V1 features

[Keyrxng]

The bot means the BotFather TG bot.

List every feature in "v1" and write a proposal for what to add in "v2"

• /start looked like this before which would allow the setting up of a forum via a short or long url. Same thing needed, yeah?
• Create new tasks. An admin replies directly to the comment which should become a task and GPT turns it into a structured templated task. The admin reply is our command e.g /newTask in direct response to the msg same way you'd ban someone via slash command. Previously all messages were processed but I feel admin-reply controlled is a better idea, it's far more precise, less GPT overheads, we can use a better model as we'll use less calls for the same cost. Admins price or close tasks anyway, makes sense they determine if it's valid at the source.
• credit user who sent the message behind the new task that was created
• Wasn't built but create new tasks via voice notes, recordings etc.
• Wasn't built but embed our chatbot into the bot. I assume via the syntax @ubiquityOS?

Two tasks, group the first three and deal with the last two as it's own task.

---

I didn't work on V1 or use it on Telegram, so I’m basing this on code review:

• Heavy Supabase dependency regarding forums and github repos etc, I assume that's to be streamlined using TG as the DB. Setting forums up should be via command like the previous /start command did but an aptly named version?
• V1 had an OAuth feature for account binding. We have that but without OAuth, unnecessary to port over, both require the user register first.
• I found three commands: register, enable_topic, start. 1. Register is done without oauth. 2. Is enable topic wanted? 3. The old start helped setup a forum. The new start lists a welcome message and available commands. What's expected of the start cmd in private and in public chats?
• What's expected of the bot in forums and public groups as I only have context on workrooms.

[0x4007]

I think much more powerful to support inline task creation as well as context aware q&a (command-ask)

Here's a sophisticated inline bot example that generates custom animations.

I think pretty much everything else can be scrapped. It seems like a lot of artifacts from when we had less infrastructure and less clarity on direction.

[Keyrxng]

right so the only feature that needs ported over directly is:

• task creation: processing every comment without admin interaction, is that what you mean? I may misunderstand, the little autocomplete part in the image is the inline part. By directly replying to a message the reply fills that spot. The server update contains both senders etc. Otherwise typing / would fill that spot with commands. Am I to replicate that screenshot, you @ the bot and it offers options/commands that an admin would select?

new features:

• command-ask:
  1. context aware as-in, pretty much development right now so global awareness via embeddings at all times?
  2. Or if it's a repo forum it should know all about just that repo? Same for workrooms, it should be aware of just that repo and task?
  3. When used in big public chats, what's the conversation depth we want to fetch from the chat? The last 20 messages prior to the query or the last 50?
• audio > whisper > task:
  1. I will keep it in mind as I build command-ask but needs it's own task, learning curve involved with docs etc.

[0x4007]

I don't think anything will be ported. Inline message seems quite different.

Try and invoke the command-ask as close as possible as it is currently without making much custom adjustments for telegram client. This already can answer org aware questions which is useful. We can focus later on adjusting the prompt to be more useful specifically in telegram context. So for now, no chat context. Just the user's query only.

[Keyrxng]

I don't think anything will be ported. Inline message seems quite different.

I will implement as I first suggested and you can see how it looks and we'll take it from there re: creating new chats from user comments.

Try and invoke the command-ask as close as possible as it is currently without making much custom adjustments for telegram client. This already can answer org aware questions which is useful. We can focus later on adjusting the prompt to be more useful specifically in telegram context. So for now, no chat context. Just the user's query only.

Got it, minimal changes to current structure and the context window is the sysMsg + embeddings search + query.

[ubiquity-os-beta[bot]]

A new workroom has been created for this task. Join chat

[rndquu]

I don't understand the long-term vision for this plugin.

Right now it supports sending telegram notifications on:

• permit generated (issue_comment.created)
• inactivity reminder (issue_comment.created)
• forced unassignment (issue_comment.created)
• PR review requested (pull_request.review_requested)

Basically it just duplicates the github functionality, so in addition to github email notifications one will also receive telegram notifications. What's so special about this plugin that can't be achieved via github directly?

[0x4007]

For context, the future vision of UbiquityOS is to span across all major productivity platforms, including but not limited to: GitHub, telegram, google drive, notion.

This is our first step into capturing, processing, and recognizing events that occur on telegram in order to have a more wholistic understanding of how team members are contributing to getting projects done.

Most organizations are not as strict about their communication policy being mainly under the GitHub issue for example. We should try and make UbiquityOS as seamless as possible to add in the background and 1. collect org intelligence and 2. Compensate the top contributors the most.

As you can imagine, traditional organizations have more varied communication channels including the likes of email, slack, zoom, in person. Ideally UbiquityOS should be context aware and understand what projects they are working on, and credit them accordingly for their productive communications related to completing that project.

---

I think the initial setup is the biggest immediate goal here. Then we can start looking into "telegram kernel plugins" to extend its capabilities.

Off hand, I think that it makes sense to:

1. Generate vector embeddings to make the system more intelligent for any command-ask
2. Allow instant task creation in conversation
3. Have a private interface to be able to interact with the system (like setting wallet, queries, label permissions, q&a)