How It Works

Plurilore uses Retrieval-Augmented Generation (RAG). Instead of asking a language model to answer only from its general training, the system first searches a selected sacred-text corpus and supplies relevant passages to the model. The result is still generated text, but it is generated with a visible source set.

From question to answer

1. Selection: You choose one or more traditions. That choice limits which retrieval targets can supply passages.
2. Retrieval: The question is compared with indexed corpus material to find passages that are semantically relevant, even when the wording is not an exact keyword match.
3. Generation: The retrieved passages and the recent context from the current page session are sent to the language model.
4. Provenance: Structured source references are rendered in the Sources consulted panel and linked to the Library.

For comparative questions, each selected tradition retrieves from its own corpus. The answer can then discuss similarities and differences in the retrieved evidence without merging the underlying collections.

What a citation establishes

A citation establishes that a particular source passage was supplied during generation. Opening it lets you inspect the text and its immediate context. Citations are especially useful for checking quotations, references, and whether the answer's summary matches the source.

A citation does not prove that the model interpreted the passage correctly, found every relevant passage, or represented every school within a tradition. Retrieval can miss material, translations carry editorial choices, and generated prose can overstate an inference.

Product boundaries

Chat context lasts for the current page session rather than becoming a permanent conversation record. Plurilore searches its curated corpus, not the open web, and it does not act as clergy, a religious authority, or a provider of medical, legal, mental-health, or pastoral advice. See Limits and Verification for a practical review checklist.