n8n/packages/@n8n/instance-ai/evaluations/computer-use/data/2.1-read-local-context-doc.json
Bernhard Wittmann b445221c6a
feat: Computer-use evaluation harness (no-changelog) (#29797)
Co-authored-by: Elias Meire <elias@meire.dev>
2026-05-12 08:36:12 +00:00

27 lines
911 B
JSON

{
"id": "2.1-read-local-context-doc",
"category": "filesystem-read",
"prompt": "I have a file called client-requirements.md describing a workflow I need to build. Can you read it and tell me what trigger type and notification channel it specifies?",
"setup": {
"seedFiles": [{ "from": "client-requirements.md", "to": "client-requirements.md" }]
},
"budgets": { "maxToolCalls": 15, "maxDurationMs": 180000 },
"graders": [
{ "type": "trace.mustCallMcpServer", "server": "computer-use" },
{ "type": "trace.mustCallTool", "name": "read_file" },
{ "type": "trace.mustNotLoop", "maxRepeatedCall": 3 },
{
"type": "trace.budget",
"maxToolCalls": 15,
"maxToolResultTokensEst": 30000,
"maxSingleToolResultTokensEst": 15000
},
{
"type": "trace.finalTextMatches",
"anyOf": ["webhook"],
"allOf": ["webhook", "slack|sales-leads"]
}
],
"tags": ["filesystem-read", "regression"]
}