CoreCortex Enterprise
Tools for preparing data, running retrieval experiments, and approving a configuration for production.
Enterprise capabilities
- Maintained source connectors, refresh schedules, and pipeline operations
- Schema and storage discovery from a representative corpus sample
- Human review workflows that keep generated integration checks separate from golden evidence
- A–E experiments across conversion, chunking, embeddings, search method, and result count
- Promotion gates based on measured quality, latency, cost, and policy
- Operational monitoring, governance, and enterprise controls
Enterprise retrieval experiments
The controller reads the candidate settings and schedules the A–E runs. Every comparable run uses the same corpus snapshot and reviewed questions. The final run uses the selected converter, chunks, embedding model, search method, and result count.
The controller saves the index-build receipt and test report for each run. It blocks approval when access tests fail, files change between runs, or a report belongs to a different configuration.
| CoreCortex Core | CoreCortex Enterprise |
|---|---|
| Public corpus and evidence contracts | Search-space and experiment planning |
| Provenance, access, adapters, retrieval plans | Converter and chunk audits; isolated projection receipts |
| JSONL evaluation files | Stage selection and final report assembly |
| Rebuildable normalized records | Final reports and production approval checks |
Documentation boundary
Public Enterprise documentation describes the product contract: commands, required inputs, machine-readable outputs, workflow states, quality gates, developer-agent skills, runtime query routing, API behavior, security expectations, and the verified onboarding path. Prospective customers can evaluate fit before receiving the licensed distribution, and delivery teams can prepare integrations against stable boundaries.
The licensed implementation retains the proprietary value: classification and schema-advice heuristics, enrichment rules, experiment-selection logic beyond the published methodology, internal prompts, connector operations, administrative automation, observability internals, and engagement-derived tuning knowledge. Customer data, customer schemas, and field examples remain outside public documentation.
Learn and evaluate Enterprise
Availability
Good design-partner fit
- You operate a consequential corpus whose answers need defensible evidence.
- Your ingestion sources or access rules change often.
- You need to compare multiple storage and retrieval strategies.
- You already have domain experts who can judge useful evidence.