Skip to content

runner

diapason.learning.spec_search.diagnose.runner

DiagnosisRunner: orchestrates phase 1 of the spec-search loop.

Builds diagnostic tools, runs the TeacherAgent, parses failure clusters from the teacher's output, and persists artifacts.

See spec §5.

DiagnosisResult dataclass

The output of a diagnosis run.

DiagnosisRunner

Orchestrates phase 1 of the spec-search loop.

Parameters

teacher_engine : The CloudEngine (or mock) for teacher inference. teacher_model : Frontier model id (e.g. "claude-opus-4-6"). trace_store : TraceStore for reading student traces. benchmark_samples : List of PersonalBenchmarkSample objects. student_runner : Callable to re-execute the student on a task. judge : TraceJudge for comparing outputs. session_dir : Path where session artifacts are written. session_id : Current session id. config : Dict with config_path and diapason_home. max_turns : Max teacher tool calls (default 30). max_cost_usd : Max teacher API cost (default 5.0).

run()

Execute the diagnosis phase.

Returns

DiagnosisResult Contains the diagnosis markdown, parsed clusters, cost, and tool call records.