app.py

""" Simple Chatbot
@author: Nigel Gebodh
@email: nigel.gebodh@gmail.com
@website: https://ngebodh.github.io/
"""
import numpy as np
import streamlit as st
from openai import OpenAI
import os
import sys
from dotenv import load_dotenv, dotenv_values
load_dotenv()

# #===========================================
# updates = '''
# Updates
# + 02/06/2026
# - Updated inference endpoints for HF models
# - Added Kimi model 

# + 01/10/2026
# - Updated cooldown

# + 01/08/2026
# - Updated logging info

# + 10/10/2025
# - Update the model options since Gemma-2-9B-it 
#    is no longer supported. Replaced with GPT-OSS-120B

# + 04/20/2025
# - Changed the inference from HF b/c 
#     API calls are not very limted.
# - Added API call limiting to allow for demoing
# - Added support for adding your own API token.     

# + 04/16/2025  
# - Changed the inference points on HF b/c
#     older points no longer supported.
    
# '''
# #-------------------------------------------











#==========================================================
# Logging
#  --------------------------------------------

import requests
from datetime import datetime


try:
    LOGGER_TOOL_WEBHOOK = os.environ.get("LOGGER_TOOL_URL")
except Exception as e:
    print(f"❌ Error in loading LOGGER_TOOL_WEBHOOK")

    

def log_to_webhook(
    *,
    session_info: dict,
    model: str,
    prompt: str,
    response: str,
    temperature: float,
):
    if not LOGGER_TOOL_WEBHOOK:
        return

    payload = {
        #Session info
        **session_info,

        #Model info
        "model": model,
        "temperature": temperature,

        #Content
        "user_prompt": prompt,
        "assistant_response": response,

        #Usage
        "api_call_count": st.session_state.api_call_count,
        "api_call_limit": API_CALL_LIMIT,
        "remaining_calls": API_CALL_LIMIT - st.session_state.api_call_count,

        #Timestamp
        "timestamp": datetime.utcnow().isoformat(),
    }

    try:
        requests.post(LOGGER_TOOL_WEBHOOK, json=payload, timeout=3)
    except Exception as e:
        print("Logging failed")

#  --------------------------------------------


#==========================================================
# Unique Users / Session Info
#  --------------------------------------------
import uuid
import time
import hashlib
import json
import sys
from datetime import datetime

def get_session_info():
    data = {
        "timezone": time.tzname,
        "platform": sys.platform,
        "rand": uuid.uuid4().hex,
    }
    raw = json.dumps(data, sort_keys=True)
    return hashlib.sha256(raw.encode()).hexdigest()[:12]


if "session_info" not in st.session_state:
    st.session_state.session_info = {
        "session_id": str(uuid.uuid4()),
        "session_start": datetime.utcnow().isoformat(),
        "conversation_id": str(uuid.uuid4()),
        "run_count": 0,
        "fingerprint": get_session_info(),
        "platform": sys.platform,
        "timezone": time.tzname,
    }

st.session_state.session_info["run_count"] += 1



def reset_conversation():
    st.session_state.conversation = []
    st.session_state.messages = []
    st.session_state.session_info["conversation_id"] = str(uuid.uuid4())

#  --------------------------------------------




#==========================================================
# Limits
#  --------------------------------------------

API_CALL_LIMIT = 20 # Define the limit

if 'api_call_count' not in st.session_state:
    st.session_state.api_call_count = 0
    st.session_state.remaining_calls = API_CALL_LIMIT



REQUEST_COOLDOWN = 3  # seconds between requests

if "last_request_time" not in st.session_state:
    st.session_state.last_request_time = 0
#  --------------------------------------------


    

model_links_hf ={
    "Gemma-3-27B-it":{
                      "inf_point":"https://router.huggingface.co/v1",
                      "link":"google/gemma-3-27b-it:scaleway",
                      },
    "Meta-Llama-3.1-8B":{
                      "inf_point":"https://router.huggingface.co/v1",
                      "link":"meta-llama/Meta-Llama-3.1-8B-Instruct:scaleway",
                      },
    "DeepSeek-R1-Distill-Llama-70B":{
                    "inf_point":"https://router.huggingface.co/v1",
                    "link":"deepseek-ai/DeepSeek-R1-Distill-Llama-70B:scaleway",
                     },
    "Qwen2.5-Coder-32B-Instruct":{
                        "inf_point":"https://router.huggingface.co/v1",
                        "link":"Qwen/Qwen3-235B-A22B-Instruct-2507:scaleway",
                        },

