codemonger-io / mumble Goto Github PK

bto and bcc must not be exposed to other than the sender.

We can form the timeline by combining outbox and inbox, though, an endpoint dedicated to the timeline may be implemented more efficiently.

According to Activity Pub's spec, inbox MUST be an OrderedCollection.

The outbox endpoint does not return the number of posts for now. Counting up all the posts every time is too expensive, so how about to remember and maintain only the number instead?

Atmark (@) does not represent the lowest ID because digit characters have smaller ASCII (0x30-0x39) code than @ (0x40). It should not be a problem in this case, but we'd better change it; e.g., to !.

Simply incorrect reference.

          I found one Mastodon server repeatedly sends `Delete` activities of different users supposed to have been in that Mastodon server.

Originally posted by @kikuomax in #22 (comment)

If we can view a profile page as an HTML page, it will make invitation easier.

To demonstrate the FlechasDB system, I want to add a feature to do similarity search over mumblings.

• #27
• #28
• Demonstrate the similarity search feature on the viewer
  • #29
  • #30

• #31
• codemonger-io/mumble-bee#15

Looks poor without the profile image.

We should take care of non-existing users and not abort the entire workflow due to missing users.

410 Gone

It takes about 3 seconds.

Tired of passwords.

Due to lack of Markdown support, past posts look messy. I think ActivityPub allow editing past posts.
https://www.w3.org/TR/activitypub/#update-activity-outbox

If there are more than one recipients from the same ActivityPub server, we should deliver the activity to the shared inbox of the server.

It only checks if "attachment" is a sequence, so far.

Once I made Mumble public, it started to receive dubious requests in the inbox. Hopefully, I could filter them before they reach the REST API to save the API calls. Since none of them bore a valid signature, they ended up with rejection by Mumble. So there is no harm so far.

Can WAF deal with these requests? Or, maybe I could analyze the pattern of dubious requests and implement a filter by myself?

• #21
• #22
• Plan the next action

We use OpenAI's embedding model "text-embedding-ada-002", because it is powerful and affordable.

• #31
• codemonger-io/mumble-bee#15

Note may have a source property, and a source property has content and mediaType.
https://www.w3.org/TR/activitypub/#source-property

Simply receive a query text? Or receive an embedding vector?

• #31
• codemonger-io/mumble-bee#15