A

Project Report on

Talk -Doc-Assistant

Submitted to
Savitribai Phule Pune University

In the partial fulfillment of the requirement of the award of the degree

of
Bachelor of Computer Application
TYBCA, Sem: VI
Academic Year 2024-25

Submitted by
Mr. Shubham Shivaji Suryawanshi

Under the guidance of

Prof. Preeti Lokhande

Department of Computer Application

Alandi (D)- Pune 412105

1
 Alandi (D) - Pune 412105

CERTIFICATE

Department of Computer Application

This is to certify that Mr. Shubham Suryawanshi of TYBCA, Sem VI,

Exam Seat No. _______________ , has/have successfully
completed project work entitled Talk-Doc-Assistant in the partial
fulfilment of the requirement of the degree of Bachelor of
Computer Application for the Academic Year 2024-25

Prof Preeti Lokhande Dr. Vikas Mahandule

Project Guide Head of the Department

Internal Examiner External Examiner

2
 Acknowledgement

The words are not enough to express my thanks to Dr. B. B. Waphare

Principal, MIT Arts Commerce and Science College Alandi (D) for
providing me with the opportunity to avail the excellent facilities and
infrastructure of the institute.

It is my proud privilege to express my profound gratitude to Dr. Vikas

Mahandule HOD, Computer Application Department, for his astute
guidance, constant encouragement and sincere support throughout my
academic course.

I thanks to my honest gratitude to Prof. Preeti Lokhande for her inspiration,

constructive suggestions and affectionate guidance in my project work
completion.

Last but not the least, I express my sincere thanks to all my dear friends and
family members for their constant motivation, moral support and invariable
direction throughout my life

3
 DECLARATION

I Shubham Suryawanshi, here by declare that this project work entitled Talk-
Doc-Assistant submitted at MIT, Arts Commerce and Science College,
Alandi(D), (Affiliated to Savitribai Phule Pune University) is a record of original
work done by me under the supervision and guidance of Prof. Preeti Lokhande
Department of Computer Application.

Signature:

Name of Student: Shubham Shivaji Suryawanshi

Exam Seat No:

Place: Alandi (D), Pune

Date:

4
 INDEX
Chapter No Contents Page no
Chapter 1 INTRODUCTION

1.1 Problem Statement 6

1.2 Objective and Goals 6

1.3 Project Scope and Limitations 7
Chapter 2 SYSTEM ANALYSIS

2.1 Existing System 8

2.2 Project Perspective and Features 8

2.3 Functional Requirement 9
2.4 Hardware 9
Chapter 3 SYSTEM DESIGN

3.1 DFD 10-11

3.2 ERD 12

3.3 Data Dictionary 13

Chapter 4 FORM DESIGN (With Input Values) 14-16

Chapter 5 OUTPUT RESULTS (With Reports) 17-19

Chapter 6 CONCLUSION & FUTURE SCOPE

20-21

Chapter 7 REFERENCES 22

5
1. Introduction:
The advancement of artificial intelligence has paved the way for intelligent conversational agents
that can simulate human-like interactions. AI-powered chatbots have emerged as a critical
component in domains such as customer service, education, and business automation. This project
aims to develop an AI chatbot that integrates FastAPI, [Link], PostgreSQL, and DeepSeek R-
1:1.5B to deliver real-time, contextually aware conversations.
The chatbot will provide text and voice-based communication, allowing users to interact
efficiently. This enhances the chatbot's usability across multiple domains, such as customer
support, academic assistance, and automated workflows.

1.1 Problem Statement

With the increasing need for efficient digital assistants, users often struggle with chatbots that
lack natural conversation capabilities, document processing features, or real-time voice
interactions. Existing cloud-based solutions come with privacy concerns, high response latency,
and dependency on third-party APIs. To address these limitations, this project aims to develop a
self-hosted AI-powered chatbot that seamlessly integrates voice, text, and document processing
while ensuring data privacy and efficient performance using locally installed AI models.

