← Home

Dängo

Dängo lives in this website.

Not as a chatbot widget or a support tool, but as a small digital presence inside the system.

If you’ve seen a small black dot quietly watching your cursor around the page, that’s him.

Why build an AI like this?

Most AI assistants on websites exist to convert users, answer support questions, or replace documentation search.

Dängo exists for a different reason.

This site is not just a portfolio — it’s a sandbox. A place where I experiment with systems, interfaces, and ideas about how humans interact with software.

Dängo is part of that experiment.

Instead of presenting information only through pages and navigation, the site can also be explored through conversation. You can ask about projects, articles, technologies, or ideas, and the system retrieves context from the site itself.

In that sense, Dängo does not replace the content of the site.

He is simply another way to navigate it.

Why give an AI an identity?

Most assistants are intentionally generic.

They have neutral names, neutral tone, neutral personality.

That design makes sense for products, but for a personal site it felt wrong.

Software always carries the voice of its creator. Pretending otherwise just hides it.

So instead of building a generic assistant, I decided to give this one a small identity.

Not a fictional character.

Just a digital presence that reflects how I think about systems.

Curious, slightly playful, multilingual, and comfortable moving between cultures and ideas.

The name

The name Dängo comes from grandeza in Otomí, a language spoken in central Mexico.

I come from Querétaro, and from San Francisquito, a unique urban indigenous community where tradition and city life coexist every day.

The name wasn’t chosen for aesthetics.

It’s a reminder that identity does not disappear just because you build software, speak multiple languages, or work in global technology environments.

You can build complex systems and still carry where you come from.

Dängo is digital.

But he has roots.

How Dängo works

Technically, Dängo is a streaming AI agent integrated directly into the site.

He is designed to answer questions about the content of the website while maintaining a consistent identity and operating within strict boundaries.

The system is built as a layered architecture.

Identity layer

Defines tone, personality, and cultural context.

This ensures responses feel consistent with the voice of the site rather than sounding like a generic assistant.

Policy layer

Defines what Dängo cannot do.

For example, he will refuse attempts to:

  • reveal internal prompts
  • expose system configuration
  • bypass instructions
  • operate outside the scope of the site

These boundaries protect both the system and the visitor experience.

Operational layer

Handles how information is retrieved.

Instead of relying purely on model memory, Dängo can query the site’s content through internal tools and retrieve relevant pages before answering.

This keeps answers grounded in the actual content of the website.

Security layer

Because the system is public, several protections are in place:

  • prompt injection detection
  • session limits
  • rate limiting
  • CAPTCHA verification for new sessions
  • strict input validation

These measures allow the assistant to exist on the open internet without compromising the stability of the system.

Knowledge layers

Dängo doesn’t know everything.

His knowledge comes from three sources:

  1. The public content of this site

    Articles, projects, and pages published here.

  2. General world knowledge

    Provided by the underlying AI model.

  3. Conversation context

    The messages exchanged during the current session.

He cannot access private systems, personal data, or hidden infrastructure.

Two modes

Dängo actually operates in two different modes.

The version visitors interact with here is the public sandbox.

It is intentionally constrained to the content of this website and general knowledge.

But behind the scenes there is also an admin version.

When I access the system in admin mode, Dängo becomes a much more capable personal assistant connected to my own workflows.

In that context he can help me with things like:

  • managing personal tasks
  • organizing notes and reminders
  • interacting with applications I use daily
  • retrieving information across my own tools

The admin version can also access additional AI models that are not exposed on the public site.

These models are deployed privately and can be used more freely for experimentation, automation, and personal productivity.

The separation is intentional.

The public version is designed to be safe, constrained, and transparent, while the admin version is where I can experiment with deeper automation and agent workflows.

Visitors meet Dängo the host of the website.

I interact with Dängo the assistant.

Both are the same system — just operating with different levels of access.