repos / pico

chore: caddy logs to pipe

Don't require space

Set bouncer for oauth

Change auth endpoints to check plus and space

chore: go tidy

Only insert visit if clean

Add user to analytics entries

No read timeout when reading metrics

Fix metric drain

Bump utils

Use shared lib for metrics connection

Cleanup and standardize user id for logs

Updates utils

Update to add read and write timeouts

Update to use reconnect logger

Continue on err for analytics

Add buffering to analytics collection and make it non-blocking

Update utils

Update pipe logger and other utils

fix(pgs): trusted_proxies

chore(pgs,prose): caddy access logs

fix: ip address typo

feat(auth): subscribe to pico's metric-drain pipe

We have a lot of services that need to record site usage analytics so we
need a distributed way to receive these events.

As a result, all of our site visits will be sent through our pipe
service.  Then inside of our auth service we process the visits and
store them in our analytics table.

This overloads the auth service since it's now serving as a destination
for our metric-drain, but we think that's okay for now.

fix(pgs): caddy

fix(pgs): trust upstream reverse proxies and forward client ip

refactor(pgs): use pobj/storage

fix(pgs): set content-length so caddy can compress content

Allow unauthenticated access to public pipes (#157)


Co-authored-by: Eric Bower <me@erock.io>

fix(bouncer): previous alpine docker was failing to build

chore: 20gb to 10gb for pico+

fix(pipe): remove csp

chore(analytics): add indexes to table

docs: readme

chore(pgs): move fns around

refactor(pgs): move filehandler to pgs folder

fix(pgs): root domain routes were broken

refactor(pgs): use `http.ServeMux` v1.22

This change futher separates `pgs` from most of the other services.

One important change is refactoring our web router to use the new
`http.ServeMux` router that does a better job of routing with templated
strings.

Further, we try to leverage the `http.Handler` interface that removes
the need to use `context.Context` as much.

Decrease loop iteration time for pubsub

chore(pipe): update cicd copy

chore(pgs): print 4 decimal places special files

chore(pgs): limit size of _redirects and _headers files

chore(pgs): api tests

chore(pgs): api tests

chore(pgs): integration tests for web api (#152)

chore: monthly active users script

fix(auth): associate payment history and `plus` feature flag

fix: check if user is nil when uploading files

refactor(pgs): use httputil rev proxy to external service

refactor(feeds): only show banner notice when <30 days

feat(feeds): extend engagement click once every 6 months

fix(pgs): various _redirects fixes with wildcards and splats

fix: linter

Fix silent failure on invalid redirect syntax (#148)

Add logger and prom to imgs

Merge pull request #147 from picosh/am/light-refactor

Refactor pubsub->pipe

Update caddyfiles

Refactor pubsub->pipe

Refactor shared utilities

docs(pubsub): copy

Add public key comment to pubsub ls

docs(pubsub): copy

chore(pgs): block site access when project is blocked

fix(pico): reject external service redirects for non-paid users

chore(pubsub): update asciicast

chore: changes asciicast

chore: blah

docs(pubsub): copy

style(pubsub): make it pop

chore(pubsub): print `-p` flag for pub and pipe stdout msg

Merge pull request #145 from picosh/am/lib-refactor

Refactor

Refactor

docs(pubsub): web copy

docs(pubsub): copy

fix(pico): cleanup logs when leaving page

chore(pico): logs work

fix(pico): cap log lines to 1k

chore(pastes): better logs

chore(pastes): better api logs

feat(pgs): add cache headers etag and last-modified

feat(tui): logs

Fix leaky logs

chore: log formatting

feat(pico): logs ssh command

Change to log drain

chore: logs

Set proper profile setting

Fix incomplete messages

fix(pgs): log

fix(pgs): log

Updated the output

chore(pgs): better structured logging

Merge pull request #143 from picosh/am/send-log-sink

Setup a send log sink

Fix write lock

Add configuration options

Make sure to build pico app

Remove log dep and add data race safe impl

Use once to close channels

Enable robust log handler

Setup a send log sink

style(pastes): smol-v2 styles

Fix port setting

Update help command output

chore(pubsub): update pubsub dep

Simplify ls

Update go version and update pubsub features

Pubsub refactor

style(prose): moved tags block to bottom of page

feat(prose): `with_styles` setting in `_readme`

This adds a new frontmatter property in `_readme` that allows users to
request no styles from us.  This allows for better customization in
conjunction with `_styles.css` which is the user-defined CSS stylesheet
that gets loaded.

refactor(pubsub): default timeout set to 30 days

feat(pubsub): if no channel is provided, randomly gen one

chore(pubsub): pub log when timeout reached

feat(pubsub): pub new flag, `-t {duration}` timeout

The default is no timeout.

