**Site index:** [https://2plot.dev/llms.txt](https://2plot.dev/llms.txt) — every page on this site, as Markdown.  
**Network index:** [https://2plot.ai/llms.txt](https://2plot.ai/llms.txt) — The 2plot network; start here to discover sibling sites.  
**Sibling sites:** 9 more in The 2plot network — listed in the site index above.  
**Sitemap:** https://2plot.dev/sitemap.xml  

> **Full documentation:** [https://llms.2plot.dev](https://llms.2plot.dev) — the dedicated dash-improve-my-llms documentation site, with the complete API reference and deeper examples. This page is the quick-start overview.



### Installation

[Visit GitHub Repo](https://github.com/pip-install-python/dash-improve-my-llms) · [PyPI](https://pypi.org/project/dash-improve-my-llms/)

```bash
pip install "dash-improve-my-llms[flask]"     # Dash's default backend
pip install "dash-improve-my-llms[fastapi]"   # Dash 4.2+
pip install "dash-improve-my-llms[quart]"     # Dash 4.2+ async
```

Requires `dash>=4.1`. The backend is detected from `app.server`; your code is
the same either way.

### Introduction

A Dash app is a JavaScript application. Ask for any page without running
JavaScript — which is what most crawlers, link previewers and LLM fetchers do —
and every URL returns the same empty `Loading...` shell. To a search engine, a
30-page documentation site looks like 30 identical thin pages. To an agent, it
looks like nothing at all.

`dash-improve-my-llms` fixes that at the framework level. You write each page's
content once, as Markdown, and the package serves it everywhere a
non-JavaScript consumer will look: rendered into the page's own initial HTML
before React mounts, as `/{page}/llms.txt`, in the site-wide `/llms.txt` index,
in `sitemap.xml` and `robots.txt`, and as an MCP resource on Dash 4.3+. One
source of truth, six surfaces, and the interactive app is untouched.

This very site runs it — the [/llms.txt](/llms.txt) index, the Markdown twin
behind every page, and the cross-host network directory that links the
`*.2plot.dev` satellites together are all this package at work.

### Full Documentation

**The complete documentation lives on its own site:
[llms.2plot.dev](https://llms.2plot.dev)** — every route, the access-control
tiers, network configuration, the MCP bridge, and the machine-readable twin at
[llms.2plot.dev/llms.txt](https://llms.2plot.dev/llms.txt). This page is a
summary; go there for the full reference.

### Quick Start

```python
import dash
from dash import Dash, html
from dash_improve_my_llms import add_llms_routes, RobotsConfig

app = Dash(__name__, use_pages=True)

# Absolute URLs in sitemap.xml, llms.txt and canonical tags come from this.
app._base_url = "https://myapp.com"
app._robots_config = RobotsConfig(block_ai_training=True)

app.layout = html.Div([dash.page_container])

add_llms_routes(app)   # ← the whole integration
```

Then give each page prose — a module-level `LLMS_DOC` string, or
`register_page_metadata(path, name=..., description=..., llms_doc=...)` when
pages are generated in a loop.

### What You Get

| Surface | What lands there |
|:--------|:-----------------|
| The page's own HTML | The prose, rendered into the initial response, before React mounts |
| `/{page}/llms.txt` | The same prose as Markdown, with links back to the site and network indexes |
| `/llms.txt` | An index of every page, plus the cross-host network directory |
| `/sitemap.xml` | Every non-hidden page |
| `/robots.txt` | Per-bot-class access policy |
| `dash.mcp` | The same prose as an MCP resource (Dash 4.3+) |

### Features

*   **Universal prerender** — crawler-ready static HTML for every page, with the interactive app untouched.
*   **`llms.txt` everywhere** — a Markdown twin of every page, plus a rendered viewer for humans who open it in a browser.
*   **Multi-backend** — Flask, FastAPI and Quart, auto-detected from `app.server`.
*   **Per-request access control** — visibility verdicts (`allow`/`gated`/`deny`) applied consistently across llms.txt, crawler HTML, prerender, sitemap and MCP.
*   **Cross-host network directory** — publish sibling sites so an agent landing on one host can find the rest; this is how the [2plot network](https://2plot.ai) hangs together.
*   **Bot management** — block AI-training crawlers while allowing AI-search citations, configurable per bot class.