Introducing Stenography - An API for understanding code

Hi Community!

I’m super happy to be launching Stenography, an API powered by GPT-3, Codex, and some tricks of the trade I’ve learned from software engineering and being a developer advocate for five years.

My goals with Stenography were threefold:

  1. Give code its own “voice”, outside of the engineer who wrote it. Give anyone the chance to interact with a codebase written in any language.

  2. Make onboarding/reading someone’s code/writing documentation involve less tedium. Reading code or writing documentation should not be a struggle or depend on your expertise with code!

  3. Make KT of code easier between local and remote teams, decreasing organizational slowdown between engineering and non-engineering teams

I think the result is pretty cool, and I’ll be actively working at Stenography in service of the above three goals.

I hope you’ll give it a go! :slight_smile:

https://stenography.dev/

P.S. there is a fourth bonus goal of making the API as simple to use as possible for any coding language or format. And also, to be able to put the API anywhere, i.e. Slack bots, Discord, VSC extensions, Obsidian, Notion, you name it!

P.P.S if you are in the SQL/Data Analysis space, check out scribe! https://twitter.com/itsjonahkaplan/status/1442696196506587146?s=20

14 Likes

I think you might get more interest if the front page is improved; it’s a bit hard to see what it does.

I see a screenshot of an API request using curl. However, it’s not clear what the response is. There is a hard-to-read comment in grey on black, but I’m not sure what exactly is generated? Is it the description of the adder function?

It would be good to have a page of API examples that you can read without signing up.

1 Like

That’s great feedback, thank you @skybrian :slight_smile:

I was about to sign up, but I don’t see a link to the Terms of Service or Privacy Policy. I need to make sure that you don’t own the code I upload before I join :slight_smile:

1 Like

That’s super cool! I wonder if I had made something like this myself, if I wouldn’t have just stored all the code people put in … for whatever really lol. I wonder that about OpenAI themselves too, if they now have an enormous database filled with all fine-tuning files, executions, etc… Kinda jealous ngl

Just released a VSC extension (mainly an API update + a VSC extension) that smart parses entire code files and runs the Stenography API!

The project has evolved greatly since the last post!

I present, automatic documentation.

No if ands or buts about it, I’m putting my money where my mouth is! Stenography can take any code file in supported languages, parse them, read them, and return coherent explanations. No highlighting code blocks, no waiting for different code blocks to return at different times, no developer interruptions. Check it out if you’re intrigued!

2 Likes

very nice! Installed and trying it out. Two comments about the VS Code extension.

  1. can I turn off
  • generated by stenography autopilot [ :red_car::woman_pilot: ]

Or at least make it appear only once in the document, near the top?

  1. I don’t see a way to tell it to create all the documentation blocks at once for a given file? Or for all files in a project?

  2. It would be nice to set a threshold where no invocation is carried out for very simple functions that really do not need a redundant comment block. E.g.:

'''
This code is getting all the paragraphs in a document
- generated by stenography autopilot [ 🚗👩‍✈️ ] 
'''
def get_all_paragraphs_in_docx(document):
        # get a list of all the paragraphs
    paragraphs = document.paragraphs
    return paragraphs
  1. The phrasing of “commit comments to file” is a bit confusing. It’s not an actual git commit, right? It’s just inserting new text into a file that will subsequently be committed by the user.
2 Likes
  1. can I turn off * generated by stenography autopilot [ :red_car::woman_pilot: ]

i didnt add that but ill put an option in to take it out in the settings :slight_smile:

  1. I don’t see a way to tell it to create all the documentation blocks at once for a given file? Or for all files in a project?

autopilot is basically parsing a code tree and trying its darndest to find “complex but not TOO complex” code blocks. So it won’t do every line, just the ones where it can add value. That said, I’m actively working on the parser layer and would love any feedback on code you’d like to see commented more :slight_smile:

But if you want to add all comments to the file at once run Cmd + Shift + P -> Stenography Autopilot

It would be nice to set a threshold where no invocation is carried out for very simple functions that really do not need a redundant comment block

gotcha, great parser feedback thank you! (n.b. edit: each language is unique and idiosyncratic, so some parsers evolve at different rates than others, if feedback is language specific thats even better)

  1. The phrasing of “commit comments to file” is a bit confusing. It’s not an actual git commit, right? It’s just inserting new text into a file that will subsequently be committed by the user.

no its not commiting to git haha, good point. any suggestions? write to file perhaps?

edit 2: the github repo for the vsc extension is open source if you wanna submit any issues there, id super appreciate it!

3 Likes

Following up on

and

Both have been added in the newest 1.1.8 version. Thank you for the feedback, @NimbleBooksLLC :slight_smile:

2 Likes