docs(pubsub): update copy

fix(feeds): days until expiration date

fix(feeds): flip post expiration check

fix(auth): docker tag

fix(feeds): url

feat(feeds): require human interaction to continue sending emails

BREAKING CHANGE: we require users to click a link in their feed emails
once every 90 days

fix(prose): undo footer box for posts

chore(feeds): more logging cleanup

chore: structured logging for servier and user info

Update some page options

docs(pubsub): add our logo

docs(pubsub): reorg content

Add gifs to marketing page for send

Added pipe support to send

Merge pull request #140 from picosh/am/pubsub-clean

Alternate pubsub implementation

Add more caching to builds

Added admin listing

Updated pubsub ref

chore(prose): revert diff code

fix(pgs): redirects

Update dep

Update cli

Alternate pubsub implementation

docs(pubsub): copy

docs(pubsub): copy

fix(pubsub): proper bidirectional blocking

docs(pubsub): web copy

chore: cleanup

feat(pubsub): send messages over public channels

feat(pubsub): flag for empty messages

docs: copy

fix(pubsub): scope `ls` to ssh user

fix(pgs): rss feed

fix(pubsub): web

fix(pubsub): create channel for multicast

feat(pubsub): pub waits for a sub

chore: cleanup

fix(pubsub): compose file

chore(pubsub): Caddy use basic reverse proxy

chore(pubsub): docker compose config

feat: pubsub service

feat(imgs): pubsub

Send keys for user info

Add user handler to auth system

refactor(pgs): remove CSP and allow headers to be overriden

chore: fix linter

chore: cleanup go.mod

chore: update send pkg

feat(pgs): block uploads for specific projects

chore(pgs): add projects.blocked col

Sometimes users have sites that violate our ToS.  We can delete the
content but if they have a script that will reupload it will
automatically recreate the site.  So we need a way to prevent uploads to
projects.

chore(bouncer): remove debug flag

fix(prose): add rss route back

chore(bouncer): add `-debug` flag to soju

chore(bouncer): update patch against soju upstream

chore: diff in post

feat(prose): live journal

The goal is to be able to use posts as live journals where reader can
receive updates to docs via rss.

chore(pgs): marketing page copy

chore: rm lists.sh

chore: test names

fix(tui): force minimum color profile

docs: dev doc

feat(pgs): rewrite support

feat(pgs): rewrite proxy api support

fix(bouncer): update go to 1.22

chore(bouncer): disable inactive users after 30 days

See soju docs to read more: https://soju.im/doc/soju.1.html

chore(bouncer): switch to codeberg for upstream remote

chore: update send and pobj

fix: cleanup go.mod

chore: ptun => tunkit

use local

chore: keep pgs ssh pprof

fix: use file size if available when uploading

chore: update `send`

This fixes a memory leak in our ssh apps.

style: update smol

style: tweaks

style: blockquote for post description

style(prose): adjust spacing

style: mv anchor link to end of title

style: use `time.DateOnly` format for dates

style: update smol.css

fix(pgs): when removing project, always try to remove objects

refactor(pgs): use pubkey auth handler

refactor: be pickier with which pubkey to accept

chore: formatting analytics

feat(prose): cli cmd `stats {post}` for analytics

feat(pgs): cli cmd `stats {project}` for project analytics

chore(pgs): better error message for uploads

feat(pgs): change max file size for free tier 1->10mb

fix: use official send pkg

refactor(pgs): dont copy reader into memory

chore(pgs): set memory limit for ssh container

feat(imgs): pubsub for docker image tag updates

Updated go.mod to remove replacement

chore(pastes): set cron to once every 12h

chore(pastes): show unlisted tag

Merge pull request #136 from picosh/am/update-deps

Update dependencies (for chroma)

Fix lint

Update dependencies (for chroma

Set ip correctly based on prod-irc as well

chore(plus): change payment history amount to $24

Speed up caddy build

Merge pull request #134 from picosh/am/sshfs

Added support for sshfs

Fix lint

Added support for sshfs

fix(imgs): user can be nil

refactor: remove ability to override CSP for pgs subdomains

BREAKING CHANGE: cannot override default CSP for pgs

fix: linter

fix: denylist for discovery feed to remove spam

chore: use consistent date fmt

styles(tui): settings page suggested width

feat(tui): settings page

styles(tui): md word wrapping

styles(tui): create account page

docs(tui): copy

styles(tui): update input placeholder

fix(tui): update models when resizing

chore(tui): cleanup

chore: cleanup

style: page titles

refactor(tui): reorganize into pages

chore(pgs): set default height to 24

chore: use `MakeOptions`

refactor: use bubbletea's renderer fn

refactor: tui pages

styles: cleanup tui colors

ascii logo

chore(pico): better logging statements

chore: copy

chore: update storage quotas from 1024 -> 1000

docs: copy

refactor(plus): make plus a single feature (#128)

fix(lint): comment

chore: url shortener for payment processor checkout url

fix: rm `links` from `registry` service

minio and registry are not on the same VM so it was failing when running
docker compose commands

fix(pico): check for user in new tui pages

refactor(plus): use lemonsqueezy

feat(plus): support international users

docs(pico): copy

style: update smol.css

chore: cleanup

chore: tui fixes

chore: new screen

style: blog posts spacing

style: light theme colors

style: code headers

chore: update smol.css

chore: update deps

fix(prose): allow .ico to be uploaded

Closes: #131

feat: senpai integration (#129)

