Start now without registering at unforget.computing-den.com.
Unforget is a minimalist, offline-first, and end-to-end encrypted note-taking app (without Electron.js) featuring:
- Offline first
- Privacy first
- Progressive web app
- Open source MIT License
- End-to-end encrypted sync
- Desktop, Mobile, Web
- Markdown support
- Self hosted and cloud options
- One-click data export as JSON
- Optional one-click installation
- Public APIs, create your own client
- Import Google Keep
- Import Apple Notes
- Import Standard Notes
Unforget is made by Computing Den, a software company specializing in web technologies.
Contact us at sean@computing-den.com
Sign up for free to back up your notes safely to the cloud fully encrypted and sync across devices.
No email or phone required.
Use it directly in your browser or install:
Browser | Installation |
---|---|
Chrome | Install icon in the URL bar |
Edge | Install icon in the URL bar |
Android Browser | Menu → Add to Home Screen |
Safari Desktop | Share → Add to Dock |
Safari iOS | Share → Add to Home Screen |
Firefox Desktop | cannot install |
Firefox Android | Install icon in the URL bar |
Notes are organized chronologically, with pinned notes displayed at the top.
This organization has proven very effective despite its simplicity. The search is very fast (and done offline), allowing you to quickly narrow down notes by entering a few phrases. Additionally, you can search for non-alphabetic characters, enabling the use of tags such as #idea, #project, #work, #book, etc.
There is no limit to the size of a note. For larger notes, you can insert a ---
on a line by itself to collapse the rest of the note.
Notes are immediately saved as you type and synced every few seconds.
If you edit a note from two devices and a conflict occurs during sync, the most recent edit will take precedence.
Unforget does not receive or store any personal data. No email or phone is required to sign up. As long as you pick a strong password, your notes will be stored in the cloud fully encrypted and safe.
Only your username and note modification dates are visible to Unforget servers.
The main differences with the Github flavored markdown are:
- If the first line of a note is followed by a blank line, it is a H1 header.
- Anything after the first horizontal rule
---
in a note will be hidden and replaced with a "show more" button that will expand the note.
# H1 header
## H2 header
### H3 header
#### H4 header
##### H5 header
###### H6 header
*This is italic.*.
**This is bold.**.
***This is bold and italic.***
~~This is strikethrough~~
- This is a bullet point
- Another bullet point
- Inner bullet point
- [ ] This is a checkbox
And more text related to the checkbox.
1. This is an ordered list item
2. And another one
[this is a link](https://unforget.computing-den.com)
Inline `code` using back-ticks.
Block of code:
```javascript
function plusOne(a) {
return a + 1;
}
```
| Tables | Are | Cool |
| ------------- |:-------------:| -----:|
| col 3 is | right-aligned | $1600 |
| col 2 is | centered | $12 |
Horizontal rule:
---
To build Unforget for production, put a .env
file in the project's root directory:
PORT=3000
NODE_ENV=production
DISABLE_CACHE=0
LOG_TO_CONSOLE=0
FORWARD_LOGS_TO_SERVER=0
FORWARD_ERRORS_TO_SERVER=0
and then run
cd unforget/
npm run build
npm run start
It is recommended to use Nginx as a reverse proxy and set up SSL certificate using Let's Encrypt.
To build and run Unforget in development mode, put a .env
file in the project's root directory:
PORT=3000
NODE_ENV=development
DISABLE_CACHE=1
LOG_TO_CONSOLE=1
FORWARD_LOGS_TO_SERVER=0
FORWARD_ERRORS_TO_SERVER=0
and then run
cd unforget/
npm run dev
This will build the project and watch for changes in the source files.
Here, all paths are relative to either the official server at https://unforget.computing-den.com or your own server if you're self hosting.
In the examples/ directory you will find example code for TypeScript and Python.
To run the Typescript example:
cd examples/
# Signup
npx tsx example.ts signup USERNAME PASSWORD
# Login
npx tsx example.ts login USERNAME PASSWORD
# Create new note
npx tsx example.ts create "Hello world!"
# Get all notes
npx tsx example.ts get
# Get note by ID
npx tsx example.ts get ID
To run the Python example:
cd examples/
# Signup
python3 example.py signup USERNAME PASSWORD
# Login
python3 example.py login USERNAME PASSWORD
# Create new note
python3 example.py create "Hello world!"
# Get all notes
python3 example.py get
# Get note by ID
python3 example.py get ID
type Note = {
// UUID version 4
id: string;
// Deleted notes have text: null
text: string | null;
// ISO 8601 format
creation_date: string;
// ISO 8601 format
modification_date: string;
// 0 means deleted, 1 means not deleted
not_deleted: number;
// 0 means archived, 1 means not archived
not_archived: number;
// 0 means not pinned, 1 means pinned
pinned: number;
// A higher number means higher on the list
// Usually, by default it's milliseconds since the epoch
order: number;
}
type EncryptedNote = {
// UUID version 4
id: string;
// ISO 8601 format
modification_date: string;
// The encrypted Note in base64 format
encrypted_base64: string;
// Initial vector, a random number, that was used for encrypting this specific note
iv: string;
}
The server only knows about EncryptedNote
and never sees the actual Note
. So, the client must encrypt before sending to and decrypt after receiving notes from the server.
