pmkelly4444

If you are building AI Agents in Python and want to improve token efficiency, please have a look at the RC of the pctx Python SDK

e2b.dev is focused on this space

yes! coming soon

this makes sense, we should support a model where the code snippets can all be stored on the filesystem rather than in the context window from the MCP response

got it, yes so currently this is built just for one user - one set of credentials, but passing user credentials through is something we want to add.

thinking a native connection to cloud auth managers is the best way to do this (clerk, auth0, etc.)

thank you, will get a more detailed benchmark out soon!

no dependency injection at the moment... this is something we are exploring. adding dependencies would require rebuilding the execution runtime, which is something we want to open up in the framework soon

hey! the way it works is that the llm is first given snippets in typescript that tell it how to use the various MCP tools. it then can code, in typescript, and execute all of the tool calls in the deno sandbox. so yes, it can only execute javascript, but this isn't meant to be a full arbitrary code execution env like E2B.dev is, this sandbox is only meant to be a place for MCP calls to happen.

we chose typescript because it's the most token efficient way to pass types and signatures to an LLM, with Python and Pydantic there are extra characters passed around

thank you!

on your second point, check out how we lock down the sandbox with a custom deno runtime! https://github.com/portofcontext/pctx/tree/main/crates/pctx_...

on third, will def get some benchmarks out... we setup OTEL so we have the data

nice! will follow your progress! love that this runs locally as well

Interesting, are you using LLM rules files?

And you're stuck maintaining the glue by hand because no generator understands the relationships between all these different systems.

An SDK generation solution that utilizes a macro! in Rust to maintain custom files added between code generations with zero configuration.

SJ Morris, HubSpot's Developer Community lead, discusses the unique challenges of developer relations (DevRel) at "developer-plus" companies, where unlike public relations, devrel must constantly justify its value despite serving an increasingly broad technical audience that extends beyond traditional developers.

extremely useful

this is very cool! I just tried using it, unfortunately, my NextJS app dir project makes most requests from the server side, so it was only capturing "posts" made from the client. Is there a way to run it from the server?

edit: we've moved all unofficial projects to the "public" subdomain: https://public.sideko.dev/hacker-news/docs

thank you! It's funny, loading a .so/.dll was our initial approach to the problem. we found that this didn't work for us for two reasons.

1. Security/packaging problems. Security problem is self-explanatory, and for packaging binaries in SDKs we ran into a lot of issues with languages that don't have first-class support for binaries. It was particularly problematic in golang. 2. Iteration time: waiting for a cross-compilable binary each time we wanted to create a new SDK was a huge testing issue. we needed to build in a way that let us test our outputted SDKs in seconds rather than minutes. we now have a testing pipeline that runs in less than a minute on a bunch of live APIs

so we've gone with an approach that uses a combination of custom macros + templating

Thank you! We will add links to the github/docs.rs to our documentation ASAP.

And yes, language-level docs will be implemented. So far we've found that our users are happy with using the snippets in the docs + autocomplete from typing in code editor. But for advanced use cases I definitely see the value in language-level docs.

Google results showing me this occurs in Firefox sometimes. will check it out

Thank you!

check out uniffi for a complete workflow for FFI in many languages: https://github.com/mozilla/uniffi-rs

it looks like a notion clone on first inspection. your github/website could be more powerful if you highlighted some differentiators! I see there is an "emails" tab. Does it allow you to track emails like HubSpot?

love this! will use for personal site

Could you give an example of how this would work with an API? Trying to understand what you mean

I am building something in the SDK generation from OpenAPI space. This is making me reconsider the roadmap as ChatGPT is now somewhat of a natural language SDK.