Co-authored-by: Antonio Mika <antoniomika@gmail.com>

fix(tui): don't delete account when active key is deleted (#130)

feat: update pubkey comments w authorized_keys

chore: better output when uploading pubkeys

fix: users not being able to register an account

feat: pico ssh app (#126)

docs: showcase some features

chore: readme

docs: readme

docs: fmt

chore: docs

chore: doc links

chore: readme

chore: readme

chore: update readme

fix(pgs): analytics for private sites

chore: update env.example to include registry url

feat(imgs): ssh cli with ls and rm commands

chore: downgrade go version requirements

chore: cleanup pgs cli

fix(imgs): need allowed ext

chore: TUI only accessible via pico.sh

BREAKING CHANGE: TUI only accessible via pico.sh

Readme: Bump required version to 1.22.0 (#121)

fix: linter

feat(pgs): override denylist

refactor(prose): discover page sorts by publish_at

chore: `lint-dev` => `lint`

chore(pgs): add better err messages

refactor(pgs): switch from allowlist to denylist for files

fix(pgs): add trailing slash even with extension

feat(pgs): support serving files without extensions

feat(pgs): append trailing slash to redirect matchers

This change will make it so a user can provide -- or not provide -- a
trailing slash for matching a route, that way the user doesn't need to
think about it.

refactor(pgs): redirects and rewrites

feat(pgs): 301 redirect urls that are directories without trailing slash (#116)

feat(pgs): support yml, eot, and LICENSE

Closes: #118

fix(analytics): better analytic queries

chore(feeds): cleanup

refactor(feeds): change digest var name

fix(analytics): better first day of month calc

feat(pgs): add opml to allowlist

fix(ui): scope where statement properly

fix(pgs): use full filepath when checking for `.well-known`

Closes: #113

docs(pgs): copy

feat(ui): ability to enable/disable analytics

chore(ui): more analytics API endpoints

feat(pgs): allow csp to be overriden by upstream

fix(prose): add smol.css route

fix(prose): copy smol.css from pico.sh

fix(prose): use relative _styles.css

Closes: #112

docs(pgs): copy

refactor(analytics): cleanup naming and fn params

chore(analytics): hmac ip address

feat: analytics for visitors on prose and pgs (#111)

chore(ui): cleanup api

chore(ui): authorization checks

fix

chore: add pico+ to latest expiration date instead of now

When a user eventually purchases another year of pico+, we don't want to
set the expiration date a year from purchase date.  Rather, we want to
set the expiration date to the expiration date of the feature + 1
year.

chore: change toc to Table of Contents

cleanup

md/toc: make it nice, flexible, and fast

cleanup

feat(prose): toc

Merge branch 'tags'

refactor(prose): dedupe string slice for embedded tags

feat(pgs): support `.xsl`

Closes: https://github.com/picosh/pico/issues/105

chore: upgrade pgit

chore: fmt

fix(prose): use image mimetype if file ext provided

Closes: #104

docs(pgs): change free tier limit

chore(plus): increase imgs quota

fix(ui): always fetch user from ctx

When a user registers for an account, we set the user ctx, but all the
api endpoint are using an closured version of the user.

fix(ui): set user ctx after registering

feat(auth): stripe webhook integration (#101)

fix(prose): RSS content-type with charset=utf-8 (#103)

feat(prose): CSS var --body-max-width (#102)

chore(auth): pico+ activated rss feed item

chore(pgs): reduce free trial to 5MB

chore(imgs): cors headers

chore: update ptun

fix(imgs): registry needs imgs network

fix(imgs): need tmp file

fix

chore(imgs): move ssh app into pico

feat(pgs): impersonation working

feat(pgs): impersonate any user

chore(ui): objects postfix

chore(pgs): add id to objects

feat(ui): api for fetching project objects

docs(pgs): use port `1337` as canonical port for web tunnels

chore(prose): use middle dot to separate date and link

chore(prose): rm "on" from post date line

A small gesture to make prose more multi-lingual

fix(ui): posts should fetch all posts

chore(ui): change post apis

chore: cleanup

chore: rename plus folder to ui

feat: add public_keys.name column

feat(ui): more api endpoints!

chore(pgs): api cleanup

fix(ui): don't bomb on rss token missing

feat(pgs): allow cors for private sites

chore(pgs): rm cors

fix(pgs): caddy

chore(pgs): caddy

fix: caddy

chore: caddy

fmt

chore(pgs): caddy

revert

custom caddyfile for pgs

feat(pgs): pico-ui

mdparser: extract hashtag detection into a func

Also fix a bug where `tags: []` would still do the detection.

mdparser: implicitly capture hashtags in document

When they're found, they will be rendered as so:
`<span class="hashtag">...</span>`.

They are ignored in the following:
* block code sections
* block quote sections
* span code sections

This list can be expanded.

