Start here · 10 minutes

Model your first corpus

Validate a declarative corpus configuration, inspect its planned storage, and prepare normalized records for ingestion.

What you will build

A CoreCortex corpus starts as an authored configuration: datasets, transformation stages, storage targets, indexes, and retrieval plans in one reviewable file.

The current alpha is for local development and evaluation. It validates and plans your memory system; hosted operation is not yet generally available.

Prerequisites

  • Rust 1.97.1 toolchain
  • PostgreSQL for control-plane and lifecycle operations
  • Optional Turbopuffer and Ollama credentials for the end-to-end vector demo

Install the corec CLI

CoreCortex uses corec as its compact command name. The examples below assume the executable is available on your PATH.

shellCoreCortex Core
cargo install --path crates/corec-cli --locked
corec --version

Validate the example corpus

Validation catches structural and semantic errors before CoreCortex writes control-plane state. Schema output lets editors provide completion and inline diagnostics.

shellCoreCortex Core
corec validate examples/fda-labels/corec.yaml
corec schema --output /tmp/corec.schema.json

See the plan before you apply it

The plan explains which retrieval strategy, index, filters, limits, and fallback path would be used. The query layer returns ranked evidence; answer synthesis remains an application concern.

shellCoreCortex Core
corec demo query-plan \
  --config examples/fda-labels/corec.yaml \
  --query "Which warnings mention liver injury?"
Continue readingUnderstand the model