extract-candidates

To extract knowledge candidates from raw evidence, you provide an in-memory Evidence Bundle to trigger Stage B-2 extraction. The process analyzes the bundle to identify confirmed decisions, anti-patterns, and established conventions, transforming them into structured vault-entry candidates.

An Evidence Bundle is an in-memory collection of raw evidence gathered during the preceding collection stage. It serves as the required input for Stage B-2 extraction, holding the unstructured data that will be analyzed and transformed into structured vault-entry candidates.

You validate domain conflicts during knowledge candidate extraction by using the integrated knowledge-gate CLI. This command-line interface performs domain validation and conflict checks on the extracted data before outputting the final structured vault-entry candidate objects.

No, the knowledge extraction process requires zero external dependencies to function. It operates directly on the in-memory Evidence Bundle and utilizes the built-in knowledge-gate CLI for domain validation and conflict checks to output candidate objects.

You need to transform evidence into vault-entry candidates when you want to populate a Knowledge Vault with confirmed team decisions and anti-patterns from incidents. This Stage B-2 extraction step guides future coding decisions by structuring raw evidence into actionable candidate objects.

You use the batch-refinement orchestrator in conjunction with the extraction process to feed the Knowledge Vault. The extractor outputs structured candidate objects, which the orchestrator then processes to systematically populate the vault with validated knowledge entries.