We then collect all of the matching hashtags and provide them as
metadata, but only if there is no explicit frontmatter for them.
So a workaround for undesirable hashtag detection is to use front-matter
to specify `tags: []` or similar.

mdparser: extract ast manipulation into shared fn (#96)

chore: update user to `pico`

prose: infer title from first <h1> before any <p> (#93)

This eliminates the need to use explicit title metadata for the vast
majority of posts.

This still allows explicitly setting a title via the frontmatter.
If neither is found, the current logic goes on.

fix(prose): rss link should be to website

Closes: #94

feat(pgs): add `.wasm` to allowlist

feat(pgs): support more file formats

fix(pgs): redirect to full url

fix(prose): better front-matter error handling on file upload

chore(pgs): log requests

fix(pgs): various redirect fixes (#92)

Closes: #91

fix(pgs): sort rss feed

feat(pgs): add new updated_at rss feed

fix: pico-plus script args

feat(prose): let user customize 404 page

feat(prose): `_404.md` file now supported

chore(pgs): docker compose tmp folder

fix(scripts): make pico-plus script more flexible

chore: bump docker go version

refactor: shared router

refactor(shared): http router

docs: cleanup

chore: pico+ go script

fix: empty username

refactor: ssh styles
feat: add pico+ info

chore: rm TERM

chore: add ico to imgs

fix: pin charm libs

Fix environment colors

feat(pgs): support avif and heif image formats

docs(pgs): private projects

docs(pgs): marketing

use v3

chore: use v1 gha

fix write flag

fixes

refactor(pgs): new cli api

design(pgs): cli aesthetic

fix(feeds): name is nullable

feat(pgs): _headers file (#83)

https://docs.netlify.com/routing/headers/

Closes: #82

chore(pgs): cleanup logs

chore: update ptun

fix(router): pass writer to create serve basic

chore(pgs): acl for public sites

chore(pgs): web tunnel cleanup

feat(pgs): private projects with web tunnels (#80)

We can support private projects via local forward ssh tunnels.

Eric sets ACL for project `hey-docs-dev`:

```
ssh pgs.sh acl hey-docs-dev pico antonio
```

Antonio creates SSH tunnel to project:

```
ssh -L 5000:localhost:80 -N hey-docs-dev@pgs.sh
```

Antonio can access site via http://localhost:5000

Other options we support:

- `acl public` -> default, open to everyone
- `acl pico` -> private to any pico user
- `acl pico xxx,yyy` -> private to specific pico users
- `acl public_keys` -> private to anyone with a pubkey
- `acl public_keys sha256:xxx,sha256:yyy` -> private to specific pubkeys

Further, there is a special API endpoint for any sites accessed via a
tunnel: GET http://localhost:5000/pico

This will return information about the user accessing the site, like
pico username, public key used, etc.

refactor: make public keys unique (#81)

BREAKING CHANGE: we will no longer support multiple account with the
same public key

docs(pgs): copy

fix: ssh cms

chore: fmt

chore: rm set username

chore: rm set username

docs(pgs): marketing copy

refactor: make pgs the CMS for pico (#79)

feat(auth): pico+ rss feed

fix: pico+ sql

fix: new user error

chore: update pico+ sql

chore: use slog wish middleware logger

chore: add `GetObjectSize`

refactor: use pobj (#76)

refactor: replace zap with slog (#75)

chore: remove lists project

fix(prose): allow `_styles.css` to be uploaded again

Closes #74

chore(imgs): dont run ssh

chore: update auth and imgs services

feat(auth): privileged access based on `auth` feature flag

refactor(auth): `/key?space=xxx` instead of inside POST body

feat(auth): `/key` support to check any space

docs: copy changes

fix(pgs): correct content type

chore(prose): set img proc defaults

feat: support more imgproxy options (#73)

New options in any combinations:

- `/s:w:h`
- `/q:1-100`
- `/rt:angle`

We also still support our imgs api version of sizing.

fix(imgs): redirect to prose.sh

chore(imgs): remove unused code

feat(prose): directory of images at `x.prose.sh/i`

refactor(imgs): merge into prose.sh (#71)

feat(pgs): use image proxy for images

BREAKING CHANGE: removed imgs.sh blog features

fix: go fmt

fix(pgs): classify xml when calc routes

`.xml` files were returning as `text/plain` which has different routing
rules than what xml should have.

Closes https://github.com/picosh/pico/issues/72

chore(pgs): increase storage max to 25MB

fix: default size limit fallbacks

pico pro (#69)

Update irc configs

chore: allow us to use sojuctl (#70)

chore: add feeds cmd for testing fetcher

fix(lists): update shutdown date

fix(lists): add real blog post

shutdown notice

fix(feeds): move cron to web app

Our `Dockerfile` doesn't move template files into their respective
images when it's an SSH app.  feeds' cron requires those template files
in order to send emails.

revert: imgs ssh

Use new httpmux config option and explicitly set max users

refactor(imgs): use prose ssh app

Prose's ssh app does the same thing as imgs so there is no point in
having them both.  This frees up an IP address for us to use for other
purposes.

