Pattern Tools
Tools for managing the automated pattern detection engine. The system scans the research graph on a schedule to detect convergence, contradictions, staleness, cross-domain connections, and untested implications.
Available Tools
pattern_list
List detected patterns for a workspace.
{
"name": "pattern_list",
"arguments": {
"workspaceId": "ws_123",
"status": "detected",
"patternType": "contradiction"
}
}
Arguments:
| Argument | Type | Required | Description |
|---|---|---|---|
workspaceId | string | Yes | Workspace ID |
spaceId | string | No | Filter by space |
status | string | No | Filter by status (detected, acknowledged, dismissed) |
patternType | string | No | Filter by pattern type |
Response:
{
"patterns": [
{
"id": "pat_001",
"patternType": "contradiction",
"severity": "high",
"title": "Contradicting hypotheses: Caching vs No-cache",
"details": {
"fromPage": "page_hyp_123",
"fromTitle": "Caching reduces latency by 50%",
"toPage": "page_hyp_456",
"toTitle": "Caching adds overhead for small payloads"
},
"status": "detected",
"detectedAt": "2024-01-15T06:00:00Z"
}
]
}
pattern_acknowledge
Acknowledge a detected pattern, indicating it has been reviewed.
{
"name": "pattern_acknowledge",
"arguments": {
"patternId": "pat_001",
"actionTaken": {
"notes": "Created a new experiment to resolve the contradiction",
"experimentId": "page_exp_789"
}
}
}
Arguments:
| Argument | Type | Required | Description |
|---|---|---|---|
patternId | string | