Cortext.io

Mission of the Project

Preserve Indigenous Languages

by expanding Cortext to 20 other languages influencing "schema.org" to implement this Semantic Calculus across the web

In Short

Powerful LLMs are held by IBM, Google, Twitter, ChatGPT ...

These companies provide consultations on how to engineer language to best accomplish sophist goals. Cortext is a lightweight LLM that extracts "hashtags" and searches the most overlapping "Blue Words" in Wikipedia.

The Cortext LLM is more elegant than "Word2Vec" style Natural Language Processing, though crudely implemented... it is based on Wierzbicka's Semantic Primes, and has the potential to function in 20+ languages.

2025-05-25

Thank you Mat Allen for all the contribution!

Command to Run MCP

## create image
sudo docker build -t cortext-python310 .
sudo docker run -it -p 5000:5000 --name cortext-python310-fastmcp cortext-python310

## inside the image run
python3 ./cortext_io/api/app.py

Microservice Architecture I run Cortext LLM within

Command to run with text already in input.txt:

./text.sh [Title] [Email] [Email?]

Command to run in Batch Job:

["/bin/bash","-c","echo '[This is clean text that will not have encoding issues in a linux command line, and is also ~7800 characters]' > /home/ec2-user/cortext_io/cortext_io_input/input.txt && rm /home/ec2-user/cortext_io/cortext_io_input/input2.txt && /home/ec2-user/cortext_io/10K_RiskFactors_PROCESS/1_CORTEXT_Run.sh [Title] [Email] ['Yes'/'No']"]

You will need to replace the emails, usernames, and passwords in the following locations for this to actually send out content:

Purpose	Location
Send Email Report with SMTP	cortext_io/10K_RiskFactors_PROCESS/sendEmail.py
Push content to your ElasticSearch instance	cortext_io/10K_RiskFactors_PROCESS/zzz_pushElastic.py

Cortext LLM uses an SQLite database file for staging

I recommend you open the SQLite Database File in a tool like DBeaver to understand the data model or see the uml diagram here: https://www.cortext.io/how-it-works

Purpose	Location
Relational Data Structure for Text Decomposition	cortext_io/cortext_io_db/cortext_io.db

Track the Java Routines --> Look for .item

The LLM has gone through multiple languages and architectures (not to say they got progressively better). I landed on Talend... Search for the routines to see where the hashtag extraction takes place.

Purpose	Location
"Scripts" in Talend Workflow	cortext_io/AA_cortext_io_linux_0.1/AA_cortext_io_linux/items/integralmass/code/routines

Perceptron to Clean the Noise

A Python Perceptron is used to clean the noise, based on hand-selection of the best hashtag extractions from 1000 actuarial articles.

D3.js Associative Web Not Currently in Workflow

It is my current understanding that email bodies cannot have executable javascript, so the Associative Web (force-directed node graph) is just hanging there.

Home Website

https://cortext.io

Risk Wiki

Cortext ran over risk factors of 100s of Annual Reports from Public Companies

https://riskrunners.com

Semantic Calculus

Foundational Mathematics

https://perrydime.com/Begin_With_The_End_In_Mind.pdf

First Implementation

https://nlpbigdata.jeffersonrichards.com/

10 Year Project in AI

https://wiki.richards.systems

About Me

https://jefferson.cloud https://richards.plus https://richards.systems