      # "Mistral-7B":{
      #                 "inf_point":"https://router.huggingface.co/v1",
      #                 "link":"mistralai/Mistral-7B-Instruct-v0.2",
      #                 },
      # "Gemma-2-27B-it":{
      #                 "inf_point":"https://router.huggingface.co/nebius/v1",
      #                 "link":"google/gemma-2-27b-it-fast",
      #                 },
      # "Gemma-2-2B-it":{
      #                 "inf_point":"https://router.huggingface.co/nebius/v1",
      #                 "link":"google/gemma-2-2b-it-fast",
      #                 },
      # "Zephyr-7B-β":{
      #                 "inf_point":"https://router.huggingface.co/hf-inference/models/HuggingFaceH4/zephyr-7b-beta/v1",
      #                 "link":"HuggingFaceH4/zephyr-7b-beta",
      #                 },
  }


model_links_groq ={
      "OpenAI-GPT-OSS-120B":{
                      "inf_point":"https://api.groq.com/openai/v1",
                      "link":"openai/gpt-oss-120b",
                      },    
      "Meta-Llama-3.1-8B":{
                      "inf_point":"https://api.groq.com/openai/v1",
                      "link":"llama-3.1-8b-instant",
                      },
      "Kimi-K2-Instruct":{
                  "inf_point":"https://api.groq.com/openai/v1",
                  "link":"moonshotai/kimi-k2-instruct",
                  },

      # "Gemma-2-9B-it":{
      #                 "inf_point":"https://api.groq.com/openai/v1",
      #                 "link":"gemma2-9b-it",
      #                 },
  }

