Skip to content

Mancasvel/FlowSight.AI

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

137 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

FlowSight

Privacy-first developer productivity intelligence — runs locally on your machine.

License: AGPL v3 Commercial License available CLA required Support on Ko-fi Ask DeepWiki

FlowSight is a desktop application that helps distributed engineering teams understand how their work flows, without the surveillance baggage of traditional productivity tools. All sensitive processing happens on the developer's machine: screen context, git metadata, and activity summaries are analyzed by a bundled local LLM and never leave the device unless the user explicitly chooses to sync aggregate signals.


Features

  • 100% local inference — bundled llama.cpp + a small Qwen3-VL GGUF model. No cloud roundtrips for sensitive data.
  • Desktop-native — Tauri 2 (Rust) shell, Vite frontend, SQLite for local state. Installs as a single .msi on Windows.
  • Activity-oriented, not surveillance-oriented — the agent surfaces meaningful work units (branches, PRs, focus windows) rather than keystroke counts.
  • Team analytics, with consent — opt-in aggregation into a Supabase backend only for users who join a team.
  • Self-hostable backend — the Community Edition can run against your own Supabase instance.

Status

FlowSight is in active development. Expect breaking changes until v1.0. Track progress on the Releases page.


Quick start

Prerequisites

  • Windows 10/11 (Linux and macOS are on the roadmap).
  • Rust stable (for building the Tauri shell).
  • Node.js 18+ and pnpm 8+.
  • Python 3.11+ (runs the prebuild script that fetches the LLM model).

Install and run

git clone https://github.com/Mancasvel/FlowSight.AI.git
cd FlowSight.AI
pnpm install
pnpm dev

The dev command starts the agent with hot reload. The first run downloads the GGUF model from the repository's GitHub Release (see scripts/fetch-models.mjs).

Build a release installer

pnpm build

The installer lands in apps/agent/src-tauri/target/release/bundle/. It bundles llama-server.exe, the required DLLs, and the GGUF model into the MSI, so the end user does not need any runtime download.


Architecture (short version)

+-------------------------------+       +-----------------------+
|  Tauri agent (Rust)           |       |  Supabase backend     |
|   - OAuth (Google)            |<----->|   - Teams             |
|   - Context capture           |       |   - Aggregated events |
|   - Local LLM (llama.cpp)     |       |   - RLS per team      |
|   - SQLite state              |       +-----------------------+
+---------------+---------------+
                |
                v
     %LOCALAPPDATA%\FlowSight\
     (logs, db, cache — local only)

The heavy lifting (context summarization, PII filtering, intent inference) runs in-process against the local llama-server.exe. Only already-filtered aggregates reach the cloud backend, and only when the user belongs to a team.

Repository layout

apps/
  agent/          Tauri desktop app (Rust + Vite frontend)
  dashboard/      Next.js team dashboard (optional)
local_llm/
  bin/            llama-server.exe + DLLs (committed, ~50 MB)
  *.gguf          Local model weights (fetched at build time, not committed)
scripts/
  fetch-models.mjs  Prebuild hook (Node-only): downloads GGUF from GitHub Releases
.github/
  workflows/      CI (build, release, gitleaks)

Configuration

All user-facing settings live in the desktop app. Local state is persisted under:

  • Windows: %LOCALAPPDATA%\FlowSight\
  • Logs: server.log, agent_error.log, crash.log
  • Database: dev-agent.db (SQLite)

No configuration file is expected on the user's side. The environment variable GITHUB_TOKEN is only needed by developers who want to fetch model assets from a private release.


Join Us

We're hiring our first engineer.

FlowSight is a privacy-first productivity tool that runs locally on your machine. No surveillance, no cloud dependency, no compromise. Backed by Microsoft for Startups, incubated at Xiji (Shanghai), part of AltaLab's accelerator.

The problem we're solving

  • Teams waste hours in meetings that could be a message
  • "Productivity tools" are surveillance with a nicer UI
  • No one trusts their data with the tools they use at work
  • Privacy and good UX shouldn't be mutually exclusive

Who we're looking for

  • You've shipped something end-to-end (side project, startup, open source)
  • You're comfortable with Rust, TypeScript, or systems-level work
  • You understand why privacy-first architecture matters
  • You want to build, not just code tickets

Why FlowSight

  • Equity-first role. Real ownership, not token options
  • Product is live. Not a pitch deck, a working app
  • Technical CEO who built it solo and needs a multiplier
  • Shanghai-based or remote. We care what you build

How to apply

Send an email to [email protected] with:

  • Something you've built that you're proud of
  • Why privacy-first AI matters to you
  • Your GitHub or portfolio

No cover letters. No culture fit essays. Show us what you've done.


License and distribution

FlowSight is distributed under a dual licensing model:

Edition License Intended for
Community GNU AGPL-3.0 Individuals, OSS projects, academic use, internal non-commercial deployments
Enterprise / Commercial Proprietary, per contract Closed-source redistribution, SaaS offerings, OEM, customers whose policies forbid AGPL

TL;DR: you can use, modify and self-host the Community Edition as long as you respect the AGPL — which, crucially, requires you to publish your modifications if you expose them over a network. If you can't live with that, buy a commercial license: [email protected].

Contributing

Contributions are very welcome. Every contributor must sign a CLA (individuals: CLA.md, companies: CLA-CORPORATE.md) so that the project can keep the dual-licensing model working. The CLA Assistant bot handles signatures automatically on your first PR. See CONTRIBUTING.md for the full flow.

Code of Conduct

Participation is governed by the Contributor Covenant.

Security

If you find a vulnerability, please do not open a public issue. Use GitHub's private Security Advisory feature on this repository, or email [email protected].


Support FlowSight

FlowSight is free and open source. If it's useful to you, consider supporting the project:

Every coffee helps keep development going. All funds go directly to:

  • Server costs — CI/CD, model hosting, Supabase backend
  • Model improvements — better local LLMs for activity analysis
  • Cross-platform — Linux and macOS builds
  • New features — team analytics, integrations, plugin system

Sponsor Tiers

Tier Amount Badge
☕ Supporter €1-4 Listed in SPONSORS.md
☕☕ Champion €5-14 Listed + name in app credits
💎 Founding Supporter €15+ Listed + featured in app About page
🔄 Monthly Backer Any recurring All above + early access to features

Other ways to contribute

  • Star this repo — helps with visibility
  • 🐛 Report bugs — open an issue
  • 💻 Submit a PR — see CONTRIBUTING.md
  • 📣 Spread the word — share with your team

Trademarks

"FlowSight" is a trademark of FlowSight. The AGPL license grants you rights to the code but not to the trademark. If you publish a fork, please pick a different name for your distribution.


Links

About

FlowSight AI is a privacy-first developer productivity tool that automatically tracks coding activity, detects blockers, and keeps project managers informed in real-time—without ever storing screenshots or sensitive data.

Resources

License

Code of conduct

Contributing

Stars

Watchers

Forks

Packages

 
 
 

Contributors