Skip to main content

Experiment Tools

Tools for registering, tracking, and completing experiments linked to hypotheses.

Available Tools

experiment_register

Register a new experiment, optionally linked to a hypothesis. Creates a typed page with pageType: 'experiment' and a TESTS_HYPOTHESIS edge in the research graph.

{
"name": "experiment_register",
"arguments": {
"workspaceId": "ws_123",
"spaceId": "space_456",
"title": "Redis cache benchmark under load",
"hypothesisId": "page_hyp_123",
"method": "Load test /api/search with 1000 concurrent users, with and without Redis"
}
}

Arguments:

ArgumentTypeRequiredDescription
workspaceIdstringYesWorkspace ID
spaceIdstringYesSpace ID
titlestringYesExperiment title
hypothesisIdstringNoID of the hypothesis being tested
methodstringNoDescription of the experimental method

Response:

{
"page": {
"id": "page_exp_456",
"title": "Redis cache benchmark under load",
"pageType": "experiment",
"metadata": {
"status": "planned",
"hypothesisId": "page_hyp_123",
"method": "Load test /api/search with 1000 concurrent users..."
}
},
"graphEdge": "TESTS_HYPOTHESIS"
}

experiment_complete

Complete an experiment with results. If passedPredictions is true, a VALIDATES edge is created from the experiment to the hypothesis. If false, a CONTRADICTS edge is created.

{
"name": "experiment_complete",
"arguments": {
"workspaceId": "ws_123",
"pageId": "page_exp_456",
"results": {
"p95Before": 800,
"p95After": 350,
"cacheHitRate": 0.87,
"memoryUsage": "384MB"
},
"passedPredictions": true,
"unexpectedObservations": [
"Cold start latency spiked to 1200ms for first 30 seconds"
]
}
}

Arguments:

ArgumentTypeRequiredDescription
workspaceIdstringYesWorkspace ID
pageIdstringYesExperiment page ID
resultsobjectYesExperiment results (freeform)
passedPredictionsbooleanNoWhether predictions were confirmed
unexpectedObservationsstring[]NoAny unexpected observations

Response:

{
"page": {
"id": "page_exp_456",
"pageType": "experiment",
"metadata": {
"status": "completed",
"results": { "p95Before": 800, "p95After": 350 },
"passedPredictions": true,
"completedAt": "2024-01-15T14:30:00Z"
}
},
"graphEdge": "VALIDATES"
}

experiment_update

Update an experiment's status or metadata.

{
"name": "experiment_update",
"arguments": {
"workspaceId": "ws_123",
"pageId": "page_exp_456",
"status": "running",
"metadata": {
"startedAt": "2024-01-15T10:00:00Z"
}
}
}

Arguments:

ArgumentTypeRequiredDescription
workspaceIdstringYesWorkspace ID
pageIdstringYesExperiment page ID
statusstringNoNew status
metadataobjectNoMetadata fields to update

Experiment Statuses

StatusDescription
plannedExperiment is designed but not started
runningExperiment is currently in progress
completedExperiment is done, results recorded
failedExperiment could not be completed

Graph Edges

When experiments are registered and completed, the system automatically manages edges in the research graph:

EventEdge CreatedDirection
Register with hypothesisIdTESTS_HYPOTHESISExperiment → Hypothesis
Complete with passedPredictions: trueVALIDATESExperiment → Hypothesis
Complete with passedPredictions: falseCONTRADICTSExperiment → Hypothesis