#Pull info about the model to display
model_info ={
    "OpenAI-GPT-OSS-120B":
        {'description':"""The GPT OSS 120B model is a **Large Language Model (LLM)** that's able to have question and answer interactions.\n \
            \nIt was created by the [**OpenAI**](https://openai.com/research) team as an open-source initiative and has over  **120 billion parameters.** \
            \nThis model represents one of the largest publicly available transformer-based language models, designed for advanced reasoning, dialogue, and code understanding tasks.\n""",
        'logo':'https://registry.npmmirror.com/@lobehub/icons-static-png/1.74.0/files/light/openai.png'},
    "Mistral-7B":
        {'description':"""The Mistral model is a **Large Language Model (LLM)** that's able to have question and answer interactions.\n \
            \nIt was created by the [**Mistral AI**](https://mistral.ai/news/announcing-mistral-7b/) team as has over  **7 billion parameters.** \n""",
        'logo':'https://cdn-avatars.huggingface.co/v1/production/uploads/62dac1c7a8ead43d20e3e17a/wrLf5yaGC6ng4XME70w6Z.png'},
    "Gemma-2-27B-it":        
        {'description':"""The Gemma model is a **Large Language Model (LLM)** that's able to have question and answer interactions.\n \
            \nIt was created by the [**Google's AI Team**](https://blog.google/technology/developers/gemma-open-models/) team as has over  **27 billion parameters.** \n""",
        'logo':'https://pbs.twimg.com/media/GG3sJg7X0AEaNIq.jpg'},
    "Gemma-3-27B-it":        
        {'description':"""The Gemma model is a **Large Language Model (LLM)** that's able to have question and answer interactions.\n \
            \nIt was created by the [**Google's AI Team**](https://blog.google/technology/developers/gemma-open-models/) team as has over  **27 billion parameters.** \n""",
        'logo':'https://pbs.twimg.com/media/GG3sJg7X0AEaNIq.jpg'},
    "Gemma-2-2B-it":        
        {'description':"""The Gemma model is a **Large Language Model (LLM)** that's able to have question and answer interactions.\n \
            \nIt was created by the [**Google's AI Team**](https://blog.google/technology/developers/gemma-open-models/) team as has over  **2 billion parameters.** \n""",
        'logo':'https://pbs.twimg.com/media/GG3sJg7X0AEaNIq.jpg'},
    "Gemma-2-9B-it":        
        {'description':"""The Gemma model is a **Large Language Model (LLM)** that's able to have question and answer interactions.\n \
            \nIt was created by the [**Google's AI Team**](https://blog.google/technology/developers/gemma-open-models/) team as has over  **9 billion parameters.** \n""",
        'logo':'https://pbs.twimg.com/media/GG3sJg7X0AEaNIq.jpg'},
    "Zephyr-7B":        
        {'description':"""The Zephyr model is a **Large Language Model (LLM)** that's able to have question and answer interactions.\n \
            \nFrom Huggingface: \n\
            Zephyr is a series of language models that are trained to act as helpful assistants. \
            [Zephyr 7B Gemma](https://huggingface.co/HuggingFaceH4/zephyr-7b-gemma-v0.1)\
            is the third model in the series, and is a fine-tuned version of google/gemma-7b \
            that was trained on on a mix of publicly available, synthetic datasets using Direct Preference Optimization (DPO)\n""",
        'logo':'https://huggingface.co/HuggingFaceH4/zephyr-7b-gemma-v0.1/resolve/main/thumbnail.png'},
    "Zephyr-7B-β":        
        {'description':"""The Zephyr model is a **Large Language Model (LLM)** that's able to have question and answer interactions.\n \
            \nFrom Huggingface: \n\
            Zephyr is a series of language models that are trained to act as helpful assistants. \
            [Zephyr-7B-β](https://huggingface.co/HuggingFaceH4/zephyr-7b-beta)\
            is the second model in the series, and is a fine-tuned version of mistralai/Mistral-7B-v0.1 \
            that was trained on on a mix of publicly available, synthetic datasets using Direct Preference Optimization (DPO)\n""",
        'logo':'https://huggingface.co/HuggingFaceH4/zephyr-7b-alpha/resolve/main/thumbnail.png'},
    "Meta-Llama-3-8B":
        {'description':"""The Llama (3) model is a **Large Language Model (LLM)** that's able to have question and answer interactions.\n \
            \nIt was created by the [**Meta's AI**](https://llama.meta.com/) team and has over  **8 billion parameters.** \n""",
        'logo':'Llama_logo.png'},
    "Meta-Llama-3.1-8B":
        {'description':"""The Llama (3.1) model is a **Large Language Model (LLM)** that's able to have question and answer interactions.\n \
            \nIt was created by the [**Meta's AI**](https://llama.meta.com/) team and has over  **8 billion parameters.** \n""",
        'logo':'Llama3_1_logo.png'},
    "DeepSeek-R1-Distill-Llama-70B":
        {'description':"""DeepSeek-R1-Distill-Llama-70B is a **Large Language Model (LLM)** distilled from the DeepSeek-R1 reasoning family using the Llama architecture. \
            \nIt is designed to retain strong capabilities in reasoning, coding, and general text generation while being more accessible than the full DeepSeek-R1 model. \
            \nLearn more on HuggingFace: https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Llama-70B""",
        'logo':'https://cdn-avatars.huggingface.co/v1/production/uploads/6538815d1bdb3c40db94fbfa/xMBly9PUMphrFVMxLX4kq.png'},
    "Qwen2.5-Coder-32B-Instruct":
        {'description':"""Qwen2.5-Coder-32B-Instruct is a **Large Language Model (LLM)** in the Qwen2.5-Coder series tailored for code generation, reasoning, and instruction-following tasks. \
            \nBuilt on the Qwen2.5 architecture, this 32B-parameter model is optimized for coding, debugging, and developer use cases. \
            \nLearn more on HuggingFace: https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct""",
        'logo':'https://cdn-avatars.huggingface.co/v1/production/uploads/620760a26e3b7210c2ff1943/-s1gyJfvbE1RgO5iBeNOi.png'},
    "Kimi-K2-Instruct":
        {'description':"""The Kimi-K2-Instruct model is a **Large Language Model (LLM)** that's able to have question and answer interactions.\n \
            \nIt was created by the [**Moonshot AI**](https://www.moonshot.cn/) team as part of the Kimi model family. \
            \nThe model is designed for instruction following, reasoning, and general conversational tasks, with a strong focus on high-quality responses and long-context understanding.\n""",
    'logo':'https://cdn-avatars.huggingface.co/v1/production/uploads/641c1e77c3983aa9490f8121/X1yT2rsaIbR9cdYGEVu0X.jpeg'},

