lunarcave

9 ms·

61.
by lunarcave 2y ago
What did you use to build it? (Might be a good candidate for a Show HN [1]) [1] https://news.ycombinator.com/show
62.
Ask HN: How many of you open up your production DBs to internal users?
1 points by lunarcave 2y ago|8 comments
63.
You don't need tool calling(inferable.ai)
2 points by lunarcave 2y ago|0 comments
64.
by lunarcave 2y ago
Oh, you mean in terms of semantics? Yeah, that's a bit tricky. I guess that's true for any LLM related result though.
65.
by lunarcave 2y ago
It's more helpful if you give it the plans you already have and then ask how you'd improve them. Less helpful as a starting point IMHO.
66.
by lunarcave 2y ago
Does this [1] look like the thing you're looking for? We built a basic Re-Act agent that can dynamically inspect the database context and generate the queries iteratively. Our experience was that it was hard to do single-shot. Happy to
67.
Show HN: Open-source conversational AI agents for internal tools(github.com)
11 points by lunarcave 2y ago|4 comments
68.
How to help coding agents help you(inferable.ai)
4 points by lunarcave 2y ago|0 comments
69.
by lunarcave 2y ago
They work, just barely, and definitely not up to the marketing hype. I'm in the space, and it's really hard to "sell" a solution to someone who had bought the marketing bs and says "X said they can do Y". Most
70.
by lunarcave 2y ago
The discourse, has been taken over by hype - I'll give you that. I'm in the space [1] and I'll try my best to answer. Building on your definition, I'd say an agent is is a collection of LLM calls with structured outputs.
71.
by lunarcave 2y ago
I'm using the apps mostly. But I think you can use machines for more lower level use cases. IIRC, apps run machines.
72.
by lunarcave 2y ago
As someone who has been building "agents" for about a year, definitely over-hyped. They work well in extremely narrow use cases where: 1. A human is there to keep it on the rails 2. Tools are domain specific Most agents being show
73.
by lunarcave 2y ago
You can confuse the LLM to reveal its system prompt even if you put it behind a service API IMHO, as long as a user has access to the chat and can freely interact with the LLM.
74.
by lunarcave 2y ago
I have a hosted code-first agent builder platform in production, so I respond these question a lot from our customers. 1. Probably the best is fly.io IMHO. It has a nice balance between running ephemeral containers that can support long run
75.
by lunarcave 2y ago
We don't hate Python at all. We've just been waiting for someone to ask for it :) We're actually language agnostic (because we're operating at the level of the API). We're going after more underserved languages from
76.
by lunarcave 2y ago
Doing something open-source [1] in the space. Self-hostable as well. Our customers actually have built internal things to help them with Zendesk / replace some manual Retool tasks. There's a slack [2] integration as well. Fun fact
77.
by lunarcave 2y ago
In pg, each unique `listen(channel)` takes up server resources, and if you don't reliably clean them up, everything comes to a screeching halt. There's also `max_notify_queue_pages` >Specifies the maximum amount of allocated pa
78.
by lunarcave 2y ago
Yes, but that's one part of it though. For example, you have to: - change how you hydrate the initial state in the web component. - rework any request-oriented configurations you do at the edge based on the payloads. (For example, if y
79.
by lunarcave 2y ago
Author here. Having done this, I don't think I'd reduce it to "just a little bit of work" to make it hum in production. Everything in between your UI components and the database layer needs to be reworked to work in the
80.
by lunarcave 2y ago
Yes, but the problems of detecting that changeset and delivering it to the right connection remains to be solved in the app layer.
81.
by lunarcave 2y ago
Good call! I forgot that.
82.
Newspeak (programming language)(en.wikipedia.org)
3 points by lunarcave 2y ago|3 comments
83.
by lunarcave 2y ago
Good point! We did consider SSE, but ultimately decided against it due to the way we have to re-implement response payloads (one for application/json and one for text/event-stream). I've not personally witnessed this, but peo
84.
Back to basics: Why we chose long-polling over websockets(inferable.ai)
258 points by lunarcave 2y ago|163 comments
85.
by lunarcave 2y ago
Congrats on the launch. Very polished. Is there a way to track the price across time for a given model? This has been something I've been wanting to track for a while.
86.
Ask HN: Who's using agents for work?
2 points by lunarcave 2y ago|0 comments
87.
by lunarcave 2y ago
Yes! I've worked on something [1] in the space lately, and can share learnings. It generates queries, which gets fed to the application layer, and it executes them based on some deterministic controls. Basically, if you're working
88.
by lunarcave 2y ago
Claude within Cursor. I sometimes disable the Cursor Tab (tab to autocomplete) when I want to think fast. The new agent-based composer has been pretty good at maintaining attention and getting things done on the periphery. I'm slowly c
89.
by lunarcave 2y ago
Yeah, but the used car salesman's explicitly disincentivised to tell you the downside. An LLM is agnostic to the outcome. It's only incentive is to produce an output. A correct analogy here would be your distant uncle who knows ab
90.
by lunarcave 2y ago
Isn't that a form of checking?

More ›