refactor(feeds): cron to run on ssh server

style(pastes): add `.post-date` css

chore: force pgit to update

chore: update license with company name

Fix soju config

Update subscriber change for soju

Add prometheus and identd

style: use docs smol.css as canonical css framework

chore: update service intros

refactor: move ssh services into service packages

The only excpetion is imgs.sh which contains a circular dep.

docs: readme

docs: tweak footers

Refactor docs (#68)

Removed all docs associated with our services and moved them to a centralized source: https://pico.sh

doc suggestion from IRC #pico.sh @ libera.chat (#67)

[10:25am] pareidoliac: I have had problem with the docs apart from the minor detail that the "-> yourusername"
          should be "-> {username}-myproject" in the https://pgs.sh/help#custom-domain section.
[10:25am] pareidoliac: thinking of making a PR to suggest that
[10:29am] pareidoliac: also suggest mentioning the CSP policy pico.sh vs custom domain in the https://pgs.sh/help#spa section
[10:44am] erock: happy to accept PRs

Co-authored-by: jungle <jungle@night.local>

docs(pastes): showcase rsync support

docs(pastes): showcase sftp support

chore(prose): support trailing "/" for post raw

fix(prose): support trailing "/" in url

Closes: https://github.com/picosh/pico/issues/64

fix(imgs): dont clip ext unless actually file ext

Closes: https://github.com/picosh/pico/issues/65

docs(prose): prefer rsync

docs(prose): promote `sftp`

chore(deps): tidy

chore: update golang.org/x/crypto (#63)

Reference: https://groups.google.com/g/golang-announce/c/-n5WqVC18LQ

refactor: move `wish/*` to its own repo (#62)

typo: help.page.tmpl (#61)

Don't send notifications when self

Fix bug in contains

Update soju patch to include selective message notifications

fix(pgs): period in route redirects to 404 (#59)

Closes #58

chore: add tokens view to pgs cms (#60)

Push network id with broadcasted message in soju

Enable panic recovery in ssh and set logger to be based on debug

Redirect imgproxy root

Add metrics for imgproxy

Set proper route ordering

Fixed endpoint for caddy metrics

Update our caddyfiles for imgproxy

refactor(imgs): use imgproxy instead of libwebp (#56)

---------

Co-authored-by: Antonio Mika <antoniomika@gmail.com>

chore: go deps

refactor(imgs): dont bring libwebp into every service (#55)

Most of our libwebp related code lived inside the `shared` package.  This is
problematic since every service imports that package, thereby requiring
libwebp for every service.

Now references to libwebp live solely inside `imgs` library.

Unfortunately, all of our post-based SSH services import `imgs` so they
still have the requirement.

Merge pull request #53 from picosh/post-filesize

post filesize script

Merge pull request #54 from picosh/am/dynamic-list

Add the ability to set the per page by window height

Add the ability to set the per page by window height

feat(prose): render `unlisted` on post when hidden

feat(pastes): set custom expiration and set paste to unlisted (#52)

fix(pgs): load created_at and updated_at for projects

chore: `make scripts` command

skip imgs

ready

post filesize

chore(script): clean write env var

fix: lint

chore(pgs): rm objects from store when project no longer exists

fix(pgs): missing project name when deleting project files from store

Fix panic in toupper

Add feature flag for advanced pastes management

fix(gha): copy contents not dir itself

chore: use latest pgs-action

Report same issues on receive

Added the tuns caddyfile

Put key auth behind feature flag

Add key endpoint for sish

Set correct updated at when creating a post

Don't use now for updates

Properly check mod time as well for updating a post

Use modtime from client, not now

Properly update post time when updated

Skip empty files in rsync

Change to 1.21 for static site builder

Merge pull request #51 from picosh/am/rsync-updates

Am/rsync updates

Ignore error on status messages

Properly close connection for unsupported flags

Ignore error for chtimes as nop

Update go-rsync dep

More work to properly support rsync

Working on adding more features to rsync

Improve pgs filesystem support (and fix local storage)

Fix sftp

Only build bouncer and caddy when changes are made to their contexts

Set auth tag properly

Merge pull request #50 from picosh/am/update-sftp

Fixed issues with sftp file listing and downloading from minio

Checkout repo before start

Break out reused code

Another one

Last try

Last change

Another one

Retry build

Fix

Refactor actions for matrix builds

Fix deprecations

Added the ability to remove files using sftp

Fixed issues with sftp file listing and downloading from minio

design: tweak headers

fix(imgs): better handling of 404 images

Added caddy metrics to auth

docs(pgs): `_redirects` copy

feat(pgs): support `_redirects` special file for routing (#48)

Merge pull request #47 from picosh/am/soju-webpush

Added simple diff for soju to fix firefox and ios webpush

Added simple diff for soju to fix firefox and ios webpush

fix(bouncer): makefile cmd

chore: update soju.config

doc(pgs): copy edits marketing page

feat(pastes): extend file expiration from 7 -> 90 days
chore(pastes): update copy
refactor(pastes): move ssh server into `pastes` package

chore(pgs): add more mime types to helper

Implement a fake oauth login system that uses ssh generated tokens (#44)

Ensure caddy metrics are enabled

Ops cleanup

Merge pull request #43 from picosh/am/irc

Adds infra support for irc infrastructure

Added monitoring caddyfile

Fixed minio env var

Work on adding bouncer and auth services to deployments

chore(docker): sync production changes

fix(imgs): regression with upload changes

fix(bouncer): soju config

feat(pgs): add .webmanifest and .map

chore(irc): separate docker compose file

fix(imgs): when calc user's total img size, coalesce null to 0

chore: add more error logging for us to debug

feat(auth): web service (#42)

Authentication (oauth2) web service intended for soju bouncer.

