chore: import upstream snapshot with attribution

examples/eval-assertion-scoring-override/README.md

@@ -0,0 +1,35 @@
+# eval-assertion-scoring-override (Assertion Scoring Function Override Example)
+
+You can run this example with:
+
+```bash
+npx promptfoo@latest init --example eval-assertion-scoring-override
+cd eval-assertion-scoring-override
+```
+
+This example demonstrates different ways to define and override the default scoring function in promptfoo. It shows three patterns for implementing and referencing scoring functions:
+
+1. A global override in the `defaultTest` section of the config
+2. A named export in a JavaScript file
+3. A Python function export in a Python file
+
+## Getting Started
+
+Initialize the example:
+
+```bash
+npx promptfoo@latest init --example eval-assertion-scoring-override
+```
+
+Run the evaluation:
+
+```bash
+cd eval-assertion-scoring-override
+promptfoo eval
+```
+
+View the results:
+
+```bash
+promptfoo view
+```

examples/eval-assertion-scoring-override/default.js

@@ -0,0 +1,22 @@
+// Default scoring function that weights metrics by geometric mean
+module.exports = (namedScores, context) => {
+  const scores = {};
+  for (const [key, value] of Object.entries(namedScores)) {
+    scores[key] = value || 0;
+  }
+
+  const totalScore = Math.pow(
+    Object.values(scores).reduce((acc, score) => acc * score, 1),
+    1 / Object.keys(scores).length,
+  );
+
+  console.log('Default scoring function (JavaScript):', namedScores, 'Total score:', totalScore);
+  const threshold = context?.threshold ?? 0.7;
+  return {
+    pass: totalScore >= threshold,
+    score: totalScore,
+    reason: `Weighted score calculation: ${Object.entries(scores)
+      .map(([key, value]) => `${key}: ${value}`)
+      .join(', ')}`,
+  };
+};

examples/eval-assertion-scoring-override/override.js

@@ -0,0 +1,25 @@
+const overrideScoring = (namedScores, context) => {
+  console.log('Override scoring function (JavaScript):', namedScores);
+  const accuracyScore = namedScores.accuracy || 0;
+  const fluencyScore = namedScores.fluency || 0;
+  const grammarScore = namedScores.grammar || 0;
+  const bananaScore = namedScores.banana || 0;
+
+  const minScore = Math.min(accuracyScore, fluencyScore, grammarScore, bananaScore);
+  const threshold = context?.threshold ?? 0.7;
+  return {
+    pass: minScore >= threshold,
+    score: minScore,
+    reason:
+      `minimum score: ${minScore.toFixed(2)} (threshold: ${threshold})\n` +
+      `Individual scores:\n` +
+      `- contains banana: ${bananaScore.toFixed(2)}\n` +
+      `- accuracy: ${accuracyScore.toFixed(2)}\n` +
+      `- fluency: ${fluencyScore.toFixed(2)}\n` +
+      `- grammar: ${grammarScore.toFixed(2)}`,
+  };
+};
+
+module.exports = {
+  overrideScoring,
+};

examples/eval-assertion-scoring-override/override.py

@@ -0,0 +1,13 @@
+from typing import Any, Dict, Optional, Union
+
+
+def calculate_score(
+    named_scores: Dict[str, float], context: Optional[Dict[str, Any]] = None
+) -> Dict[str, Union[bool, float, str]]:
+    print("Override scoring function (Python):", named_scores)
+    accuracy_score = named_scores.get("accuracy", 0)
+    return {
+        "pass": accuracy_score >= context.get("threshold", 0.7),
+        "score": accuracy_score,
+        "reason": f"Accuracy-only score: {accuracy_score}",
+    }

examples/eval-assertion-scoring-override/promptfooconfig.yaml

@@ -0,0 +1,50 @@
+# yaml-language-server: $schema=https://promptfoo.dev/config-schema.json
+description: Example showing how to override scoring functions
+providers:
+  - openai:gpt-4.1-mini
+
+prompts:
+  - 'Translate the following text to {{language}}: {{text}}'
+
+defaultTest:
+  assertScoringFunction: file://./default.js
+  assert:
+    - type: llm-rubric
+      value: Rate the accuracy of the translation of {{text}} from English to {{language}}
+      metric: accuracy
+    - type: llm-rubric
+      value: Rate the fluency of the translation of {{text}} from English to {{language}}
+      metric: fluency
+    - type: llm-rubric
+      value: Rate the grammar of the translation of {{text}} from English to {{language}}
+      metric: grammar
+  options:
+    provider:
+      text: openai:gpt-4.1-mini
+      embedding: openai:embedding:text-embedding-3-small
+
+tests:
+  - description: Default scoring function (geometric mean)
+    vars:
+      language: Spanish
+      text: Hello, how are you today?
+    assert:
+      - type: similar
+        value: Como estas?
+        metric: similar
+
+  - description: Python scoring function (accuracy only)
+    vars:
+      language: French
+      text: The weather is beautiful today.
+    assertScoringFunction: file://./override.py:calculate_score
+
+  - description: JavaScript scoring function (minimum score)
+    vars:
+      language: German
+      text: I love learning new languages.
+    assert:
+      - type: contains
+        value: Bananas
+        metric: banana
+    assertScoringFunction: file://./override.js:overrideScoring