Register new handlebars helper to remove backticks from page.content

[rshmhrj]

Hi @coddingtonbear -- amazing work on this extension! I definitely think it is a must-have for any Obsidian user!

I'd like a way to remove all backticks from the page content BEFORE it is written to the Obsidian note. In my use-case, I'm trying to pass the entire page content to OpenAI to autogenerate a summary. The below works, but ONLY if page.content doesn't have backticks. If a backtick is there, Templater blows up and can't process the note.

This entire thing is pasted into the Capture page snapshot template. You can see the Templater start <%* and end %> tags. The second line is the problem. For some articles, it works smoothly -- the entire content is temporarily added to the Obsidian note, Templater kicks in and processes the OpenAI call, and then returns the summary, which is then embedded in the template with <% summary %>.

<%*
const words = `{{page.content}}`.split(" ");

let startIndex = 0;
const keyPointPromises = [];
const MAX_TOKEN_SIZE_IN_A_REQUEST = 4096;
  
async function fetchSummaryFromOpenAI(content) {
  const prompt =
	"Below is a transcript of a Youtube video. Can you give me at least 5 key points and return the results as an unordered markdown list? \n --- \n";
  const MAX_TOKEN_SIZE_IN_A_REQUEST = 4096;
  var url = "https://api.openai.com/v1/chat/completions";
  var token = "Bearer " + "sk-open-ai-token";
  var systemRole = "You are an expert note taker and summarizer. You have been tasked with creating a concise summary of an online article using its page content. Make a summary of the content. Additionally make a short complex analogy to give context and/or analogy from day-to-day life from the transcript. Create at least 10 bullet points (each with an appropriate emoji) that summarize the key points or important moments from the content. In addition to the bullet points, extract the most important keywords and any complex words not known to the average reader as well as any complex acronyms mentioned. For each keyword and complex word, provide an explanation and definition based on its occurrence in the content. Please ensure that the summary, bullet points, and explanations fit within the 330-word limit, while still offering a comprehensive and clear understanding of the content."
	const p = {
		url: url,
		method: "POST",
		headers: {
		"Authorization": token,
		"Content-Type": "application/json"
		},
		body: JSON.stringify({
			model: "gpt-3.5-turbo",
			messages: [{"role": "assistant", "content": `${systemRole} Here is the video transcript: ${content}`}]
		})
	}
  const rawResponse = await tp.obsidian.request({url: p.url, method: p.method, headers: p.headers, body: p.body, contentType: "application/json"})
  const jsonResponse = JSON.parse(rawResponse);
  return jsonResponse?.choices[0]?.message?.content;
}

while (startIndex < words.length - 1) {
	const endIndex =
		startIndex + MAX_TOKEN_SIZE_IN_A_REQUEST > words.length
			? words.length - 1
			: startIndex + MAX_TOKEN_SIZE_IN_A_REQUEST;
	const contentChunk = words.slice(startIndex, endIndex).join(" ");
	startIndex = endIndex;
	keyPointPromises.push(fetchSummaryFromOpenAI(contentChunk));
}

const keyPoints = await Promise.all(keyPointPromises);
const summary = `${keyPoints.join("\n")}`;
%>---
title: {{article.title}}
page-title: {{json page.title}}
url: {{page.url}}
date: "{{date}}"
tags: {{article.siteName}}
up: ""
---

### Notes:
<% tp.file.cursor() %>

### Summary:
<% summary %>

---
{{#if article.excerpt}}
### {{article.excerpt}}
---{{/if}}{{#if page.selectedText}}
 
{{quote page.selectedText}}
 
---
 
{{/if}}{{page.content}}

I think adding something like this in handlebars.ts might help:

Handlebars.registerHelper("removeBackticks", (content): string => {
  return content.replaceAll("`", "");
});

Then at the top of my obsidian-web template, I could do something like this:

<%*
const words = `{{removeBackticks page.content}`.split(" ");
...

If you are okay with the idea, I can try to put together a small PR with the changes.

Thanks again for this awesome tool!

[coddingtonbear]

Before we think through options around how to work around the problem, I wonder if you could help me understand why the backticks are a problem? I might be misunderstanding, but, if you're running some sort of custom Javascript via Templater, wouldn't it be possible to remove the backticks in that logic?

[rshmhrj]

Hi @coddingtonbear , thanks for the swift response! I'll try to clarify.

const words = `{{page.content}}`.split(" ")

The entire page.content is added to the note before Templater can do any processing.

So the literal entirety of the page.content, with all the whitespace and formatting and everything is added to the note between backticks. Templater can then run and sees this is a string, and can perform the split operation to turn it into an array of words (which is later on used for chunking before passing on to Open AI).

If the page.content contains a backtick, then Templater no longer can tell that it is a string, its parsing fails, and then it aborts the processing of the note.

This is why I'm thinking the option to pre-process the backticks in obsidian-web would be useful. Also, since it's hidden behind a helper, it should not affect any existing templates.

[Dnarts]

I came here with the opposite request))
In my Obsidian, uncommented code arrives, and I'd like to insert it as it was on the website in ``` (preferably with the option to enable and disable the language if it was already specified on the website), because, on the contrary, I run the note through DeepL and then it translates everything that was uncommented, and I have to manually comment everything, which is very inconvenient.
Is there an option to enable what I'm asking for?
I am translating programming articles from English into my native language.