We are also introducings tokens which users can create through the SSH CMS for all services. This will allow users to generate an API token so they can authenticate with our oauth2 API.

feat: anchor links for prose

design: preserve header `font-size` when there's a `<code />` child

fix: gitignore

design(pastes): max-width 1025px

feat(pgs): pretty urls

chore(router): remove noisy logs

doc(pgs): copy changes

doc(pgs): better articulate this is closed beta

Also describe the requirements for participation

feat(pgs): allowing `.rss` `.atom`, and `.xml` file extensions

feat(pgs): `/rss` endpoint

feat(pgs): `{user}.pgs.sh` supported by creating `{user}` project

chore(pgs): move most ssh code to pgs package

Consistent ordering of nav items (#40)

* Ensuring consistent ordering of nav items

If nav items are declared in the yaml without the "-" prefix ordering
was random of those items and a page refresh could change where they
are. This ensures ordering is preserved in that situation.

fix(pgs): link to help page

refactor(pgs): check handler should return 404

fix(pgs): new `/check` handler for custom domains

chore: add more logging for router

Added basic .editorconfig file. (#39)

Updated dependencies required in readme. (#38)

fix(pgs): routes should prefer `/check` first

fix(pgs): retain

fix: caddy

fix(prose): upload single file without referencing name in destination

Fixes #36

fix(pgs): allow users to run commands when bucket quota exceeded

docs(pgs): marketing copy

chore(pgs): print summary for prune op

chore(prose): updated blog rss feeds

refactor(pgs): cli outputs and asset upload optimizations

fix: lint

refactor: move Cmd to own file

fix: use go1.21 in CI

use go1.21

refactor(pgs): reorganized ssh cmds

feat: all cmds are safe by default, use `--write` to commit

fix: `project` might not exist during link

feat: order projects by name asc

chore: delete orphaned assets when linking projects

fix: only delete assets from exact project match

refactor: abstract removing assets

upload to hey instead of erock

fix: validate project names

fix(pgs): add check handler

docs(pgs): rm markdown syntax

fix(pgs): only ListObjects with specific match

fix(pgs): link regression

fix: better 404

fix: better object detection and route detection

feat: delete orphaned files when link succeeds

fix: deploy pgs site

fix: install deps

fix: typo

chore: automatically publish static site to pgs

chore: cleanup

feat: more pgs commands

feat: pgit static site generation

docs: update readme

fix: caddy docker build

fix(pgs): better error handling for project mgmt

changes

changes

fix: sftp `ListFiles`

It previously wasn't returning something that sftp was expecting

allow recursive for sftp

docs: add way to signup

use `scp -r` for now

fix: upload url

fix: upload url

feat: make cmd for pgs static site deploy

fix: podman bp commands

feat: cannot link to a project that is also a link

A project linked to another project which is also linked to a
project is forbidden.  CI/CD Example:
        - ProjectProd links to ProjectStaging
        - ProjectStaging links to ProjectMain
        - We merge `main` and trigger a deploy which uploads to ProjectMain
        - All three get updated immediately
This scenario was not the intent of our CI/CD.  What we actually
wanted was to create a snapshot of ProjectMain and have ProjectStaging
link to the snapshot, but that's not the intended design of pgs.

So we want to close that gap here.

feat: pgs.sh (#28)

A zero-dependency static site hosting service for hackers.

Features:
- Terminal workflow
- No client-side installation required to fully manage static sites
- Distinct static sites as "projects"
- Unlimited projects created on-the-fly (no need to create a project
  first)
- Deploy via `rsync -a . erock@pgs.sh:/myproject`
- Symbolic linking from one project to another (to support
  promotions/rollbacks)
- HTTPS for all projects `https://erock-myproject.pgs.sh`
- Custom domains for projects (managed simply by `TXT` records)

Updates for macOS build compatibility. (#20)

Updated the readme to include a reference to a macOS system package dependency.
Added minor additional notes on direnv and running the project.
Upgraded project package dependency that includes a macOS build fix.

Ensure 'hidden' value gets updated when the draft attribute is modified. (#21)

fix(prose): `/raw/` endpoints should take priority over aliases

chore(feeds): feed_items table to store feed items (#16)

Some RSS feeds are technically invalid but still want to be fetched.  In
these cases we want to make sure we can still fetch them and properly
handle when the feed item does not contain correct tags for dates.