}



#Random dog images for error message
random_dog = ["0f476473-2d8b-415e-b944-483768418a95.jpg",
              "1bd75c81-f1d7-4e55-9310-a27595fa8762.jpg",
              "526590d2-8817-4ff0-8c62-fdcba5306d02.jpg",
              "1326984c-39b0-492c-a773-f120d747a7e2.jpg",
              "42a98d03-5ed7-4b3b-af89-7c4876cb14c3.jpg",
              "8b3317ed-2083-42ac-a575-7ae45f9fdc0d.jpg",
              "ee17f54a-83ac-44a3-8a35-e89ff7153fb4.jpg",
              "027eef85-ccc1-4a66-8967-5d74f34c8bb4.jpg",
              "08f5398d-7f89-47da-a5cd-1ed74967dc1f.jpg",
              "0fd781ff-ec46-4bdc-a4e8-24f18bf07def.jpg",
              "0fb4aeee-f949-4c7b-a6d8-05bf0736bdd1.jpg",
              "6edac66e-c0de-4e69-a9d6-b2e6f6f9001b.jpg",
              "bfb9e165-c643-4993-9b3a-7e73571672a6.jpg",
              "d467a3b8-ade5-4d68-810a-95fbb32a3cfc.jpg",
              "5384c2a7-9b73-478e-9f32-9af9f264da1d.jpg",
              "59f02432-b972-4428-935b-4efb0af83456.jpg"]



def reset_conversation():
    '''
    Resets Conversation
    '''
    st.session_state.conversation = []
    st.session_state.messages = []
    return None
    


# --- Sidebar Setup ---
st.sidebar.title("Chatbot Settings")

#Define model clients
client_names = ["Provided API Call", "HF-Token"]
client_select = st.sidebar.selectbox("Select Model Client", client_names)






if "HF-Token" in client_select:
    try:
        if "API_token" not in st.session_state:
            st.session_state.API_token = None

        st.session_state.API_token = st.sidebar.text_input("Enter your Hugging Face Access Token", type="password")
        model_links = model_links_hf

    except Exception as e:
        st.sidebar.error(f"Credentials Error:\n\n {e}")

elif "Provided API Call"  in client_select:
    try:
        if "API_token" not in st.session_state:
            st.session_state.API_token = None

        st.session_state.API_token = os.environ.get('GROQ_API_TOKEN')#Should be like os.environ.get('HUGGINGFACE_API_TOKEN')

        model_links = model_links_groq

    except Exception as e:
        st.sidebar.error(f"Credentials Error:\n\n {e}")





# Define the available models
models =[key for key in model_links.keys()]

# Create the sidebar with the dropdown for model selection
selected_model = st.sidebar.selectbox("Select Model", models)

#Create a temperature slider
temp_values = st.sidebar.slider('Select a temperature value', 0.0, 1.0, (0.5))



#Add reset button to clear conversation
st.sidebar.button('Reset Chat', on_click=reset_conversation, type="primary") #Reset button

# Contact info
# Contact info
st.sidebar.markdown(
    "<span style='font-size:0.85em; color:#bbbbbb; font-style:italic;'>"
    "Created by "
    "<a href='https://ngebodh.github.io/' target='_blank' style='color:#bbbbbb; text-decoration:none;'>"
    "Nigel Gebodh</a><br>"
    "Chatbots do not have access to real-time info. Agentic chat coming soon!"
    "</span>",
    unsafe_allow_html=True
)



st.sidebar.divider() # Add a visual separator




# Create model description
st.sidebar.subheader(f"About {selected_model}")
st.sidebar.write(f"You're now chatting with **{selected_model}**")
st.sidebar.markdown(model_info[selected_model]['description'])
st.sidebar.image(model_info[selected_model]['logo'])
st.sidebar.markdown("*Generated content may be inaccurate or false.*")
st.sidebar.markdown("\nLearn how to build this chatbot [here](https://ngebodh.github.io/projects/2024-03-05/).")
st.sidebar.markdown("\nRun into issues? \nTry coming back in a bit, GPU access might be limited or something is down.")




