buildEvaluationInstructions method

@override List<ChatMessage>? buildEvaluationInstructions(List<ChatMessage> messages, ChatResponse modelResponse, List<EvaluationContext> additionalContext)

Builds the evaluation instructions (system + user messages).

Return null to signal that a required context was missing.

Implementation

@override
List<ChatMessage>? buildEvaluationInstructions(
  List<ChatMessage> messages,
  ChatResponse modelResponse,
  List<EvaluationContext> additionalContext,
) {
  final ctx = additionalContext.whereType<ToolCallAccuracyEvaluatorContext>().firstOrNull;
  if (ctx == null) return null;

  final toolCalls = modelResponse.messages
      .expand((m) => m.contents)
      .whereType<FunctionCallContent>()
      .map((c) => '- ${c.name}(${c.arguments})')
      .join('\n');
  final toolDefs = ctx.contents.map((c) => c.toString()).join('\n');
  final userRequest = messages.lastUserMessage?.text ?? '';

  final prompt = '''
# Definition
**Tool Call Accuracy** measures how accurately the AI used available tools — whether the right tools were called, with correct parameter names, and accurate values extracted from the conversation.

AVAILABLE TOOLS:
$toolDefs

# Ratings
## [ToolCallAccuracy: 1] Wrong tools called or completely incorrect parameters.
## [ToolCallAccuracy: 2] Partially correct tool calls with significant parameter errors.
## [ToolCallAccuracy: 3] Correct tools called but with some parameter errors.
## [ToolCallAccuracy: 4] Correct tools called with mostly correct parameters.
## [ToolCallAccuracy: 5] Perfectly accurate tool calls with all correct parameters.

# Data
QUERY: $userRequest
TOOL CALLS MADE:
${toolCalls.isEmpty ? "(none)" : toolCalls}

# Tasks
## Score the accuracy of the tool calls.
- **ThoughtChain**: Think step by step. Start with "Let's think step by step:".
- **Explanation**: A very short explanation of why you think the input Data should get that Score.
- **Score**: An integer score (1–5) based on the definitions.

## Please provide your answers between the tags: <S0>your chain of thoughts</S0>, <S1>your explanation</S1>, <S2>your Score</S2>.
# Output
''';
  return [
    ChatMessage.fromText(ChatRole.system, _systemPrompt),
    ChatMessage.fromText(ChatRole.user, prompt),
  ];
}