feat: convert latex to plain text

[ldelossa]

Newest GPT models love to provide latex markup.

Unfortunately this makes things pretty impossible to read for larger questions.

Add a conversion to change latex markup to plain text.

Closes #87

Before:

───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────── user ──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
show me a gamma correction formula
─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────── assistant ───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
Gamma correction is a nonlinear operation used to encode and decode luminance or tristimulus values in image processing. The formula for gamma correction can be expressed as follows:

For a pixel value \( I \) (where \( I \) is typically in the range [0, 1] or [0, 255]), the gamma corrected value \( I_{\text{corrected}} \) is given by:

[
I_{\text{corrected}} = I^{\gamma}
\]

Where:
- \( I \) is the original pixel value (normalized to [0, 1] if necessary).
- \( \gamma \) is the gamma value (a positive real number). A common value for gamma is 2.2.

If the pixel values are in the range [0, 255], the formula can be adjusted as follows:

[
I_{\text{corrected}} = 255 \times \left( \frac{I}{255} \right)^{\gamma}
\]

In this case, you first normalize the pixel value by dividing by 255, apply the gamma correction, and then scale it back to the original range.

### Inverse Gamma Correction

To reverse the gamma correction (to get back the original pixel value), you can use the inverse formula:

[
I_{\text{original}} = I_{\text{corrected}}^{\frac{1}{\gamma}}
\]

This allows you to retrieve the original pixel value from the gamma-corrected value.
─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────

After:

───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────── user ──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
show me a gamma correction formula
─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────── assistant ───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
Gamma correction is a nonlinear operation used to encode and decode luminance or tristimulus values in image processing. The formula for gamma correction can be expressed as follows:

For a pixel value I (where I is typically in the range [0, 1] or [0, 255]), the gamma corrected value I_corrected is given by:


    I_corrected = I^γ


Where:
- I is the original pixel value (normalized between 0 and 1).
- γ is the gamma value (a positive real number). A common value for γ is around 2.2 for standard displays.

If the pixel values are in the range [0, 255], the formula can be adjusted as follows:


    I_corrected = 255 ×( I/255)^γ


In this case, you first normalize the pixel value by dividing by 255, apply the gamma correction, and then scale it back to the original range.

### Inverse Gamma Correction
To reverse the gamma correction (to get back the original pixel value), you can use the inverse formula:


    I_original = I_corrected^1/γ


This allows you to retrieve the original pixel value from the gamma-corrected value.
─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────

[npiv]

Hi @ldelossa

This looks great, but it seems to break markdown answers. For example "write a python fibbonaci recursive example"

gives me

vs (the old behavior)

so the detect_and_format_message would need to be a little bit less permissive, or the option for latex formatting should be hinted by a param.

I'm happy to have a look/play around myself if you don't have time, as I would like very much to include the latex option, but will probably only have time this weekend.

[ldelossa]

Uh oh! Should have tested a bit more. I wonder if simply doing the latex conversion after markdown parsing would help.

[ldelossa]

Ah, I see a bit of an issue, the markdown processing turns things right into term codes for the console. Where to interject latex processing then is a bit tougher. 🤔

[ldelossa]

@npiv latest push seems to fix markdown parsing in our simple testing here. I saw that the latex parsing would change '``' sequences to smart quotes. We just swap them back now.

[ldelossa]

This is looking pretty good.

[npiv]

yeah looks good, go ahead. I will release it and bump homebrew tomorrow.

Thanks a lot

[ldelossa]

Thanks a ton. While I use it, if i find any edge cases ill make more touch up prs that fix weird rendering.

Thanks again for chatblade, i literally use this 100 times a day.

[ldelossa]

Just a heads up, I can't hit merge cuz I don't have write access :)

[ldelossa]

Hey @npiv was this ever released? Got a new laptop today and installed latest chatblade, but I don't think this is there.

[npiv]

Hmm something went wrong, should be fixed now at least on pypi if you install via pip.

Homebrew will follow, but they can take a few hours usually

[ldelossa]

@npiv thanks! just upgraded with pip, working :)