Dazzi

Open source alternative to Rewind AI, DEVONthink and Fenn

13 stars
FSL-1.1-MIT
Active this week
Last commit today

Last verified Nov 5, 2026

Search local files, chats, and mail by asking questions in plain language, powered by a local embedding model that keeps your data on-device.

Visit website GitHub13

Read the full breakdownWho it is for, how it works, trade-offs, the quick start and the FAQ

who it's for

Who Dazzi is for#

Knowledge workers managing large file archives

Researchers, analysts, and consultants who build up thousands of documents over years can query across all of them in one prompt. The citation links let you verify every answer against the specific source passage, rather than opening files one by one.

Skip if:

If your files are mainly structured data like spreadsheets, databases, or CSVs, Dazzi's document search adds little. SQL or a spreadsheet formula will be faster.

Privacy-conscious users replacing Rewind AI or DEVONthink

Anyone who finds Rewind AI's continuous screen recording too invasive, or who wants a cross-platform alternative to DEVONthink's macOS-only ecosystem, can get semantic document search without sending files to an external service.

Skip if:

If you need Rewind AI's screen and audio capture specifically, to recall a website you visited or what was said in a meeting, Dazzi does not record the screen and cannot substitute for that.

Teams searching across multiple messaging apps

When conversations span Slack, Telegram, WhatsApp, and email, context is scattered. Dazzi connects all of them and searches across threads, channels, and inboxes in one query without switching apps or logging in and out.

Skip if:

If your team uses only one messaging app and already has strong built-in search, adding Dazzi is unnecessary overhead.

the problem

The problem it solves#

Finding specific information across your own computer gets harder as the number of files, folders, and connected apps grows. Standard file search matches filenames and metadata, but most useful context is buried inside documents, including a contract clause, a decision from a Slack thread three months ago, or a note in a PDF scanned last year. Switching between apps to track down information wastes time and breaks focus.

The challenge compounds when files arrive in different formats from different sources. PDFs, Word documents, markdown notes, chat exports, and email all live in separate places and require separate search tools. Most people end up with rough personal filing systems that work until volume or variety increases enough to overwhelm them.

how Dazzi solves it

How it solves it#

Local AI Search

Asks questions across the full contents of files, mail, and connected chat apps in plain English. The default embedding model (Qwen3-Embedding-0.6B) runs on-device, so documents do not leave the machine. Every answer links back to the source passages it used, with citation numbers in the response.

Connected App Search

Searches Slack channels and threads, Discord servers, Telegram chats, WhatsApp messages, GitHub repos and issues, Obsidian vaults, and mail inboxes alongside local files. One search across everything you work in, not just the filesystem.

Version Grouping

Groups duplicate and near-duplicate files into single rows in the file list, with the newest on top labeled Latest and similar variants labeled Similar or Identical. Cuts through the proliferation of files named proposal-final, proposal-v3, and proposal (1). Powered by the dupey library.

Quick Look Preview

Press Space on any file to preview it inline. Markdown, code, PDF, video, audio, Office documents, and HWP/HWPX files render immediately without opening another app. HWP and HWPX pages render locally without Hancom Office installed.

Folder Automations

Create folder-level rules that run when documents arrive or change, such as checking a PDF for a signature and filing it automatically. Automations trigger without the app being open, so a new document lands, gets checked, and is filed while you are away.

strengths · trade-offs

Strengths and trade-offs#

Strengths

Privacy-first local embeddings by defaultThe default embedding model (Qwen3-Embedding-0.6B) runs on-device. File contents stay on the machine unless the user explicitly opts into cloud embedding on a paid plan. No documents are uploaded without consent, and no content is sent to train a third-party model.
Cross-platform macOS and Windows supportInstalls from a single signed package on both Apple Silicon Macs (arm64 DMG) and Windows x64 machines. Each installer bundles the embedding model and every dependency so the first launch is the full app, with nothing extra to download.
Source-available with a clear MIT conversion pathFSL-1.1-MIT means every release converts to full MIT two years after publication. You can read the source, file issues, and build on it. The underlying AutoRAG engine is Apache-2.0 licensed, and the full codebase becomes OSI-compliant open source over time.
Broad file format support in Quick LookRenders PDF, Office documents, markdown, code, video, audio, and HWP/HWPX natively. Korean HWP pages render locally without Hancom Office, which is useful for anyone working with Korean government or enterprise documents.

