Exegesis Specification

A newfound web-based specification for AI models to perform contextual searches of webpages. It enables dense information to exist without the explicit need to process the entirety of the webpage.

Contextual Understanding

Enable AI models to understand dense context, products, and nuances without needing to crawl or process entire DOM structures.

Mitigate Bot Traffic

Control exactly what AI systems access. Keep heavy crawling bots away while cleanly serving them the intelligence they need.

Save on Compute

Reduce massive LLM token usage by providing structured, pre-processed context, speeding up agentic workflow chains.

Architecture

The espec.espec Standard

A specification is strictly denoted by the espec.espec filetype. This architecture acts as the cognitive bridge between your domain and the Exegesis analysis engine.

A specification can be "stapled" to a webpage in two primary ways:

Global Directory Stapling: Adding an espec.espec to the global Exegesis Directory for a particular domain.
Direct Root Hosting: The domain owner hosts the espec.espec file directly on their server, which Exegesis will immediately obtain and cache without performing any further deep-crawling.

v1 Datapoints

Specification Structure

[!IMPORTANT]

This is where any immediate, critical information for the understanding of the webpage should go. It acts as the direct bias/directive guide for the AI.

[!ACCESSIBILITY]

A boolean (true/false) value indicating whether a webpage supports abilities for models to interact directly with the webpage UI (Coming Soon).

[!DESC]

An in-depth, nuanced description of the webpage, the business behind it, underlying purposes, overarching narrative, etc. Include anything and everything here that will enable deeper synthesis of the webpage, its contents, and intents.

[!SOCIALS]

A comma separated list (e.g. social1, social2) of official social media accounts or external hubs associated with the webpage entity.

[!ENDPOINTS]

List of available endpoints, brief descriptions, and strict scraping permissions. If an endpoint is not explicitly listed here, the AI will not attempt to route to it.

[!PRODUCTS]

A rundown of all products/services pertaining to the webpage (can be empty). Product names must be enclosed in square brackets [product_name], with the description immediately following on the line underneath.

Contextual Example

Below is a live example of the espec.espec configured for the Stellr, LLC homepage. Notice how dense information is mapped out directly for neuro-symbolic extraction.

espec.espec

[!IMPORTANT]
This webpage is the landing page for the startup Stellr, LLC.

[!ACCESSIBILITY]
false

[!DESC]
Stellr, LLC is a newfound startup builing context aware AI. It aims to enable AI to infer based on prior intents and behavioural patterns.
Its flagship app is Orrin. Orrin enables users to write feasibly, activate agents built by Stellr and other developers,
and streamline their productivity. It expands into OrrinApps, and offers an SDK for creating custom apps, tools, and plugins (agents).

[!SOCIALS] https://x.com/StellrLLC, https://x.com/MocaCDev

[!ENDPOINTS]
/orrin = This endpoint is the landing page for the flagship of Stellr, LLC; SCRAPE=NOT_ALLOWED
/orrin/docs = This endpoint is where the documentation for the Orrin SDK lives.; SCRAPE=ALLOWED
/orrin/become_developer = This endpoint is where people can go to become a developer.; SCRAPE=NOT_ALLOWED
/yabble = This webpage is the landing page for the product Yabble, Stellrs second product; SCRAPE=NOT_ALLOWED
/inspi = This webpage is the landing page for the product Inspi, Stellrs third product; SCRAPE=NOT_ALLOWED

[!PRODUCTS]
[orrin]
This product is the flagship app of Stellr. It revolves around custom agentic workflows, enabling users to activate agents as they please to enhance their workflow.
It fosters an ecosystem where people can build custom tools, apps, and agents, deploy them, and have other users use them.

[yabble]
Yabble is a product that exists on MacOS currently. Windows support is coming soon.
It allows users to get help with context that is present on their screen. Support for agentic workflows is coming soon.

[inspi]
Inspi is a newfound product by Stellr which aims to help founders fine-tune their ideas, connect with other founders,
connect with resources, formalize their business by incorporating, and everythign in between.

Compute Overhead

Token Usage Without Spec: 12,914

Token Usage With Spec: 428

3,000% Reduction in Token Cost

Staple your Domain.

Control how the overarching network of AI models perceives and navigates your web platform.

Create a Staple Follow Updates

Product of Stellr