Directories
¶
| Path | Synopsis |
|---|---|
|
Command benefitreport renders the benefit records under evals/results/benefit/ into docs/evals/benefit-report.md and docs/evals/benefit-scatter.svg.
|
Command benefitreport renders the benefit records under evals/results/benefit/ into docs/evals/benefit-report.md and docs/evals/benefit-scatter.svg. |
|
Command chartgen renders the durable selection-run artifacts under evals/results/ into the SVG charts embedded in README.md.
|
Command chartgen renders the durable selection-run artifacts under evals/results/ into the SVG charts embedded in README.md. |
|
internal
|
|
|
benefit
Package benefit is the Jev benefit benchmark's durable core (ADR-001 amendment A1): the frozen ground truth, the run record, the grader, and the statistics.
|
Package benefit is the Jev benefit benchmark's durable core (ADR-001 amendment A1): the frozen ground truth, the run record, the grader, and the statistics. |
|
evalserver
Package evalserver is the evaluate server under test: the real shipped binary, built from source, spoken to over MCP stdio, never an in-process stand-in.
|
Package evalserver is the evaluate server under test: the real shipped binary, built from source, spoken to over MCP stdio, never an in-process stand-in. |
|
modelclient
Package modelclient is the agent under test as the eval harnesses reach it: an OpenAI-compatible chat-completions endpoint, temperature 0, read as a stream so tool arguments arrive as the model wrote them.
|
Package modelclient is the agent under test as the eval harnesses reach it: an OpenAI-compatible chat-completions endpoint, temperature 0, read as a stream so tool arguments arrive as the model wrote them. |
|
runrecord
Package runrecord owns the durable selection-run record: the one definition of the evals/results/selection-*.json format, shared by the harness that writes it and chartgen that reads it.
|
Package runrecord owns the durable selection-run record: the one definition of the evals/results/selection-*.json format, shared by the harness that writes it and chartgen that reads it. |
|
Package selection implements the closed-world tool-selection eval harness (docs/evals/PLAN.md §4–§6).
|
Package selection implements the closed-world tool-selection eval harness (docs/evals/PLAN.md §4–§6). |
Click to show internal directories.
Click to hide internal directories.