Trade-offs

-FSL-1.1-MIT is not an OSI-approved open source licenseDazzi is Fair Source, not fully open source. You can use, modify, and self-host it for any purpose, including internal business and commercial team use. What is not permitted: offering the software (or software with substantially similar functionality) as a commercial product or service to others. Each release converts to MIT after two years, so the restriction applies only to recent versions.
-Early-stage project with 13 stars and 44 open issuesThe repo was created in September 2026 and has 13 stars at the time of listing. With 44 open issues logged, users should expect rough edges, missing documentation, and breaking changes between releases. It is not a good fit for a team production workflow that requires stable, well-tested releases.
-macOS and Windows only; no Linux buildsThe installer ships macOS arm64 and Windows x64 builds only. Linux users cannot install the signed release without building from source, which requires Bun, Node.js 24+, and the Cargo toolchain for the dupey dependency.

versus alternatives

Dazzi vs alternatives#

Dazzi vs Rewind AI

Both tools help you find information stored on your computer, but they index it differently. Rewind AI records your screen and audio continuously; Dazzi indexes the contents of files and connected apps on demand.

Feature	Dazzi	Rewind AI
License	FSL-1.1-MIT (Fair Source)	Proprietary
Self-hosting	Yes (desktop installer)	No
Indexing method	File and app contents	Continuous screen recording
Platforms	macOS, Windows	macOS only
Local AI processing	Yes (default)	No
Pricing	Free tier; paid from $4/mo	Subscription required

Dazzi is the better choice when privacy matters. Documents do not leave your machine by default, and there is no continuous screen capture. It also covers Windows users and connects to chat apps and GitHub for cross-service search. Rewind AI has an advantage for recalling anything you have seen on screen, including web pages, meetings, and apps Dazzi does not index; if screen-and-audio replay is what you need, Dazzi is not a substitute.

Dazzi vs DEVONthink

DEVONthink is a macOS document manager with an extensive set of organizational features built up over two decades. Dazzi is a cross-platform AI search tool that prioritizes querying over organizing.

Feature	Dazzi	DEVONthink
License	FSL-1.1-MIT (Fair Source)	Proprietary
Platforms	macOS, Windows	macOS, iOS only
AI search	Yes (local or cloud)	Proprietary feature set
Cross-service search	Files, mail, chat apps	Local files and databases
Pricing	Free tier; from $4/mo	Paid, proprietary

Dazzi wins for cross-platform coverage, chat app integration, and users who do not want to invest in DEVONthink's organizational model. DEVONthink is still the better fit for macOS-only teams with large document collections who need DEVONthink's classification, smart groups, and deep macOS integration, as well as its track record of stability over many years.

install · quick start

Quick start#

bash

Download the signed macOS installer or Windows installer from GitHub Releases: https://github.com/NomaDamas/AutoRAG-app/releases/latest

tech stack · detected from GitHub

What it's built on#

Languages: JavaScriptTypeScript
Frameworks: ElectronReact

frequently asked

FAQ#

Is Dazzi free to use?

There is a free tier that includes chat, embeddings, reranking, and a small credit allowance. Paid plans start at Air ($4/month), which includes a larger credit allowance and the option to use an API key from an existing AI subscription. ProMax ($20/month) and Max ($200/month) add progressively larger credit allowances. You can also connect your own API key directly in the app settings on any plan.

Does Dazzi send my files to the cloud?

By default, the embedding model (Qwen3-Embedding-0.6B) runs locally on your machine and your documents are not uploaded. File contents are sent to a remote embedder only if you choose Dazzi's cloud embedding option, which requires a paid plan and explicit consent in the app. Sign-in tokens and API keys live in the system's encrypted store, not in a config file.

What does the FSL-1.1-MIT license allow, and what does it prohibit?

FSL-1.1-MIT (Functional Source License, Version 1.1) is a Fair Source license. You can use, modify, and self-host Dazzi for any purpose, including running it internally at your business or for commercial team use. What is prohibited: offering Dazzi (or software with substantially similar functionality) as a commercial product or service to others. This restriction applies only until two years after a release, at which point that release converts to full MIT with no restrictions.

Does Dazzi support Windows?

