distill::cascade::TeacherCascade¶

Public Functions¶

	Name
	init(self self, teacher_client teacher_client, benchmark_table benchmark_table, level1_model level1_model, confidence_threshold confidence_threshold)
	execute(self self, messages messages, domain domain, ** kwargs)

Protected Attributes¶

	Name
	_teacher
	_benchmark_table
	_level1_model
	_confidence_threshold

Detailed Description¶

class distill::cascade::TeacherCascade;

Multi-teacher cascade orchestrator with benchmark-routed escalation.

Constructor arguments map directly to config values so that
``TeacherClient.__init__`` can wire them from ``pipeline.yaml``::

    cascade = TeacherCascade(
        teacher_client=self,
        benchmark_table=config["teacher_benchmark"],
        level1_model=config["teacher"]["level1"],
        confidence_threshold=config["teacher"]["confidence_threshold"],
    )

Public Functions Documentation¶

function init¶

__init__(
    self self,
    teacher_client teacher_client,
    benchmark_table benchmark_table,
    level1_model level1_model,
    confidence_threshold confidence_threshold
)

Initialise the cascade orchestrator.

Args:
    teacher_client: A ``TeacherClient`` instance whose
        ``generate_with_logprobs()`` method is used for all
        teacher calls within the cascade.
    benchmark_table: The ``teacher_benchmark`` dict from
        ``pipeline.yaml`` — domain key → ``{model: score}``.
    level1_model: The always-first teacher model name
        (e.g. ``"deepseek-v4-fast"``).
    confidence_threshold: Minimum logprobs confidence
        (0.0–1.0) to avoid Level 2 escalation.

function execute¶

execute(
    self self,
    messages messages,
    domain domain,
    ** kwargs
)

Run the confidence-gated teacher cascade.

Args:
    messages: List of message dicts (OpenAI format).
    domain: Specialist niche name (e.g. ``"code"``, ``"medical"``).
        Mapped to a benchmark table key via ``_DOMAIN_MAP``.
    **kwargs: Extra parameters forwarded to
        ``TeacherClient.generate_with_logprobs()``.

Returns:
    ``CascadeResult`` with the best-available response.

Raises:
    TeacherConfigError: If every teacher in the cascade raises an
        exception (no response could be produced).

Protected Attributes Documentation¶

variable _teacher¶

_teacher;

variable _benchmark_table¶

_benchmark_table;

variable _level1_model¶

_level1_model;

variable _confidence_threshold¶

_confidence_threshold;

Updated on 2026-07-25 at 22:56:57 +0000