Skip to content

fix(eval): apply experiment config to the runtime and record requested/effective (#114) #116

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
helsome merged 2 commits into from
Sep 18, 2026
Merged

fix(eval): apply experiment config to the runtime and record requested/effective (#114) #116

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
25fd95f
fix(eval): apply experiment config to the runtime and record requeste…
qxr666-sga Sep 17, 2026
d1c9830
fix(eval): split the CLI model shorthand and stop labelling traces wi…
qxr666-sga Sep 18, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
31 changes: 31 additions & 0 deletions packages/core/src/evaluation.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -444,6 +444,32 @@ export interface EvaluationScore {

export type EvaluationRunStatus = 'completed' | 'failed' | 'cancelled' | 'timeout' | 'skipped';

/**
* A requested experiment-config dimension that has NO runtime control surface
* and was therefore NOT applied (#114). Recorded explicitly — never silently
* dropped, and never presented as if it were in effect.
*/
export interface UnappliedConfigItem {
key: 'model' | 'provider' | 'thinkingLevel' | 'strategyId';
reason: string;
}

/**
* Runtime configuration confirmed by READBACK after application (#114).
* Only values observed from the runtime's own state land here; a requested
* value without readback proof never masquerades as effective. Runs recorded
* before #114 simply omit this field (historical unknown — never backfilled).
*/
export interface EffectiveRuntimeConfig {
model?: string;
provider?: string;
thinkingLevel?: string;
/** Requested dimensions that could not be applied, with the reason why. */
unapplied?: UnappliedConfigItem[];
/** Epoch ms of the runtime readback that confirmed these values. */
confirmedAt: number;
}

export interface EvaluationRun {
id: string;
experimentId: string;
Expand All @@ -459,6 +485,11 @@ export interface EvaluationRun {
failureModes: EvaluationFailureMode[];
traceRef?: TraceReference;
error?: ApiError;
/**
* Runtime config actually in effect for this run, confirmed by readback
* (#114). `undefined` on historical records = unknown, not "default".
*/
effectiveConfig?: EffectiveRuntimeConfig;
}

/** Backend the trace lives in; `none` when observability is off (spec §89). */
Expand Down
Loading
Loading