_prompt=system_prompt, prompt=prompt)
+    return prompt
+
+def generate_response(watsonx_llm, prompt_data, params):
+    generated_response = watsonx_llm.generate_text_stream(prompt=prompt_data, params=params)
+    for chunk in generated_response:
+        yield chunk
+
+def chat_interface():
+    st.subheader("Jimmy")
+
+    # User input
+    user_input = st.chat_input("You:", key="user_input")
+
+    if user_input:
+        # Add user message to chat history
+        st.session_state.chat_history.append({"role": "user", "content": user_input})
+
+        # Prepare the prompt
+        prompt = prepare_prompt(user_input, st.session_state.chat_history)
+
+        # Apply prompt syntax
+        prompt_data = apply_prompt_syntax(
+            prompt, 
+            genparam.SYSTEM_PROMPT,
+            genparam.PROMPT_TEMPLATE,
+            genparam.BAKE_IN_PROMPT_SYNTAX
+        )
+    
+        # Setup client and model
+        client = setup_client()
+        watsonx_llm = ModelInference(
+            api_client=client, 
+            model_id=genparam.SELECTED_MODEL,
+            verify=genparam.VERIFY
+        )
+    
+        # Prepare parameters
+        params = {
+            GenParams.DECODING_METHOD: genparam.DECODING_METHOD,
+            GenParams.MAX_NEW_TOKENS: genparam.MAX_NEW_TOKENS,
+            GenParams.MIN_NEW_TOKENS: genparam.MIN_NEW_TOKENS,
+            GenParams.REPETITION_PENALTY: genparam.REPETITION_PENALTY,
+            GenParams.STOP_SEQUENCES: genparam.STOP_SEQUENCES
+        }
+
+       # Generate and stream response
+        with st.chat_message("Jimmy", avatar="😒"):
+            stream = generate_response(watsonx_llm, prompt_data, params)
+            response = st.write_stream(stream)
+
+        # Add AI response to chat history
+        st.session_state.chat_history.append({"role": "Jimmy", "content": response})
+
+def main():
+    initialize_session_state()
+    chat_interface()
+
+if __name__ == "__main__":
+    main()