CoreCortex documentation
Start with the corpus: the information, purpose, datasets, and source boundaries. Read representative records and collect real questions. Then follow the tutorial from source files to a measured search configuration.
Start with a corpus
Name the information you want to search, its purpose, its datasets, and the people who understand it.
Getting started →02Learn the concepts
Use one glossary for corpus, records, passages, indexes, retrieval, models, evaluation, and promotion.
Concepts →03Choose indexes and schema
Use the data and questions to choose keyword, vector, filter, relational, or graph search.
Indexes and schema →04Build the tutorial corpus
Model reference guides and change records as one corpus, then retrieve across both datasets.
Two-dataset walkthrough →05Improve retrieval
Compare document conversion, chunks, embedding models, search methods, and result counts.
Retrieval experiment guide →06See the Enterprise-assisted path
Inspect real input and output artifacts from classification through certification and Agent Skills.
Verified Enterprise tutorial →Follow one example from source files to tested retrieval
First define the corpus and inspect its datasets. Then collect real questions, create records and searchable passages, build an index, and run the questions. Inspect the misses before changing the configuration.
The tutorial uses two datasets. Reference guides contain prose. Change records contain dates, categories, and short notes. You will use keyword search, vector search, filters, and a query that needs both datasets.
Enterprise can read JSONL or Apache Parquet exports. It writes records, candidate test questions, and miss reports as files you can inspect and review in Git.