1.2 Objective and Goals

The primary objectives of this Talk-Doc chatbot project are:
• Develop an AI-powered chatbot capable of engaging in real-time conversations.
• Enable voice and text-based interactions to enhance accessibility for diverse users.
• Ensure data privacy and security by hosting the chatbot locally instead of using cloud-based
APIs.
• Optimize system performance by leveraging FastAPI for backend processing and
PostgreSQL for efficient data storage.
• Create a user-friendly UI using [Link] to provide a smooth and intuitive interaction
experience.
• Reduce response latency and improve efficiency by using a self-hosted DeepSeek AI model
instead of cloud-based alternatives.

6
Project Goal :
• To build a highly efficient, locally hosted AI chatbot capable of handling voice and text
queries.
• To provide a seamless user experience with [Link] UI and real-time voice processing.
• To improve accessibility by integrating speech recognition and text-to-speech capabilities.
• To maintain high-speed responses and reduce dependence on external AI models.

1.3 Project Scope and Limitations

Scope
• Frontend Technologies: [Link] with the SpeechRecognition library for voice input and an
interactive UI.
• Backend Technologies: FastAPI as a lightweight, high-performance API framework
integrated with PostgreSQL for structured data storage.
• AI Model: DeepSeek R1 for natural language processing, text generation, and question-
answering tasks.
• Core Features:
o Conversational AI with voice-enabled support.
o Text-based chatbot functionality.
o Secure and private AI assistant (self-hosted).

Limitations
• Limited Model Customization: The AI model operates based on DeepSeek R1’s capabilities
and may require additional fine-tuning for specific domains.
• Resource Intensive: Running a local AI model requires sufficient hardware resources
(RAM, GPU/CPU).
• Language Constraints: The chatbot’s performance depends on the supported languages of
DeepSeek R1.
• No Cloud Integration: While ensuring privacy, the chatbot does not use cloud-based AI
APIs, which might limit external service integrations.

7
2. System Analysis :
2.1 Existing System:

• The existing AI chatbot solutions in the market rely heavily on cloud-based services such as
OpenAI's GPT models, Google Dialogflow, and IBM Watson. While these solutions provide
robust conversational AI capabilities, they come with several drawbacks, including high API
costs, data privacy concerns, and limited customization options.
• Traditional chatbot architectures follow a client-server model, where the frontend interacts
with a cloud-based AI model, which processes the user query and returns a response. This
leads to longer response times due to network latency and API call overhead. Furthermore,
most existing systems lack a built-in knowledge retrieval feature from documents , making
them less effective for enterprise and educational applications.
• Another limitation of current systems is their dependence on external services for text-to-
speech (TTS) and speech-to-text (STT), which increases complexity and cost. Many
platforms also do not support offline functionality, making them impractical for organizations
concerned about data security and sovereignty.

2.2 Project Perspective and Features:

The Talk-AI chatbot is designed as a self-hosted, AI-powered conversational assistant that integrates
voice, text, and document processing capabilities. Unlike cloud-based chatbots, this system operates
locally, ensuring data privacy, low latency, and reduced operational costs. By utilizing DeepSeek R-
1:1.5B, FastAPI, [Link], and PostgreSQL, the chatbot delivers a fast, interactive, and secure user
experience.
Key Features
1. Self-Hosted AI Model: Uses DeepSeek R-1:1.5B for natural conversation processing without
relying on third-party APIs.
2. FastAPI Backend: Supports asynchronous request handling for high-speed response times.
3. PostgreSQL Database: Stores chat history, enabling context-aware conversations.
4. Voice Recognition (STT) & Text-to-Speech (TTS): Allows users to speak queries and receive
spoken responses.
5. Interactive UI: Built using [Link] to provide a smooth user experience.
6. Scalable and Modular Design: Allows future integrations with business applications,
education tools, and IoT devices.
7. Secure Data Handling: Eliminates data privacy concerns associated with cloud-hosted
models.

8
2.3 Functional Requirements :
The functional requirements define the expected operations and interactions of the system.
User Requirements
• Users should be able to ask questions via text or voice.
• Users should receive text or voice-based AI-generated responses.
• Users should have a seamless chat experience with contextual memory.
• The chatbot should provide quick responses with minimal latency.
System Requirements
• The chatbot should process voice inputs and convert speech to text (STT).
• The chatbot should generate voice responses using Text-to-Speech (TTS).
• The system should support document uploads and extract relevant content for responses.
• The AI model should be self-hosted, ensuring data privacy.
• The database should store user interactions for contextual understanding.
• The UI should be responsive and accessible across devices.

