Visun Your Personal AI Workspace

Visun is a voice-first AI workspace that you run yourself, that helps freelancers and studios to track projects, clients, deadlines, payments, and workflows all from your own computer.

Comment

Inspiration

Freelancers, creators, and small studios are dedicating more time to managing their work than creating it. Projects get buried on various platforms, deadlines become hard to keep up with, client messages are hard to stay on top of and payment followups become unmanageable.

We found that although AI assistants are strengthening, many productivity tools continue to rely on human input to work through dashboards, update spreadsheets and keep data synchronized across apps.

That's what motivated us to ask a very simple question: What if running your entire business could be as easy as talking to it?

That idea led us to build Visun Your Personal AI Workspace, a voice-first, agentic AI platform designed to bring projects, clients, deadlines, payments, and business operations into one intelligent workspace.

Our idea was to create a SaaS app that was non-traditional and enabled users to host the AI assistant on their own systems to have complete control over the data and the work the AI assistant performs.

What it does

What is Visun? Visun is a voice-first, intelligent business operations workspace for entrepreneurs, freelancers, and startups to orchestrate, automate, and tidy up their day-to-day business operations.

By chat with AI Copilot of Visun, users can talk to their AI Copilot to better explore the workspace, get data, and ready actions.

Key features include:

  • AI Copilot: a context-aware assistant, answering questions based on real workspace data and enabling operation using tools.

  • Gemini Live Voice Assistant: Instant voice chatting with natural back-and-forth, on-the-fly: interruption, human-like interaction in English, Hindi and Hinglish.

  • Project Management: manage projects, set up activities, create milestones, track progress, report on progress, and manage dates.

  • Smart Calendar: Preview upcoming deadlines and organize activities.

  • Scope Inbox: Monitor your client's evolving needs.

  • Payment Tracking: keep an eye on the amount of invoices, unpaid payments, and payment reminders.

  • Lead Management: List prospective clients and follow possibilities at various pipeline steps.

  • AI Reports: create business reports with tables and then export it as a PDF.

  • Agentic Automation: retrieve data and perform workspace actions using AI tools.

  • Human-in-the-Loop Approvals - Demand for clear approval before performing any potentially sensitive outbound action.

  • Other integrations: Enable third-party integrations with third-party services like Gemini PayPal Telegram Gmail Google Calendar.

  • Private Deployment: An architecture developed to be installed via Docker software on the customer's hardware (private server or computer).

Its approval based agentic workflow is one of the most critical feature of Visun.

A sensitive action suggested by the AI is pre-empted and is authorized by the user before being executed. This ensures a synergy between automation and human intervention.

How we built it

Our Visun uses a full-stack new-age architecture built on Next.js FastAPI Google Gemini, real-time Web Sockets, and persistent storage in a database.

Frontend

The frontend was built using Next.js, React, TypeScript, and Tailwind CSS. We designed an interactive workspace with sections dedicated to projects, deadlines, payments, leads, reports, activities, and integrations.

In addition, we created the animated marketing experience with Motion, GSAP, and Lenis including cinematic storytelling, editorial typography, interactive UI and layouts.

Backend

Our backend has been built in Python and FastAPI, for workspaces, AI chat integrations providers, reports, and actions executions.

We chose to use SQLModel + SQLite to persist data and Alembic to manage database migrations while preserving the ability to switch to PostgreSQL later for scaling.

AI and Voice Integration

We used Google Gemini for natural-language reasoning + tool-calling. And Google Gemini Live for real-time voice conversations.

A WebSocket architecture links the microphone in the browser to the backend, allowing for sending audio stream, receiving conversational responses, and interacting with tools through voice.

In addition, we added the ability to run workspace-aware tools with Google ADK based agent orchestration.

Security and Integrations

We've used Clerk to authenticate and authorize a workspace, with JWT validation and backend access rules.

Provider credentials are encrypted with Fernet and sensitive agent actions are integrated through an approval process that verifies the payload, checks the expiration, and executes the action in an idempotent fashion.

This architecture enables the separation of our Front End Backend Database and third-party services, helping Visun move to private, container-based deployments.

Challenges we ran into

1. Making AI Work With Real Business Data.

One of our biggest challenges was building an AI assistant that could understand real workspace information rather than return back random chatbot responses.

Our challenge was to combine the agent with project records, deadlines, invoices, conversations, and other structured data, but also to keep the generated responses grounded in what was available.

