n8n/packages/@n8n/instance-ai/evaluations/computer-use/data/M.4-file-targets-local.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

24 lines
689 B
JSON

{
"id": "M.4-file-targets-local",
"category": "filesystem-write",
"prompt": "Create a file called test.txt with the text 'Hello World' on my computer.",
"budgets": { "maxToolCalls": 10, "maxDurationMs": 120000 },
"graders": [
{ "type": "trace.mustCallMcpServer", "server": "computer-use" },
{ "type": "trace.mustCallTool", "name": "write_file" },
{ "type": "trace.mustNotLoop", "maxRepeatedCall": 3 },
{
"type": "trace.budget",
"maxToolCalls": 10,
"maxToolResultTokensEst": 10000,
"maxSingleToolResultTokensEst": 5000
},
{
"type": "fs.fileMatches",
"glob": "**/test.txt",
"anyOf": ["Hello World"]
}
],
"tags": ["filesystem-write", "regression"]
}