2.4 Hardware and Software Requirements :

Hardware Requirements
• Processor: Multi-core CPU (Intel i7 or Ryzen 7 and above).
• Memory: Minimum 16GB RAM for optimal performance.
• GPU: NVIDIA GPU (RTX 3060 or higher) for AI inference acceleration.
• Storage: High-speed SSD (512GB or more) for faster data processing.
Software Requirements
• Operating System: Linux or Windows Server (Ubuntu 22.04 recommended).
• Backend Development: FastAPI (for high-speed asynchronous processing).
• Frontend Development: [Link] with SpeechRecognition for voice input.
• Database Management: PostgreSQL for structured chat history storage.
• AI Model: DeepSeek R-1:1.5B for natural language processing.

9
3. System Design :
3.1 DFD :
LEVEL 0 :

0.0
Ask Query Update 1
User 1
Chatbot Admin
Get Result Application

LEVEL 1:

Ask Query
1.0
User 1
Process
Get Output Query

Store Update
2.0 Query Database
1
Update

10
Level 2:

1.1
USER 1 ENTER

QUERY

1.2
Process input
with LLM

1.3
Database
Get Results

11
3.2 ERD

Id

1
Gives
User Input

Input

Process
Input &
Produces

Output

12
3.3 Data Dictionary :

Stores the chat History in Database :

13
4. Form Design :

Home Page :

14
User Interface with input values :

15
INPUT :
1. System Overview
The Talk-Doc Assistant is an AI-powered chatbot that accepts user input through voice-to-text or
typed text. The system processes the input, generates a response, and delivers the output via text-to-
speech (TTS) or text display.
2. Input Modes Supported
The assistant allows users to interact in two ways:
1. Voice Input (Speech-to-Text - STT)
o Users speak into the microphone.
o The assistant converts speech into text for processing.
2. Text Input (Typing)
o Users can type their message into the input field.
o The assistant processes the text and generates a response.
3. Observed Input Status (From Image)
• Microphone Icon (Red & Glowing): Indicates that voice input is active, meaning the assistant
is currently recording speech.
• Mute Icon (Yellow): Suggests that text-to-speech output is muted, preventing spoken
responses.
• Stop Icon (Red): Indicates that the user can stop the voice input manually.
• Text Input Field: The placeholder text "Type or speak your message..." suggests that both
voice and text inputs are available.
4. Expected Input Behaviour
When voice input is activated, the assistant should:
• Accurately convert speech to text.
• Detect pauses and auto-submit queries when the user stops speaking.
• Allow users to manually stop recording using the red stop button.
When text input is used, the assistant should:
• Accept typed messages.
• Process queries upon pressing Enter or clicking the send button (paper plane icon).
5. Potential Issues & Considerations
Possible Input Issues:
• Background noise could affect speech-to-text accuracy.

16
5. OUTPUT RESULTS :

17
5.1 Outputs Report :
1. System Overview
The Talk-Doc-Assistant is an AI-driven chatbot that processes user queries using voice-to-
text technology. It then generates responses in text format and delivers them using text-to-
speech (TTS) technology.
2. Input Processing (Voice-to-Text)
• Users speak their queries into the microphone.
• The assistant converts speech to text and processes the request.
• The system supports natural language processing (NLP) for better understanding.
3. Output Generation (Text-to-Speech)
• The assistant formulates a text-based response.
• The response is converted into speech using TTS technology.
• The system ensures clear pronunciation and natural intonation for an engaging
experience.

4. User Interactions & Assistant Responses :

User Inputs:
1. "how r u"
2. "suggest me a Sci fi movies"
Assistant Responses:
1. "I'm just a computer program created by Shubham. If you have any questions or
need assistance, feel free to ask!"
2. *"Sure! Here are some popular sci-fi movies:
1. The Star Wars: Episode IV - A New Hope
2. Inception
3. The Matrix
4. Interstellar
5. The Dark Knight

18
5. System Performance & Considerations
Strengths:
• Fast speech recognition for accurate text conversion.
• High-quality TTS voice output for natural responses.
• User-friendly and interactive chat experience.
Possible Improvements:
• Multilingual support for diverse users.
• Customizable voice tones for a personalized experience.
• Faster response time for real-time conversations.

