🤖 Conversa – AI-Powered Chat Assistant

📌 Overview

Conversa is a modern, high-performance AI-Powered Chat Assistant App built using Jetpack Compose and a Java Room Data Layer.

Developed as Task 2 during the 1-Month Android Development Internship at Codec Technologies (AICTE Recognized).

Designed with a fintech AI aesthetic (#0B0E14 dark slate, Electric Cyan #00E5FF, Deep Teal accents), Conversa provides intelligent conversational assistance, Speech-to-Text voice input, AI model selection via floating dialogs (Llama-3, DeepSeek-R1, Mistral), chat history persistence, and transcript sharing.

---

📱 App Screenshots

🏠 Welcome Screen & Starter Prompts	💬 AI Chat Stream & Floating Input Pill
🤖 Floating AI Model Selection Dialog	⚙️ Settings & API Configuration

---

🎯 Key Features

• 💬 Conversational AI Stream: Interactive LazyColumn chat stream with cyan gradient user bubbles, AI model cards, and typing indicators.
• 💊 Floating Pill Input Bar: Sleek floating rounded input pill bar docked right on top of the system navigation bar with smooth keyboard motion.
• 🎙️ Speech-to-Text Voice Input: Native Android RecognizerIntent voice recognition with automated prompt input insertion.
• ⚡ Free Public AI API Integration: Direct completion requests to free public AI endpoints (Llama-3, DeepSeek-R1, Mistral) plus a 100% offline fallback engine.
• 🗄️ Java Room History Storage: Thread-safe database data layer written in Java (ChatMessageEntity, ChatMessageDao, ConversaDatabase, ChatRepository).
• 🤖 Floating Model Selection Dialog: Clean floating modal dialog menu to switch between Llama-3.3-70B, DeepSeek-R1, Mistral-7B, and OpenAI-Free.
• 📤 Transcript Export & Share: Generate and share timestamped plaintext chat logs via Android FileProvider and system share sheet.
• ⚙️ Custom API Configuration: Option to enter custom OpenAI-compatible API endpoint URLs and API keys.
• 🌓 Fintech AI Gradient Theme: Adaptive dark and light modes with glowing cyan borders and Vidora-style inverted header corners.

---

🌐 API Integration

Conversa integrates directly with free public AI completion APIs (no subscription or account required), supporting models like Llama-3, DeepSeek, and Mistral.

Example completion request structure:

curl https://blockrun.ai/api/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4-flash",
    "messages": [
      {"role": "user", "content": "Hello, how are you?"}
    ]
  }'

---

🛠️ Architecture & Tech Stack

Component	Technology
Architecture	MVVM (Model-View-ViewModel) + Repository Pattern
Languages	Kotlin (UI & App Logic) + Java (Room Data Layer)
UI Framework	Jetpack Compose + Material Design 3 System
Local Database	Android Room Database (Java DAOs & Entities)
Speech Recognition	Android Native RecognizerIntent.ACTION_RECOGNIZE_SPEECH
Networking	OkHttp3 & Gson JSON Parser
File Sharing	Android FileProvider

---

📜 Internship Context

Developed as Task 2 during the 1-Month Android Development Internship at Codec Technologies (AICTE Recognized):

• Week 1: Core Architecture & Jetpack Compose Fundamentals
• Week 2: User Interaction, Forms & Navigation
• Week 3: Data Persistence (Room DB) & Networking
• Week 4: AI Integration, Voice Input & Final Submission

---

🚀 How to Run

Prerequisites

• Android Studio Ladybug (2024.2.1) or newer
• JDK 17
• Android Device / Emulator (API Level 24+)

Setup Instructions

1. Clone the repository:

   git clone https://github.com/Maheswara660/Conversa.git

2. Open project in Android Studio.
3. Allow Gradle to sync all dependencies.
4. Select target device and click Run (Shift + F10).

---

📧 Contact & Submission

• Developer: Rakamanda Maheswara Rao
• Repository: https://github.com/Maheswara660/Conversa
• Submitted To: vaishali@codectechnologies.in

---

🔖 License

This project is developed exclusively for educational and evaluation purposes under the Codec Technologies Internship Program. Redistribution or commercial use is not permitted.