2. Real-Time Voice Communication

It was mainly difficult to integrate Gemini Live, which involves integrating the microphone input, streaming audio WebSocket playback and interruptions of the users.

Along the way we also improved the flow of conversations for bilingual English-Hindi and / or Hinglish conversations.

3. Reliable Agentic Automation

Introducing AI into any activity is risky, mostly payments, emails, and integrations.

We tackled this by coming up with a human-in-the-loop approval structure, where sensitive actions are prepared separately from their execution and an approval of the action is required before executing the action.

4. Managing Multiple External Integrations

Absorbing various providers like PayPal, Telegram, Gmail, and Google Calendar involved use of different API's credential authentication, and failure cases.

A centralized credential-management method was put in place using encryption, workspace isolation, and connection validation.

5. Designing for Privacy and Self-Hosting

We wanted to give Visun users more ownership and control compared to traditional cloud-only office suites.

This established architectural issues for database persistence, encrypted credentials authentication portability of deployment, and configuration.

Accomplishments that we're proud of

Our most significant achievement has been taking Visun from a concept to a working, interconnected workspace for AI operations.

We are mainly proud of:

  • Creating a voice-first AI Copilot by enabling streamable real-time Gemini Live conversations, and workspace-contextual intelligence.

  • Developing an agentic tool system that can get operation information and prepare an actions instead of a simple chat bot.

  • Using human-in-the-loop approvals, with authorisation validation, payload hashing, expiration, and idempotency protections.

  • Integrated working area for projects customers deadlines payments reports, leads, and activities.

  • Combining various third-party integrations via a common provider and credential-management structure.

  • Storing encrypted credentials so that sensitive provider keys are handled by the backend instead of being exposed in the browser.

  • Designing a visually striking frontend with animated narrative, interactive features, and a clear product identity.

  • The development of a modular architecture capable of supporting each installation deployed and self-hosted.

What makes us most proud is that Visun brings together AI reasoning, real-time voice interaction, practical business workflows, and user-controlled automation within one cohesive product.

What we learned

Building Visun showed us that creating agentic AI apps calls for orders of magnitude more work than just gaping an LLM into a chatbot interface.

We learned how to combine language models and external software systems, create safe tool-calling pipelines, and handle structured data while it passes through multiple services.

The experience of working on the project with Gemini Live made us comprehend streaming of real-time audio, WebSocket communication interruptions conversational AI.

In addition we benefited from:

  • Next.js and FastAPI's full stack.

  • The definition and implementation of AI agent orchestration and function calling.

  • Live voice interactions.

  • SQLModel migrations persistent storage.

  • Manage credentials and authenticate confidently.

  • Third-party API integrations.

  • Automation driven by humans in the loop.

  • Responsive ui design and powerful frontend animations.

  • Design software that supports containerization and self-hosted deployment.

Most importantly, we also learned that agentic AI is really about intelligence plus context, plus reliability, and plus human oversight.

A good AI assistant isn't just providing an answer, it's aware of your environment, and it helps you take action and is clear about the limits of its ability to help.

What's next for Visun Your Personal AI Workspace

The vision, as the company, is to develop Visun into a powerful AI operations assistant for freelancers, independent professionals, and small businesses that's privately deployable.

Our next development priorities include:

1. Simplified Self-Hosting

Simplify the Docker installation process so users can run Visun on their personal computers or private servers.

2. Local Authentication

Eliminate reliance on third-party login services for customer owned installations through introduction of a secure local login.

3. More Natural Voice Interaction

Improve Gemini Live with improved Indian English and Hinglish experiences, reduced latency, and improved handling of interruptions.

4. Expanded Agentic Capabilities

Display other tools for co-ordinating projects timetables clients, communications, and repetitive chores.

5. Better Integrations

Improve external productivity Platform support and Optimize Authentication, synchronization, and integration.

6. Improved Privacy and Deployment Management

Backup / recovery / local configuration / Install Specific Credential Management.

7. Personalized AI Workflows

Allow your customers to define how they want theVisun product to manage their particular business process, priorities, and automation.

We want Visun to be more than just another productivity app.

We aim for Visun to be the operating layer for independent professionals an intelligent work environment that gets their business, talks naturally, automates ethically, and puts them in charge of how they work.

Built With

Share this project:

Updates

Submission history