Your homepage has no screenshots to inspire confidence to signup. All links/buttons just jump to signup. Maybe add ^ video to improve conversion.
Also having nothing open source on GitHub was unexpected but understandable. It's ok to be private for a private company. It's just that the entire ecosystem is open source. A day may come when it's strange not to have code running on your laptop to be open source. But it's totally your call.
PS: you should spend some money on a designer. Too many margins and wastage of space in your demo app. And adding some polish could go a long way in adoption/fomo/conversion.
Your readme doesn't mention any example of multiple agents. We don't need another wrapper to anthropic/openai/*. Without being able to select between multiple tools/agents - you only need 2 lines of code to achieve llm calls, and another 2 lines for state management.
That said the interface is more idiomatic that others agent frameworks that show up here.
It also wasn't clear where the last N messages is being persisted. Db/file/adapters/?
The goal of this reference agent SDK is to showcase the flexibility of Pixeltable (the underlying unified storage and orchestration system) that is Open Source and Apache 2.0.
Will add it to my todo list, also what do you mean by Canva gets free traffic because it creates kinds of slides and I could do the same. Do you mean like add boilerplate invoices or smth? cheers
Tldr; it's json array passed to llms. Swagger would have sufficed. How and why youcome up with the array shouldn't matter. We shouldn't need 10000 redundant servers.
I'd like to remind you that tools is a json array to any modern llm inference api. That rather than returning text, tells you which function to call.
I'm all for abstraction of a level of indirection. But this is pushing things too far.
We now have an entire ecosystem, layers of unneeded engineering, cohorts of talent and capital going to create man in the middle servers that forces us to get this array from around the world + maintain a server with several gb of deps to get a json array that you should't trust.
2) It makes sense if every server has a tools.txt equivalent of their own swagger. Eg i would trust google photos to maintain and document their tools rather than the 10,000 MCP servers possibly alive for no reason and already out of date by the time you are done reading this comment. In addition to being over engineered, to trust a random server as a proxy never made any sense.
3) nobody wants to run servers. Can't find this meme, but found it here on HN several times.
Sorry but I would rather not wait a year for this industry to crash and burn and take down genai apps galore or worse, start leaking this data and your bills.
Here is my feedback after spending 15 mins on your docs.
Nice work on the lite mode, open source, logging, dx interface.
You may want to replace Hello world examples with real world scenarios.
The workflows that involve multiple steps tasks, dag in your terminology - the code simply isn't intuitive.
You now have to get into the hatchets mindset, patterns, terminology.
Eg: the random number example is riddled with too many. How many of the logos on your homepage did you have to write code for? Be honest.
Knowing to program should be 90% enough. Eg for js:
// send("hi", user => user.signed_up_today)
// .waitFor("7d")
// .send("upgrade", user => !user.upgraded)
Just made this up, but something like this is more readable. (PS:would love to be proved wrong by an implementation of exactly the above example here in the comments). The whole point of being smart is for your team at hatchet to absorb difficulty at the benefit of an easy interface that looks simple and magic. Your 5 line examples has types to learn, functions to learn, arguments to know, 5-10 kinds of things to learn. It showed little effort to make it easy for customers.
An engineering post on what's under the hood makes sense. But customers really don't care about your cloud infra flexes in a post introducing your company pitching the product. It's just koolaid.
Same with complete rewrite so early. I'm glad you are open to change. But the workflow market today with so many options, i don't belive this is the last rewrite or pivot to come.
The DAGs itself aren't very readable. You are better off switching to something like react flow that lets you nocode edit as well.
Focus on automation journeys that are common. Like cookbooks. And allow folks to just import them or change some configurations. like drip marketing, renewals, expired cards, forgot password handlers, shortlink creators, maybe pdf merging, turning a bunch of saved links to a daily blog post, etc
How does a workflow replace a saas they are paying $99 for. That's powerful.
Tough to serialize a worflow to json . Or atleast didn't see it. this makes it easy to have workflows as code, create nocode editors in your own roadmap. You want people to hop from 1 company to another taking their hatchet workflows with them
Good luck, and sorry for coming off as rude. It's just a space I am very passionate about.