Side note: the reason for using number (0 and 1) instead of boolean is to make it easier to store notes in SQLite which doesn't support boolean. And the reason why some of these fields are flipped (not_deleted
instead of deleted
) is to facilitate the use of IndexedDB which doesn't support indexing by multiple keys in arbitrary order.
To sign up, send a POST request to /api/signup
with a JSON payload of type SignupData
:
type SignupData = {
username: string;
password_client_hash: string;
encryption_salt: string;
}
To log in, send a POST request to /api/login
with a JSON payload of type LoginData
:
type LoginData = {
username: string;
password_client_hash: string;
}
In both cases, if the credentials are wrong you will receive a 401 error. Otherwise, the server will respond with LoginResponse
and code 200:
type LoginResponse = {
username: string;
token: string;
encryption_salt: string;
}
To log out, send a POST request to /api/login?token=TOKEN
In the following sections, all the requests to the server must include the token
either as a query parameter in the URL (e.g. /api/delta-sync?token=XXX
) or as a cookie named unforget_token
.
Notice that we never send the raw password to the server. Instead we calculate its hash as password_client_hash
which is derived from the username, password, and a static random number. It is important to use the exact same algorithm for calculating the hash if you want to be able to use the official Unforget client as well as your own. The encryption_salt
is a random number used to derive the key for encryption and decryption of notes. It is stored on the server and provided on login. The examples show how to calculate the hash and generate the salt.
Send a POST request to /api/get-notes?token=TOKEN
to get all notes. Optionally you can provide a JSON payload of type {ids: string[]}
to get specific notes.
You will receive EncryptedNote[]
.
Send a POST request to /api/merge-notes?token=TOKEN
with a JSON payload of type {notes: EncryptedNote[]}
.
If the note doesn't already exist, it will be added.
If its modification_date
is larger than the existing note, it will replace the existing note.
Otherwise, it will be thrown away.
To delete a note set its text: null
and not_deleted: 0
and merge it. This way, the stub will stay in the database and the fact that it was deleted will propogate to all the other clients.
For a long-running client, instead of using Get Notes and Merge Notes, you can use sync in the following manner.
The client and the server each maintain a queue of changes to send to each other as well as a sync number. The exchange of these changes is called a delta sync.
The sync number is 0 at login and is incremented by each side only after all the received changes have been merged and stored. At the start of each delta sync, if their sync numbers differ, it indicates that something went wrong in the last delta sync and so they must do a queue sync.
A queue sync is when each side sends its sync number along with a list of IDs and modification dates of all the notes that it knows about. After a queue sync, both sides will know which changes the other side lacks and therefore can update their own queue and sync number.
When the sync number is 0 (immediately after login), the server will send all notes in the first delta sync.
To perform a delta sync, send a POST request to /api/delta-sync?token=TOKEN
with a JSON payload of type SyncData
:
type SyncData = {
notes: EncryptedNote[];
syncNumber: number;
}
If the server agrees with the syncNumber
, it will respond with DeltaSyncResNormal
which includes the changes stored on the server for that client since the last sync. Otherwise, the server will respond with PartialSyncResRequireQueueSync
requiring the client to initiate a queue sync.
type DeltaSyncResNormal = {
type: 'ok';
notes: EncryptedNote[];
syncNumber: number;
}
type DeltaSyncResRequireQueueSync = {
type: 'require_queue_sync';
}
To perform a queue sync, send a POST request to /api/queue-sync?token=TOKEN
with a JSON payload of type SyncHeadsData
including the heads of all the notes known by the client and its sync number. You will then receive another SyncHeadsData
including the heads of all the notes known by the server for that user along with the server's sync number for that client.
type SyncHeadsData = {
noteHeads: NoteHead[];
syncNumber: number;
}
type NoteHead = {
id: string;
modification_date: string;
}
After a queue sync, each side updates its queue to include the changes the other side is mising as well as setting the new sync number to be the larger sync number + 1.
It is important that the client and the server agree on how the merging of the notes is done so that they end up with a consistent state. We say that note A must replace note B if A.id == B.id
and A.modification_date > B.modification_date
.
The details of encryption and decryption are more easily explained in code. See the Examples section.
All the API calls will return an object of type ServerError
when encountering an error with a status code >= 400:
type ServerError {
message: string;
code: number;
type: 'app_requires_update' | 'generic';
}
If you receive an error with type app_requires_update
that indicates that you are using an older version of the API that is no longer supported.