The job
What the Llama strip HTML tool keeps and what it removes
The fastest way to know whether this tool fits your workflow is a list of what comes out the other side. The pass is fixed, so the answer is the same every time and you can plan around it.
- Removed
- Every tag, and the code inside script blocks and the CSS inside style blocks.
- Kept
- All text content, all line breaks and all indentation, since whitespace is never touched.
- Kept unexpectedly
- Text from the head of a document, including the page title, which arrives at the top of the output.
- Converted
- Exactly four entities: , &, < and >. Nothing else is decoded.
Entities
The entity question, answered with the actual list
Every tool in this category handles entities differently, and most pages describe it vaguely enough that you have to test it yourself. Here is the tested behaviour, so you do not have to.
The practical consequence for an editing workflow: if your source is a CMS field that escapes quotation marks and apostrophes, you will see " and ' in the output and you will need a second pass or a find and replace to clear them. If your source is ordinary article markup, four entities is usually all it needs.
- Handled
- becomes a space. &, < and > become their characters.
- Left as text
- ", ', ©, —, ’ and every other named or numeric entity.
- Over-decoded
- Double-encoded markup such as &lt; comes back as a bare angle bracket instead of as <, because the replacements run in sequence.
How to
How to strip HTML from Llama text in three steps
Three actions and nothing to configure. Your original stays in the input box, so a bad result costs you nothing.
- Step 1
- Paste the markup. A fragment or a whole saved page both work, and there is no character limit.
- Step 2
- Run it. The plain text fills the output box straight away.
- Step 3
- Scan the result for leftover entity text and, if you pasted a full document, delete the page title from the top before copying.
The pitfall it avoids
Script and style contents go with the tags
The standard failure in this category is a stripper that removes script tags and leaves the JavaScript between them. You get an article with a few hundred lines of tracking code in the middle, and in a long document it is genuinely easy to miss until someone else finds it.
Style blocks behave the same way, dumping selectors and colour values into the middle of your prose.
Both block types are removed with their contents here, before the general tag pass runs. That is a small implementation detail with a large effect on whether the output is safe to paste without reading all of it.
Known limits
Two inputs that defeat this tool
This is pattern matching, not document parsing, and two inputs break it in a predictable way. Neither is common in published article markup, and both are worth a look if you are pasting hand-written HTML.
- Angle bracket in an attribute
- Markup like <a title="a > b" href="#"> ends the tag match early, and the rest of the attribute text appears in your output.
- Angle bracket in a comment
- A comment containing a > leaks its remainder the same way.
- Broken markup
- Unclosed tags and bad nesting are not repaired. Anything the pattern does not read as a tag is treated as text.
- What to do
- For those sources, read the output rather than copying it blind. For ordinary content HTML, the pass is clean.
Structure
Line breaks are kept and there is no toggle
Competing tools offer a keep-line-breaks checkbox. This one has no options, and line breaks are always kept, because the pass removes tags and never touches whitespace.
That means a source with markup on its own indented lines produces output with those same blank lines and that same indentation. If you want the result compacted afterwards, that is a separate whitespace pass rather than a setting here.
Workflow
Where stripping HTML fits before a model run
Searches that combine a model name with stripping tags usually come from one task: turning a saved web page into clean input. Markup adds length without adding meaning, and the tags say nothing useful about the prose you want summarised.
Stripping first is the cheap version of that step. It will not select the article out of a page for you, so a full document still arrives with navigation text, footer links and the page title attached as plain words. Paste the region you want rather than the whole file and the output is much closer to usable.
In practice
Who uses the Llama strip HTML tool, and when
Editors pulling body copy out of a CMS field that stores raw markup. Support teams reducing an email template to its words before rewriting it. People preparing a saved article as plain text input.
It is the right tool when the markup is disposable. It is the wrong tool when structure matters, because nothing here turns tags into headings, lists or links. Removal is the whole feature.