# llm-food — Document preparation — ALTAI

> llm-food prepares model-readable content: it turns PDF, DOCX, RTF, PPTX and HTML files into Markdown and keeps their structure visible.

## llm-food. From document to Markdown.

Keep the structure visible. llm-food prepares model-readable content: PDF, DOCX, RTF, PPTX and HTML files become Markdown, through an API, a CLI or Python.

[GitHub](https://github.com/altaidevorg/llm-food)

1. **Drops the clutter:** Menu, ad, cookie banner and footer fall away. The content stays.
2. **Keeps the structure:** The heading stays a heading, the table a table. Now in Markdown.
3. **Reads five formats:** PDF, DOCX, RTF, PPTX and HTML files all come out as Markdown.

llm-food turns a web page into Markdown. The menu, related links, ad, cookie banner and footer are dropped; the heading, paragraph and table remain: “Device maintenance guide”, then the paragraph, the “Maintenance schedule” heading and a two-column table. Supported formats: PDF, DOCX, RTF, PPTX, HTML.

### Turn documents into text that AI workflows can use.

llm-food converts supported documents and web pages into Markdown as a foundation for search, review, and model-input preparation. It is available through a FastAPI service, Python client, and command-line interface. Processing depends on the file format and selected backend; some PDF paths send data to Google Cloud and Gemini.

#### Convert common document formats

Convert PDF, DOC/DOCX, RTF, PPTX, and HTML or webpage content into Markdown.

#### Choose a PDF processing path

Single-file PDF conversion uses configured backends. Batch PDF processing runs through Gemini Batch API and Google Cloud Storage.

#### Connect document preparation to a workflow

Call the service through its API, Python client, or CLI and pass the resulting Markdown to downstream tools.

### From input to output.

1. **Input:** A PDF manual, DOCX procedure, slide deck, or HTML page.
2. **Process:** Submit the file or page to the service and process it with the configured text extractor or PDF backend.
3. **Output:** Markdown text organized by page or section for search, review, or another language-model workflow.

### In client projects.

ALTAI uses llm-food to build conversion workflows that prepare an organization's documents for search and review applications. We select a processing path for the file formats, specify the services and data path, and connect the outputs to the application.

### Run the first example.

The converters were run for these examples; the full server setup was not tested.

````
git clone https://github.com/altaidevorg/llm-food.git
cd llm-food
python -m pip install ".[server]"

# Local text-based PDF backend
# Current source imports pypdf for the pypdf2 option
python -m pip install pypdf
PDF_BACKEND=pypdf2 llm-food-serve

# In a second terminal, with the same environment
llm-food convert-file /path/to/document.docx
````

### Spec sheet

- **Runtime:** Python 3.10+
- **Interface:** API / CLI / Python
- **License:** MIT core
- **Source:** [GitHub](https://github.com/altaidevorg/llm-food), [Documentation](https://github.com/altaidevorg/llm-food#installation--usage)
- **Scope:** Scanned PDFs go to Gemini; a local server does not make every path local. The optional PyMuPDF path is AGPLv3.

### Write to us to use it in your organisation.

[Talk to us](/contact/?interest=llm-food)

Source: https://altai.dev/open-source/llm-food/