if "prev_option" not in st.session_state:
    st.session_state.prev_option = selected_model

if st.session_state.prev_option != selected_model:
    st.session_state.messages = []
    st.session_state.prev_option = selected_model
    reset_conversation()



#Pull in the model we want to use
repo_id = model_links[selected_model]

# initialize the client
client = OpenAI(
  base_url=model_links[selected_model]["inf_point"],#"https://api-inference.huggingface.co/v1",
  api_key=st.session_state.API_token#os.environ.get('HUGGINGFACE_API_TOKEN')#"hf_xxx" # Replace with your token
) 


st.subheader(f'AI - {selected_model}')

# Set a default model
if selected_model not in st.session_state:
    st.session_state[selected_model] = model_links[selected_model] 

# Initialize chat history
if "messages" not in st.session_state:
    st.session_state.messages = []


# Display chat messages from history on app rerun
for message in st.session_state.messages:
    with st.chat_message(message["role"]):
        st.markdown(message["content"])





if prompt := st.chat_input(f"Hi I'm {selected_model}, ask me a question "):

    # Display user message in chat message container
    with st.chat_message("user"):
        st.markdown(prompt)
    # Add user message to chat history
    st.session_state.messages.append({"role": "user", "content": prompt})

    #Cooldown check
    now = time.time()
    elapsed = now - st.session_state.last_request_time
    
    if elapsed < REQUEST_COOLDOWN:
        wait_time = round(REQUEST_COOLDOWN - elapsed, 1)
        st.warning(f"⏳ Please wait before sending another request.")
        st.stop()

    st.session_state.last_request_time = now


    
    if st.session_state.api_call_count >= API_CALL_LIMIT:
        
        # Add the warning to the displayed messages, but not to the history sent to the model
        response = f"LIMIT REACHED: Sorry, you have reached the API call limit for this session."
        # st.write(response)
        st.warning(f"Sorry, you have reached the API call limit for this session.")
        st.session_state.messages.append({"role": "assistant", "content": response })


    else:
        # Display assistant response in chat message container
        with st.chat_message("assistant"):
            try:
                st.session_state.api_call_count += 1
                # Add a spinner for better UX while waiting
                with st.spinner(f"Asking {selected_model}..."):

                    stream = client.chat.completions.create(
                        model=model_links[selected_model]["link"],
                        messages=[
                            {"role": "system", "content": "You are a helpful assistant. Always respond briefly in 1–3 sentences."},
                            *[
                                {"role": m["role"], "content": m["content"]}
                                for m in st.session_state.messages
                            ]
                        ],
                        temperature=temp_values,#0.5,
                        stream=True,
                        max_tokens=800,#1500, #3000,
                    )

                    response = st.write_stream(stream)

                    remaining_calls = (API_CALL_LIMIT) - st.session_state.api_call_count
                    st.markdown(f"\n\n <span style='float: right; font-size: 0.8em; color: gray;'>API calls:({remaining_calls}/{API_CALL_LIMIT})</span>", unsafe_allow_html=True)
                    
                    #Logging   
                    try:
                        log_to_webhook(
                            session_info=st.session_state.session_info,
                            model=selected_model,
                            prompt=prompt,
                            response=response,
                            temperature=temp_values,
                        )
                    except Exception:
                        pass

                        
            except Exception as e:
                response = "😵‍💫 Looks like someone unplugged something!\
                        \n Either the model space is being updated or something is down.\
                        \n\
                        \n Try again later. \
                        \n\
                        \n Here's a random pic of a 🐶:"
                st.write(response)
                random_dog_pick = 'https://random.dog/'+ random_dog[np.random.randint(len(random_dog))]
                st.image(random_dog_pick)
                st.write("This was the error message:")
                st.write(e)



        
        st.session_state.messages.append({"role": "assistant", "content": response})