Data

Tools

Indexes

Applications

Experiments

Awesome

An open API service indexing awesome lists of open source software.

https://github.com/karim-elkadhi/ai_chatbot

AI chatbot for customer service with real time text and voice interaction
https://github.com/karim-elkadhi/ai_chatbot

chatbot lang langfuse llm openai rag rag-chatbot sentiment-analysis speech-recognition speech-to-text text-to-speech tts

Last synced: 3 months ago
JSON representation

AI chatbot for customer service with real time text and voice interaction

Awesome Lists containing this project

README 

          
# 🧠AI Chatbot Platform for Insurance Services



> **MediBot** is an intelligent, multimodal chatbot platform built with Python.  

> It’s designed to enhance **insurance services** through voice and text interactions, emotion analysis, and data-driven dashboards.  



## 🧩 Overview



MediBot offers two main modules:  



- 👥 **User Module** — customers interact with the chatbot using **voice or text**, view their conversation history, and receive emotion-aware responses.  

- 🧑‍💼 **Admin Module** — decision-makers explore **KPIs, dashboards, and sentiment analytics**, manage users, and monitor global activity.



---

## 🚀 Key Features

### 🗣️ Multimodal Interaction



- Supports **Text** 💬 and  **Voice** 🎙️ input using Speech-to-Text (STT) and Text-to-Speech (TTS) technologies.



- Provides natural, human-like responses with audio output.



### 💬 Conversational Intelligence



- Real-time **emotion & intent detection** (e.g., *happy, sad, angry, neutral*).

- Generates automatic satisfaction scores based on emotion trends.



### 🧠 Sentiment & Emotion Monitoring



- Tracks sentiment evolution across each conversation.



- Displays **interactive graphs** showing emotional evolution.



### 🔐 Secure Authentication

- **JWT-based authentication** with role-based access control:



  - 🧍 **Users** — interact with the bot and review their conversation history.



  - 🧑‍💼 **Admins** — access advanced dashboards and user management tools.



### 💾 Persistent Data Storage



- Stores conversations, users, and sentiment scores in a SQLite database.



- Enables filtered queries and historical analysis.



---



## 🧍‍♂️ User Module



🧭 Click to expand



### Features

- Interact with MediBot via **text** or **voice commands**.  

- Enjoy **real-time voice replies** through Text-to-Speech (TTS).  

- Access **chat history** and review previous interactions.  

- Experience adaptive responses based on emotional tone.  



### Example User Flow

1. Log in with credentials.  

2. Start a conversation (voice or text).  

3. MediBot detects emotions, intent, and provides contextual responses.  

4. View conversation history revisit previous conversations.



![User Chat Interface](./utils/int_user.png) 



------------------------------------------------------------------------------------------------------



![User History](./utils/historique.png) 



---



## 🧑‍💼 Admin Module (Decision-Makers Dashboard)

The admin area empowers decision-makers with data-driven insights:



📊 Click to expand



### 📊 KPI & Analytics



- View real-time **metrics and visual dashboards** (user activity, sentiment trends, satisfaction scores).  

- Access aggregated analytics across all users and sessions.



### 🧾 Conversations Management



- Access a complete list of all users’ conversations.



- Filter, search, and export conversation histories.



### 👥 User Management



- Add 🟢, modify 🟡, or delete 🔴 users.  

- Manage **roles and permissions** dynamically.



### 😊 Sentiment Monitoring



- Visual dashboards for sentiment analysis across users and time periods.



- Identify negative trends or dissatisfaction in customer interactions.



 ![Admin Dashboard](./utils/int_admin.png) 



------------------------------------------------------------------------------------------------------



 ![Admin Sentiment](./utils/admin.png) 



---



### ⚙️ Technical Details



MediBot is built using a modular, scalable, and RAG-enabled AI architecture to provide context-aware and accurate responses.



### 🔹 Core AI Components



- Retrieval-Augmented Generation (RAG)



Uses LangChain to orchestrate retrieval of relevant information and generate responses.



Enhances factual accuracy by combining dataset knowledge with LLM generation.



- Vector Store & Semantic Search



Embeddings generated via SentenceTransformers (all-MiniLM-L6-v2 or multilingual variants).



FAISS vector store enables fast semantic search on the dataset (question, intent, category, response).



- Large Language Model Integration



Groq LLM generates natural, context-aware responses based on retrieved passages.



Prompt templates adapt dynamically to the user’s language using get_system_prompt(language).



- Multilingual Support



Detects input language automatically.



Retrieval and generation respect the detected language, ensuring consistent responses in French, English, or other languages.



- Observability & Monitoring



Langfuse tracks every RAG interaction: user query, retrieved context, LLM output, and metrics.



Enables debugging, analytics, and continuous improvement of the chatbot’s accuracy.



---

## 🧱 Project Structure



AI-chatbot/

├── main.py



├── utils



│ └── lang.py



├── dataset



│ ├── dataset.csv



├── database/



│ └── databaset.db



│ └── conversations.db



├── requirements.txt



├── README.md



## 📦 Installation



1. **Clone the repository**:



```bash

git clone https://github.com/Karim-ElKadhi/AI-chatbot.git

cd AI-chatbot

pip install -r requirements.txt

python app/main.py



## Access the interface



🧍 User Dashboard → http://localhost:5000/user



🧑‍💼 Admin Dashboard → http://localhost:5000/admin

```

📄 License



This project is licensed under the MIT License.