# The algo records whose domain is mode-driven-response-schema

This index as JSON: https://banes-lab.com/json/api/facets/algo/domain/mode-driven-response-schema

## Entries

- [Composed Turn Contract](https://banes-lab.com/records/algo/composed-turn-contract.md): Compose a model turn's validation schema and its instruction from one self-registering field registry, projected per mode, so the schema the response is graded against and the instruction the model is given can never drift.
- [Loop-Owned Mode Selection](https://banes-lab.com/records/algo/loop-owned-mode-selection.md): The loop assigns the (phase, activity) mode that selects which contract a response is validated against; the model's output carries no contract-selecting field, so a model can never grade itself against a contract it chose.
- [Versioned Turn Provenance](https://banes-lab.com/records/algo/versioned-turn-provenance.md): Persist each accepted model turn as an append-only, mode-tagged, schema-versioned, causally ordered record with change-only provenance, so the turn history is a queryable, auditable log and a turn is never graded against a schema version its stored state never saw.
- [Mode Contract Validation](https://banes-lab.com/records/algo/mode-contract-validation.md): Validate the model's response against the loop-selected mode contract and, on a validation miss, drive an error-specific self-healing retry bounded by the contract, so a response is accepted only when it satisfies the schema it is graded against.
- [<Mode-Driven Response Schema>](https://banes-lab.com/records/algo/mode-driven-response-schema.md): Make the model's structured output a mode-driven composed contract — the output mirror of composed input context: per loop-owned mode the schema and instruction are projected from one field registry, the response is validated and self-heals against it, the typed fields are governed, and every turn is persisted as a versioned queryable record — so the output has one source, cannot drift from its instruction, and can be audited.