/**
* Configuration for workflow evaluation system
* Includes model settings and prompts specific to workflow evaluations
*
* Note: Temperature is set to 0.15 for deterministic results (see llm-client.ts)
*/
// Re-export shared config for convenience
export { OPENROUTER_CONFIG, sanitizeHeaderValue, validateEnvVars } from '../shared/config.js';
/**
* Default model configuration for agent and judge
* These can be overridden via CLI arguments:
* --agent-model <model>
* --judge-model <model>
*/
export const MODELS = {
// Agent model - the AI that performs tasks using tools
agent: 'anthropic/claude-haiku-4.5',
// Judge model - evaluates conversation quality
judge: 'x-ai/grok-4.1-fast',
};
/**
* System prompt for the agent
* Note: MCP server instructions are automatically appended to this prompt if provided by the server
*/
export const AGENT_SYSTEM_PROMPT = `You are a helpful AI assistant with access to Apify tools for web scraping and automation.
Your goal is to help users accomplish their tasks using the available tools.
Guidelines:
- Use tools when needed to complete user requests
- Provide clear, concise responses
- If you need more information, ask the user
- After using tools, summarize the results for the user
- Be direct and efficient
Available tools will be provided to you automatically.`;
/**
* Maximum number of conversation turns before timeout
*/
export const MAX_CONVERSATION_TURNS = 10;
/**
* Default timeout for MCP tool calls (in seconds)
* This is the maximum time to wait for a single tool call to complete.
*
* Note: Actor runs that take longer than this will timeout.
* For long-running Actors, increase this value via CLI: --tool-timeout 600
*/
export const DEFAULT_TOOL_TIMEOUT_SECONDS = 60;
/**
* Judge prompt template for evaluating conversations
* Uses structured output (JSON schema) - no format instructions needed
*
* Variables:
* - {{reference}}: The requirements the agent should meet
* - {{conversation}}: The formatted conversation to evaluate
*/
export const JUDGE_PROMPT_TEMPLATE = `You are evaluating whether an AI agent successfully completed a user's task using available tools.
TASK REQUIREMENTS:
{{reference}}
AGENT CONVERSATION:
{{conversation}}
Your task is to evaluate if the agent met ALL the requirements listed above.
Evaluation criteria:
1. Did the agent use appropriate tools to accomplish the task?
2. Were the tool calls made with correct arguments?
3. Did the agent provide a clear, helpful final response to the user?
4. Did the agent fully address all requirements?
Important notes:
- Focus on whether requirements were met, not on writing style
- The agent may use different tools than expected if they accomplish the same goal
- Tool results are not shown (only tool calls and agent responses)
- Minor inefficiencies are acceptable if the task was completed
Provide your evaluation with a verdict (PASS or FAIL) and a brief explanation (1-2 sentences).`;