Yes. Dazzi ships a signed Windows x64 installer alongside the macOS arm64 DMG. Both installers bundle the app and the local embedding model, so there is nothing extra to download. Windows support is part of the initial release.

What file formats does Quick Look preview?

Markdown, code files, PDF, video, audio, Office documents (.docx, .xlsx), and HWP/HWPX files all preview with the Space bar. HWP and HWPX pages render locally without Hancom Office installed. The preview panel opens instantly beside the file list without launching a separate application.

also worth a look

Similar open-source tools#

OpenClaw

Featured

Alternative toMuse

Self-hosted AI assistant for WhatsApp, Telegram, and more

393,917 GitHub stars
Licence MIT

Inbox Zero

Featured

Alternative toSuperhuman

AI email assistant that drafts replies and clears your inbox

12,889 GitHub stars
Licence AGPL-3.0

rakazo

Featured

Alternative toGrok Bot

AI teammates you own: your keys, your model, your machine.

7,028 GitHub stars
Licence Apache-2.0

paperclip

Featured

Alternative toGrok Bot

Self-hosted AI agent management with org charts and budgets

111,113 GitHub stars
Licence MIT

Teable

Alternative toAirtable

Spreadsheet-simple database and AI app builder on PostgreSQL

21,953 GitHub stars
Licence AGPL-3.0 with brand protection terms

Metabase

Alternative toTableau

Self-hosted BI with dashboards, SQL editor, and AI-native analytics.

50,133 GitHub stars

by category

Open source tools in these categories#

472

12,889 GitHub stars
Licence AGPL-3.0
#Email

rakazo

Alternative toGrok Bot

AI teammates you own: your keys, your model, your machine.

7,028 GitHub stars
Licence Apache-2.0
#AI Agents

SnitchFeed

Live monitoring on LinkedIn, X, Reddit & more. Agentic social listening for startups.

Marketing & Growth

Advertise your product

The projects we add and every new guide. Free, unsubscribe any time.

Find open source alternatives to paid software

Dazzi vs Rewind AI

Feature	Dazzi	Rewind AI
License	FSL-1.1-MIT (Fair Source)	Proprietary
Self-hosting	Yes (desktop installer)	No
Indexing method	File and app contents	Continuous screen recording
Platforms	macOS, Windows	macOS only
Local AI processing	Yes (default)	No
Pricing	Free tier; paid from $4/mo	Subscription required

Dazzi vs DEVONthink

DEVONthink is a macOS document manager with an extensive set of organizational features built up over two decades. Dazzi is a cross-platform AI search tool that prioritizes querying over organizing.

Feature	Dazzi	DEVONthink
License	FSL-1.1-MIT (Fair Source)	Proprietary
Platforms	macOS, Windows	macOS, iOS only
AI search	Yes (local or cloud)	Proprietary feature set
Cross-service search	Files, mail, chat apps	Local files and databases
Pricing	Free tier; from $4/mo	Paid, proprietary

Who Dazzi is for#

Knowledge workers managing large file archives

Skip if:

Privacy-conscious users replacing Rewind AI or DEVONthink

Skip if:

Teams searching across multiple messaging apps

Skip if:

The problem it solves#

How it solves it#

Local AI Search

Connected App Search

Version Grouping

Quick Look Preview

Folder Automations

Strengths and trade-offs#

Strengths

Trade-offs

Dazzi vs alternatives#

Dazzi vs Rewind AI

Dazzi vs DEVONthink

Quick start#

What it's built on#

FAQ#

Similar open-source tools#

Open source tools in these categories#

AI & Machine Learning

Business & Productivity

Find open source alternatives to paid software

Who Dazzi is for#

Knowledge workers managing large file archives

Skip if:

Privacy-conscious users replacing Rewind AI or DEVONthink

Skip if:

Teams searching across multiple messaging apps

Skip if:

The problem it solves#

How it solves it#

Local AI Search

Connected App Search

Version Grouping

Quick Look Preview

Folder Automations

Strengths and trade-offs#

Strengths

Trade-offs

Dazzi vs alternatives#

Dazzi vs Rewind AI

Dazzi vs DEVONthink

Quick start#

What it's built on#

FAQ#

Similar open-source tools#

Open source tools in these categories#

AI & Machine Learning

Business & Productivity

Find open source alternatives to paid software