typos marking typos for words not in words.csv

[jamesbraza]

I am using typos to monitor the emergence of typos across an LLM training.

I have noticed words such as reductamol are typos, but typos doesn't flag them as such because there's not an entry in words.csv.

Is there some CLI flag to configure typos to mark any word not matched as a typo?

---

Here is my current typos configuration:

[tool.typos.default]
extend-ignore-re = ["(?Rm)^.*(#|//)\\s*spellchecker:( )?disable-line$"]

[tool.typos.default.extend-words]
arange = "arange"  # numpy
yhat = "yhat"  # \hat{y}

[epage]

There is not. We do not keep a list of valid words, along with variants.

[jamesbraza]

Doesn't https://github.com/crate-ci/typos/blob/v1.32.0/crates/typos-dict/assets/words.csv have a huge list of typos?

I am wondering if I can get typos to flag this somehow:

reductamol

The way I think it can be done is reporting any word that does not have a key in the above linked words.csv

[epage]

The keys in that file are typos; the values are valid words. The list of valid words is too incomplete to be relied on. We also don't at least time generate a way to lookup the presence of values in the the way we use that data structure generated from that file.

[jamesbraza]

Sounds good, thank you.

I think this would be a nice opt-in feature of typos, if you'd entertain something like this as a request