19
[Link] & Future Scope:
The Talk-Doc-Assistant successfully integrates voice-to-text (STT) and text-to-speech (TTS)
technologies to provide an interactive AI-driven experience. It allows users to communicate using
both voice and text inputs, delivering responses in text format or as spoken output.
This project enhances user accessibility, engagement, and convenience by enabling real-time AI-
powered conversations. It effectively processes queries, provides relevant responses, and can be
further customized and improved with advanced AI, multilingual support, and enhanced
security features.
Looking ahead, the system has great potential for growth, including better speech recognition,
deeper AI-driven conversations, integration with third-party platforms, and improved
accessibility features. With continuous development, the Talk-Doc Assistant can become a highly
intelligent and versatile AI assistant, catering to various user needs across different domains.

Future Scope for Talk-Doc Assistant :

The Talk-Doc Assistant has great potential for future advancements. Below are key areas where it
can be improved and expanded:

1. Enhanced Speech Recognition (STT - Speech-to-Text)

• Multi-Language Support – Enable recognition and processing of multiple languages

and dialects.
Accurate Context Understanding – Improve NLP algorithms to better understand
accents, slang, and conversational tone.
Noise Reduction – Implement AI-powered noise filtering for better speech recognition in
noisy environments.

2. Improved Text-to-Speech (TTS) Capabilities

• Natural Voice Output – Upgrade to AI-driven human-like voices with better emotional
tones.
Customizable Voice Profiles – Allow users to choose different voices, tones, and
speaking speeds.
Multilingual TTS Support – Enable the assistant to respond in different languages
based on user preference.

20
3. Smarter AI and Conversational Intelligence

• Personalized Responses – Use machine learning to adapt responses based on user

behaviour and history.
Context Retention – Improve memory to keep track of previous conversations for better
continuity.
Emotion Detection – Recognize user emotions and adjust responses accordingly (e.g.,
supportive tone for sad users).

4. Advanced Integration with Other Platforms

• Social Media Integration – Allow direct posting to platforms like Instagram, Twitter,
and Facebook.
Cloud Storage & Document Handling – Enable saving and retrieving documents from
Google Drive, OneDrive, etc.
CRM & Business Tools – Integrate with customer support systems (e.g., Slack,
Microsoft Teams, Salesforce).

5. Enhanced UI/UX for Better User Interaction

• Dark & Light Mode Options – Provide theme customization for better readability.
Animated Assistant Avatars – Use AI avatars to make interactions more engaging.
Haptic Feedback & Gestures – Allow touch-based controls and haptic responses for
mobile users.

6. Security & Privacy Enhancements

• End-to-End Encryption – Secure conversations to protect user data.

On-Device Processing – Minimize cloud dependencies for improved privacy.
Customizable Data Retention Policies – Let users control how long their data is stored.

7. AI-Powered Accessibility Features

• Sign Language Support – Convert text responses into sign language videos for hearing-
impaired users.
Voice Command Shortcuts – Enable quick actions using voice (e.g., "Summarize this
document").
Adaptive Interfaces – Adjust UI based on user disabilities or preferences.

21
7. References :

1. Author Name, “LangChain: A Comprehensive Guide,” 2023.

2. PGVector Documentation, “Efficient Vector Search in PostgreSQL,” 2022

3. ReactJS Official Documentation, 2024.

4. Author Name, "Generative AI in Information Retrieval Systems," 2023.

5. Voice Recognition Insights, "Trends in Speech-to-Text Technology," 2022

6. OpenAI, "Building AI Chatbots with FastAPI and PostgreSQL," 2024.

7. DeepSeek Documentation, "DeepSeek R-1:1.5B Model Overview," 2024.

8. Mozilla TTS, "Text-to-Speech Implementation for AI Assistants," 2023.

9. SpeechRecognition Library, "Handling Real-time Voice Input in Web

Applications," 2023.

10. Framer Motion Docs, "Creating Smooth Animations in React," 2024.

22