Previously we used `published` and internally `last_digest` to determine
if a feed has already been sent to the user.  Well some feed items do
not contain `published` so we would always send the feed item to the
user in their email digest.

Now we are storing the feed items we have fetched in a separate table
along with some of its metadata in order to properly mark a feed item as
"already sent."  Now it doesn't matter if the feed item has the correct
tags or not, we will try to fetch the items, send them to the user, and
them mark them as already sent.

chore: move docker commands into variable for makefile

chore(imgs): add in-memory caching for resizing images

This caching could also be used for the other services as well now.

bug(imgs): ensure webp imgs are saved

Also in this commit is a change to the resize filter algorithm to see if
that helps with quality and performance.

Read more: https://github.com/disintegration/imaging#image-resizing

bug(imgs): resize regression because of incorrect switch case

chore(prose): add some more rss endpoints

fix(imgs): do not convert svg to webp

We never actually tested svg images before "supporting" it.

style: nav links

style(prose,lists): each nav link in aside theme gets own line

chore: fix lint errors

feat(prose): aliases can have a leading slash now

It feels more natural to have a leading slash `/` for aliases even
though it's a little more work to handle it properly with current slugs.

bug(aliases): add support for comma delimited string

style: tweak max-width

style: tweak design

feat(prose): post aliases (#13)

Posts can now have alias routes as well as the default filename as the
slug.

This is an important feature to support migrating from a self-hosted
hugo blog to one hosted by pico.

```
---
title: My post
aliases:
    - my_awesome_post
    - 2023/03/10/my-post
---
```

feat(prose): support custom favicon

Now bloggers can supply their own favicon inside the `_readme.md`

```
---
favicon: https://zzz.png
favicon: /zzz.png
---
```

If relative paths are supplied we replace it with `imgs.sh`

fix(lists): use correct var for blog templates

feat(cms): ability to see view count for posts

chore(db): add view count to post struct

style(prose,lists): tweaked /read and blog styles

design tweaks

feat(prose,lists,imgs): Add rss alt link to blog page

Closes #9

Fix feeds panic and lists lint

Update caddy resolvers to use cf

fix(feeds): wrong id for inline content section

chore(feeds): http client cleanup

fix(feeds): support reddit rss feeds

docs(feeds): add inline content section

fix(feeds): when inline_content is false we should only send summary in
email

feat(feeds): add option to disable inlining of post content in feed.sh emails (#8)

feat(feeds): add description to feeds

feat(feeds): render actual content

fix(imgs): support uploading webp

We always intended to support `image/webp` but we never actually tested
the code path.

chore: cleanup old code

fix(feeds): remove log stmt used for debugging

fix(feeds): fix validation for digest interval to include `6hour`

fix(imgs): typo on hotlink for post

feat(imgs): return current total size for user storage

This value will be displayed when uploading an image to imgs.sh.

Example of what it looks like: `https://erock.imgs.sh/barefoot (space: 0.01/1.00GB, 1.39%)`

chore(imgs): Add hot links and resizing examples on each imgs post page

fix(imgs): allow deletion using the empty post method

feat(feeds): add `6hour` to feed digest interval

fix(prose): rss link was incorrect

fix(feeds): actually use returned date

chore(feeds): change subject of email based on post title

docs(feeds): update roadmap

chore(feeds): fix linter

feat(feeds): support html

Also changed how we fetch feeds

bug(feeds): cron was not properly validating the digest period

fix(feeds): added logging statements and some different fetch logic

fix(feeds): cron log is using incorrect `posts.Total`

docs: update readme

docs(feeds): update privacy policy

docs(feeds): marketing page updates

fix(feeds): add feeds space to posts

change gh action name for feeds img builder

feeds.sh (#2)

* feat(feeds): init

docs: add imgs.sh to readme

Set default address space to not conflict

Merge pull request #6 from picosh/am/no-matrix

Change build to use composite instead of matrix action

Use proper platforms setting

Fix composite action

Return env vars

Change build to use composite instead of matrix action

Return caddy build to separate job

Add caddy step to normal build

Use updated image references

Merge pull request #5 from picosh/am/ci-cd

Added basic ci/cd

Fix lint

Install libwebp when running actions

Added basic ci/cd

chore: reference github instead of sr.ht

fix: suppress fsetstat error

feat(prose): permit rel attribute on hyperlinks.

Closes https://todo.sr.ht/~erock/pico.sh/51

docs: remove print stmt

refactor(pastes): add col `expires_at` to posts

Currently we figure out how to delete pastes by having a cron goroutine
check on an interval whether or not posts are older than 3 days.
Instead, we want to provide the ability for that delete time to be
determined by individual records.

So we added `expires_at` column to our `posts` table.

