Datasets:
roanbrasil
/
devops-gitops-corpus

Dataset card Data Studio Files
xet
 Community
1
text
stringlengths
0
59.1k
scorers.exactMatch; // output === expected
scorers.levenshtein; // edit distance (0-1 score)
// Numeric and data comparison
scorers.numericDiff; // normalized numeric difference
scorers.jsonDiff; // JSON object comparison
scorers.listContains; // list element matching
```
### LLM-Based Scorers
For LLM-based evaluation, use the native VoltAgent scorers that explicitly require a model:
```ts
import {
createAnswerCorrectnessScorer,
createAnswerRelevancyScorer,
createContextPrecisionScorer,
createContextRecallScorer,
createContextRelevancyScorer,
createModerationScorer,
createFactualityScorer,
createSummaryScorer,
} from "@voltagent/scorers";
import { openai } from "@ai-sdk/openai";
// Create LLM scorers with explicit model configuration
const answerCorrectness = createAnswerCorrectnessScorer({
model: openai("gpt-4o-mini"),
options: { factualityWeight: 0.8 },
});
const moderation = createModerationScorer({
model: openai("gpt-4o-mini"),
threshold: 0.5,
});
```
### Custom Scorers
For custom scoring logic, use `buildScorer` from @voltagent/core:
```ts
import { buildScorer } from "@voltagent/core";
const customLengthScorer = buildScorer({
id: "length-validator",
label: "Length Validator",
})
.score(({ payload }) => {
const output = String(payload.output || "");
const minLength = Number(payload.minLength || 10);
const valid = output.length >= minLength;
return {
score: valid ? 1.0 : 0.0,
metadata: {
actualLength: output.length,
minLength,
},
};
})
.reason(({ score, results }) => ({
reason:
score >= 1
? `Output meets minimum length of ${results.raw.minLength}`
: `Output too short: ${results.raw.actualLength} < ${results.raw.minLength}`,
}))
.build();
```
## Result Structure
`runExperiment` returns:
```ts
interface ExperimentResult {
runId?: string; // VoltOps run ID (if connected)
summary: ExperimentSummary; // aggregate metrics
items: ExperimentItemResult[]; // per-item results
metadata?: Record<string, unknown> | null;
}
```
### Summary
```ts
interface ExperimentSummary {
totalCount: number;
completedCount: number;
successCount: number; // items with status "passed"
failureCount: number; // items with status "failed"
errorCount: number; // items with status "error"
skippedCount: number; // items with status "skipped"
meanScore?: number | null;
passRate?: number | null;
startedAt: number; // Unix timestamp
completedAt?: number; // Unix timestamp
durationMs?: number;
scorers: Record<string, ScorerAggregate>; // per-scorer stats