OpenWiki: CLI that writes and maintains agent documentation for your codebase

https://github.com/langchain-ai/openwiki

by handfuloflight • 21 days ago

96 28 21 days ago

github.com

dcreater• 21 days ago

What does this do better than just asking your agent to "write docs" or a more robustly defined prompt/skill?

_pdp_• 21 days ago

This is what we do. The same agent writing the code can also write the docs.

b212• 21 days ago

I swear most of these tools are made for the sake of it…

While good old prompting is often better than plan mode or superpower skills.

SpicyLemonZest• 21 days ago

I've had a number of people send me tools like this at work, and easily 50% of them can't answer basic questions like "what's the reason someone would adopt this tool" or "how do you know that it will achieve its stated goals". Agents are good at reading code, I can't imagine what the point of autogenerating agent context could be if it's not showing demonstrable cost benefits.

capplexham• 21 days ago

Have a look at the prompts in the GitHub [0]. It defines a System Prompt and specifies the documentation structure. This would allow you to switch coding agents, instead of relying on how your coding agent interprets the command "write docs".

[0] https://github.com/langchain-ai/openwiki/blob/main/src/agent...

LudwigNagasena• 17 days ago

That's just cruft unless there is a benchmark demonstrating it's actually better than just asking an agent to write docs and/or using one of the thousands of document-writing skills like Anthropic's doc-coauthoring.

rrvsh2• 21 days ago

maintaining an LLM wiki has been a lot more effort than I thought, at least if we are trying to maintain a high quality in structure and writing comprehension (for easier lookups both for the agent and human). Are people just shotgunning their agent wikis or how

bad_username• 21 days ago

maintaining an LLM wiki has been a lot more effort than I thought

Same here. Wikis start out good, but either devolve in a journal-y mess after a while and many updates, or require constant expensive rewrites. (I didn't use the software of the OP.)

reacharavindh• 21 days ago

Wikis start out good, but get stale too quick and become useless or worse confusing.

I’ve experienced this over and over again to strongly believe it.

I genuinely wonder if throwing LLMs at this problem would solve it at least to some extent. Make a LLM agent whose sole purpose is to act as a librarian. It periodically reviews _all_ of the wiki and validates them against codebases, newer docs, anything. Whatever it finds, it should be allowed to intelligently quiz the team/dev whether something is right/stale/wrong and updated it accordingly.

If one tolerates that toil - answering questions of a library bot, would it result in a usable wiki base?

TeeWEE• 21 days ago

This is mostly a thin clintypescript wrapper around the prompts.

This could have been a SKILL

themgt• 18 days ago

Yeah I went looking: https://github.com/langchain-ai/openwiki/blob/main/src/agent...

_andrei_• 17 days ago

it's langchain, what did you expect

mthoms• 21 days ago

We need to make an effort to distinguish “this is a thing for humans” vs “this is a thing for bots” in our naming IMHO. In that respect, “open wiki” is not such a great name. “Agent Wiki” or similar would be better.

Without such a qualifier, “wiki” carries a strong connotation of (usually collaborative) human involvement. That’s literally what it’s famous for.

Thats just my $0.02 on the naming. I definitely think it’s a worthwhile idea. All the best.

salviati• 17 days ago

I recently noticed people think "wiki" is a set of deeply interlinked documents. But to me it means ability to edit without logging in.

Ferret7446• 17 days ago

I don't know of any wikis that let you edit without logging in. That is simply not viable, it'd be like trying to live without an immune system.

zhengsihua• 18 days ago

How does the ability to search code snippets and symbols compare to Codegraph?

esafak• 18 days ago

Unless it's about motivation and other things that can't be inferred from the code (and comments on such are missing), just ask the agent. Give it an LSP or code intel MCP to do it better.

mellosouls• 18 days ago

I think it's own sub-dir OpenWiki serves as an example of the sort of output you might get:

https://github.com/langchain-ai/openwiki/blob/main/openwiki/...

I agree with others this seems somewhat over-engineered; you can get similar results with a good prompt/skill; I guess the rest of the implementation here is intended as an agent-maintainer.

commandersaki• 17 days ago

Is this the tool that blasted away all the wiki pages for Azure in the repos?

bignet• 16 days ago

extension for pi.dev https://github.com/barvhaim/pi-openwiki

ricardobeat• 16 days ago

Needs actual evidence that this is better than not having documentation, only AGENTS.md and letting agents explore.

Having verbose, duplicated, incorrect, outdated or incomplete docs often leads to worse performance for AI

OpenWiki: CLI that writes and maintains agent documentation for your codebase

Related Stories