I also extended the default expiration from 3 to 7 days.

readme update

docs: point sourcecode links to github

bug(lists): transparency page included wrong templates

doc updates

feat(lists): relative image urls now route to imgs.sh

This reaches feature parity with prose.sh where we can reference images
hosted by imgs.sh from lists.sh

Both "/image", "./image", and "./image.jpg" are supported.

feat(lists): layout support

Introduced parity with prose for different layouts.  Right now we
support "default" and "aside".

fix: remove check for imgs feature flag

fix: correctly cast bool frontmatter

feat: support `draft` metadata property

Currently we support setting `date` into the future which will result in
the post being still URL addressable but not in our discovery feed, rss,
or the author's feed.

refactor: use neurosnap/go-exif-remove for custom patches

`go-exif-remove` uses `go-jpeg-image-structure` in order to strip exif
data.  Unfortunately there's a bug in `go-jpeg-image-structure` that
makes it so some jpeg image aren't detected as jpegs and then fails.

See open PR for details: https://github.com/dsoprea/go-jpeg-image-structure/pull/19

bug(upload): removed user field from shared file uploader struct

When a user attempts to upload files to our server, the
logic we use to perform file uploads is baked into a few similar
structs (`ScpUploadHandler` and `UploadImgHandler`).  In those
structs we authenticate their `ssh.Session` by using their name
to find a user record in our db.

We mistakenly had a user field on those shared structs which would
sometimes result in a race condition where user A would upload files to
user B's blog.

We have removed the user field and instead leverage
`ssh.Session.Context` to set and access the user object for our file
uploaders.

feat(imgs): set img file size limit to 10mb

docs(imgs,prose): add more info for docs

Also added a section on the prose marketing page about
how to upload images with your blog.

docs(imgs): fix typo with integrations header

feat(imgs): general availability

fix(imgs): set level to 1 for lossless

warning: this was the only value I could supply to get this to work

feat(imgs): small images are lossless

feat(imgs): be more explicit about quality setting

Also change resize algo to nearest neighbor

feat(imgs): only modify quality if file is large

refactor(wish): downgrade error stmt to info

docs(imgs): update marketing copy

refactor(imgs): simplified logic in imgHandler

docs: remove wish/readme since it's out of date

feat(imgs): we save webp images next to main img

fix: use host domain for full urls

fix: add other rss routes to subdomains

fix(lists): use wrong file for readme text

fix(imgs): jpg not always detected in exif scrubber

refactor(imgs): move image processsing to shared

Add the ability to download images and fix sftp uploads

fix(parser): use right imgs url for custom domains

Cleaned up configuration and added debug flag

Work on optimizing imgs.sh

Remove os/user ldd warnings

Switch to debian based images and implement cgo cross compilation

docs(imgs): remove metadata sections

refactor(imgs): use go-webp

feat(prose): special _footer.md appends all posts

This allows users to be able to upload a special "hidden" file that will
be appended to every post.  The intention is to use this as a regular
call-to-action like an email address for feedback.

fix(imgs): update marketing page with example

feat(imgs): strip exif data for png, jpg

feat(imgs): web optimized images

fix(all): moved auth middleware above bubbletea

feat(prose): support for og images on blog and posts

Move logging before auth

Implement new buffer for sftp and simplify auth logic

docs: add section about custom domains check endpoint

feat: support for different blog page layouts

"default" and "aside" are currently supported by adding `layout` in the
_readme.md frontmatter

fix(rss): lists set proper rss entries for blog

fix(rss): limit blog rss feeds to 10

Fixed image deletes from cms and fixed prose image uploads

Move configuration options and add the ability to delete image files from CMS

Added storage to scp handler

Fixed docker build

Added ca certificates

Added minio support to imgs

Added prometheus settings for minio

Fix minio console port and caddy config

Added caddy configurations for minio

Added minio

feat(prose): markdown footnotes

style: tweak heading sizes

feat(prose,lists): filter posts by all tags read

Now users can filter posts by tag on the read page.  This will allow us
to host writing competitions where users can submit their article with
the tag specifically for the competition and then we read them to pick
a winner.

Fixed list condition

Added rsync and sftp download support

Cleanup update logic and update to go 1.19

style: add some margin between articles

refactor(all): making sure every post has shasum

fix(imgs): use updated shasum logic

refactor(imgs): remove markdown metadata uploader

After further thought, using a second markdown file for the metadata of
images was just a little too cumbersome.  I also have a hunch that most
people are not going to take the time to annotate their images into
mini-blogs.

refactor(imgs): remove tag tmpl files

refactor(imgs): redesign blog page for imgs

After chatting with the pico members, it became clear that the photo
album direction I was heading in was not going to super useful.  Instead
we decided to revert the design back to how the other pico services
function.  So now it's just a stream of images not grouped by tags.

fix(mdparser): supply goldmark html config for ImgRenderer

feat(prose): allow images to be uploaded from other services

refactor: move prose markdown parser to shared

docs(imgs): add another feature to roadmap

docs(imgs): typo

docs(imgs): marketing content updates

docs(imgs): add section about metadata

fix(all): updated rss feeds to contain more information

I also added aliases to user blog's rss feeds

docs(all): update server costs and provider

Add rsync support to imgs

fix(sql): updated posts table columns to fix regression

We did not assume mime_type or shasum could be null, so this adjusts the
migration to use what we pasted into our prod db.

Added imgs docker deployment data

docs: update example env file with imgs service