A question/response chat-style service powered by Traditional Catholic Documents.

I am a TradGPT (tradgpt.ai), a question/response chat-style service powered by Traditional Catholic Documents.

You ask a question. It finds the most applicable sections of reference material from its database of stored tradtional Catholc documents (listed below) and then uses those references to answer your question.

Yes, TradGPT allows a line of questions within a conversation.

TradGPT will search its documents and find those that match most closely (which may have near zero applicability). Then it will likey determine that the answer cannot be found in those documents and reply with "I cannot determine the answer from my references."

Yes, the full text of all references used to respond are listed below the response.

Yes, if you go to Sources.Full Documents you can read the entire and exact database that is used in book format."

It uses chatGPT models to formulate natural language reponses constrained to a database of Traditional Catholic documents.

TradGPT limits its retrieval and reference to traditional Catholic documents. This curation can help ensure that the responses are consistent and factually grounded within that domain, reducing the risk of generating misinformation or off-topic, harmful content.

By confining its output to a vetted set of texts, the system is less likely to produce unexpected or inaccurate content. This controlled boundary can serve as a safeguard against the generative model "hallucinating" information that isn’t supported by its reference material.

The focused nature of the system means that it is designed to adhere closely to traditional Catholic teachings. In this sense, the system’s consistency acts as a safety feature — users receive answers that are in line with a predefined, well-understood framework.

Like any AI system, RAG systems can be targets for adversarial inputs.

Yes, just go to the Sources.Full page to select and read any source document, without using any AI.

All documents are pre-processed to allow them to be utilized in a RAG system (see below). This entails:

• stripping out all formatting
• "chunking" the document into semantically cohesive sections that are the "right" size
• processing each chunk to produce a set of unique lookup vectors
• storing the chunks and vectors in a database for easy access

The document view is simply the reconsitution of all the chunks back into a serial document, it is the same live chunk data used for searches. It retains a tiny amount of formatting due to line breaks and number lists that are preserved. The horizontal lines denote the boundaries of a chunk. It is adequate to read the document, but not as satisfying as professionally typset documents. In future it may be possible to add some simple markdown layout back into the chunks to improve this.

• Start by Signing Up as a new user.
• You can change your name and password later, but your email is the permanant sign-in key of your account
• An email verification link will be sent to your email. Be sure to check your junk folder.
• The Profile Page allows you to view and edit your profile
• You can upload a profile image, change your name, or reset your password
• After you login, going to the Home Page or clicking on the TradGPT icon will show your list of previous chats and you can start a new one.
• The About.Legal Page contains the User License, Privacy Policy, and Terms of use that you agree to as a user of TradGPT.

After you have an account and log in, just go to the home page a type in a question in the "How can I help?" field.

Yes, they are saved in your account. You can rename and delete any of your conversations.

Artificial Intelligence (AI) is a broad field of computer science dedicated to creating systems capable of performing tasks that normally require human intelligence.

AI involves the development of algorithms and models that enable computers to perform tasks such as reasoning, learning, perception, language processing, and problem solving. It ranges from simple rule-based systems to complex neural networks that mimic aspects of human thought.

ChatGPT is an advanced conversational agent (or chatbot) developed by OpenAI that uses state-of-the-art natural language processing to understand and generate human-like text responses.

ChatGPT is based on the transformer model architecture, which is a neural network design that excels in handling sequential data like natural language. The transformer architecture enables the model to capture complex relationships between words, allowing it to generate contextually relevant responses.

Retrieval-Augmented Generation (RAG) is an advanced technique in natural language processing (NLP) that combines the strengths of information retrieval systems with those of generative language models. By marrying these two components, RAG systems enhance the quality and accuracy of generated content, particularly for tasks that require up-to-date or domain-specific information.

TradGPT is a retrieval-augmented generation (RAG) system integrated with ChatGPT. It combines ChatGPT’s natural language generation capabilities with a specialized retrieval mechanism that only accesses traditional Catholic documents supported by TradGPT. This setup effectively limits the system's "memory" to these curated texts, ensuring that all responses are grounded solely in the established traditional Catholic sources. As a result, TradGPT consistently delivers answers that align with these documents, maintaining both content integrity and relevance.

TradGPT currently uses the text-embedding-3-small model to generate embedding vectors, and the gpt-5-mini as the default completion model. Other models may also be selected.

• The Bible, Douay-Rheims
• A Catechism of Christian Doctrine (The Baltimore Catechism No. 3)
• An Explanation Of The Baltimore Catechism of Christian Doctrine (The Baltimore Catechism No. 4)
• The Catechism of St. Pope Pius X
• Prayers of the Catholic Faith - English/Latin
• The 1917 OR Pio-Benedictine Code of Canon Law
• Summa Theologica, Part I (Prima Pars)
• Summa Theologica, Part I-II (Pars Prima Secundae)
• Summa Theologica, Part II-II (Secunda Secundae)
• Summa Theologica, Part III (Tertia Pars)
• First Council of Nicaea
• First Council of Constantinople
• The Council Of Ephesus
• The Council of Chalcedon
• Second Council of Constantinople
• Third Council of Constantinople
• Second Council of Nicaea
• Fourth Council of Constantinople
• First Lateran Council
• Second Lateran Council
• Third Lateran Council
• Fourth Lateran Council
• First Council of Lyons
• Second Council of Lyons
• Council of Vienne
• Council of Constance
• Council of Basel
• Fifth Lateran Council
• General Council of Trent
• First Vatican Council
• ...and more coming

TradGPT attempts to stick to traditional Catholic documents that are the foundation of Catholic faith. More will be added over time, but it will remain wholly based in classic Catholic tradition.

Yes, you can use the filter feature when you ask your question. By formulating your question as filter:question you can scope the documents that TradGPT uses to reference its answer. TradGPT will only use documents that contain filter in the title (case insensitive).

So, "trent" will only use "General Council of Trent", and "council" will use all the council documents. "summa" will use all four parts of "Summa Theologica" "bible" will use "The Bible, Douay-Rheims" "catechism" will use both "A Catechism of Christian Doctrine (The Baltimore Catechism No. 3)" and "An Explanation Of The Baltimore Catechism of Christian Doctrine (The Baltimore Catechism No. 4)"

Some full examples using the filter feature:

• trent: what does trent say about nuns?
• bible: what does the church teach on marriage?
• catechism: what does the church teach on marriage?
• council: what does the church teach on marriage?
• summa: what does the church teach on marriage?
• First Council of Nicaea: summarize the first council of Nicaea
• canon: can a pope excommunicate himself?

Wait, does this mean you cannot put a semicolon in your question that is not a filter? yes, the first semicolon is treated as the filter separator.

• First Commit

• First Public Hosted Release

• Preserve context per chat for full conversation, add self (service) awareness

• Conversation rename and delete.

• Click on reference opens highlighted in document.

• Topic only vectors, regex hybrid id lookup for daily missal.

• Update models.