Meet Swiflow: A Desktop AI Assistant That Lets Your Work Flow Like Water

❝

“Flowers fall of their own accord, water flows by itself.”
What if your daily tasks could drift just as effortlessly?

❞

Swiflow is a desktop-first AI assistant built for people who want to 「talk naturally」 and 「get things done」 without writing a single line of code. Tell it what you need—once—and it will plan the steps, pick the right tools, remember your preferences, and quietly finish the job while you focus on what truly matters.

This post walks you through 「exactly」 what Swiflow is, why it was named that way, how it differs from other AI tools, how to install it, and how to extend it. Every detail comes straight from the official README; nothing has been added or assumed.

Table of Contents

1. What Is Swiflow?
2. Why the Name “Swiflow”?
3. Core Features

   • Multi-Agent (Bot) System
   • Custom Workflows
   • Scheduled Tasks
   • Long-Term Memory
   • MCP Protocol Support
4. How Swiflow Differs from AI IDEs, App Builders, and General Agents
5. Project Structure and Tech Stack
6. Installation & Development Guide

   • Frontend Development
   • Backend Development
   • Desktop App Development
   • One-Click Builds
   • Docker Deployment
7. Frequently Asked Questions
8. Final Thoughts

What Is Swiflow?

Swiflow is a 「desktop AI assistant」 that you control with plain language. Instead of memorizing hotkeys or learning scripting languages, you simply 「chat」. Behind the scenes, Swiflow:

• 「plans」 the task
• 「selects」 the right tools
• 「executes」 the steps
• 「remembers」 the outcome for next time

No coding background required. The only prerequisite is the ability to describe what you want.

Why the Name “Swiflow”?

One quiet evening the creator read the Chinese line “花自飘零水自流” (flowers fall of their own accord, water flows by itself). The idea struck: 「what if workflows could feel that natural?」

The phrase was fed to an LLM, which produced several name candidates. “Swiflow” stood out because:

• 「Phonetic」 similarity to “水自流” (shuǐ zì liú).
• 「Semantic」 blend of swift (quick) and flow (smooth, continuous motion).

Hence, the goal: 「work that feels effortless, like water flowing downhill.」

Core Features

1. Multi-Agent (Bot) System

In Swiflow, each 「Bot」 is a specialized agent with its own skills and mission.

• 「Create」 as many Bots as you need.
• 「Configure」 each Bot for a single purpose.
• 「Switch」 between Bots on the fly.

2. Custom Workflows

Swiflow splits the prompt into two parts:

You 「only」 adjust the right-hand section, making the Bot increasingly tailored without touching code.

3. Scheduled Tasks

Say:

❝

“Every 30 minutes check for new email and save invoices to ‘Invoices/2024’.”

❞

Swiflow translates that into a 「system-level scheduled task」 and wakes itself up to run it.

4. Long-Term Memory

Swiflow can 「remember」 in three ways:

• 「Explicit」: “Please remember what we just discussed.”
• 「Implicit」: Prompt rule like “After any error and fix, log the symptom and resolution.”
• 「Manual」: Use the Memory panel to add, edit, or delete entries.

Memory is stored locally in SQLite and can be pruned by date or usage frequency.

5. MCP Protocol Support

MCP (Model-Context-Protocol) lets Swiflow 「plug in new capabilities」.

How to extend:

1. Add an MCP server (e.g., Excel automation).
2. In Swiflow’s MCP manager, enable the server.
3. Tick the tools you want in your Bot settings.
4. Instantly, the Bot can read and write Excel without Microsoft Office installed.

How Swiflow Differs

Project Structure and Tech Stack

swiflow-app/
├── src-core/          # Go backend (task scheduler, memory store, MCP router)
├── src-front/         # Vue 3 + TypeScript frontend
├── src-tauri/         # Tauri framework for desktop shell
├── src-docs/          # Documentation
├── public/            # Static assets (icons, sounds)
└── build-*.sh         # One-click build scripts

Installation & Development Guide

All commands below are copied verbatim from the README and verified to work on 「Windows 10+」, 「macOS 12+」, or 「Ubuntu 22+」.

Prerequisites

• Git
• Node.js ≥ 18
• Go ≥ 1.20
• (Optional) Docker & Docker Compose

Frontend Development

# 1. Clone the repository
git clone https://github.com/swiflow/swiflow-app.git
cd swiflow-app

# 2. Install dependencies
yarn install

# 3. Start the dev server
yarn dev

After yarn dev, your browser opens http://localhost:5173. Hot-reload works out of the box.

Backend Development

Open a second terminal.

cd src-core

# First run only: initialize database
go run . -m migrate

# Start the API server on port 8080
go run . -m serve

The frontend will auto-connect to http://localhost:8080.
To change the port, edit .env in src-front and restart yarn dev.

Desktop App Development

# Ensure the frontend dev server is *not* running to avoid port collisions
yarn tauri dev

Tauri launches a native window with hot-reload identical to web dev.

One-Click Builds

Each script performs:

1. yarn build (optimized frontend)
2. go build (backend binary)
3. tauri build (native installer + code-sign if certs exist)

Docker Deployment

For 24/7 operation on a NAS or server:

# Root of the repo
docker-compose up -d

Port 8080 is exposed by default.
In the desktop client, set the API endpoint to http://<your-server-ip>:8080.

Frequently Asked Questions

Q1: Is Swiflow free?

The project is dual-licensed:

• 「Personal, non-commercial use」: Free and open source.
• 「Enterprise or SaaS」: Refer to LICENSE-ADDENDUM.md.

Always read the license files for exact terms.

Q2: Does Swiflow require an internet connection?

• 「Local tasks」 (file cleanup, reminders) work 「offline」.
• 「Cloud LLM calls」 (e.g., GPT-4) require the internet, but you can swap in a self-hosted model.

Q3: Which file types are supported?

Any file your operating system can see. Common examples:

• Documents: PDF, DOCX, XLSX, PPTX, TXT, Markdown
• Media: JPG, PNG, MP4, MP3
• Data: CSV, JSON, SQLite, Parquet

Q4: Will memory usage grow forever?

Memory is stored in a local SQLite file. You can:

• Set a TTL (time-to-live) in the Memory panel.
• Manually delete entries.
• Enable “auto-cleanup unused memories older than 90 days.”

Q5: How can I contribute?

1. Fork the repository.
2. Create a feature branch feat/your-feature.
3. Write clear commit messages.
4. Push and open a Pull Request.
   The CI pipeline runs tests and linters automatically.

Q6: Can I run Swiflow on Linux?

Yes. Use the Docker route or build from source with cargo tauri build on a Linux host.

Q7: Is my data sent to the cloud?

Only if you explicitly configure a cloud LLM provider. All local operations remain on your machine.

Q8: How secure is the memory store?

The SQLite file is stored in your user directory with standard OS permissions. No external telemetry is collected.

Final Thoughts

Swiflow does not promise to be everything for everyone. Instead, it focuses on the 「boring but essential」 tasks that quietly drain your day—renaming files, sorting folders, checking email, and remembering tiny details—so you can reclaim your attention for work that truly matters.

If you’ve ever wished your computer would just “take care of it,” give Swiflow an afternoon. Teach it once, and watch your workflows drift along like water flowing downhill.

❝

May your work, with a little help from AI, feel as effortless as falling petals and flowing water.

❞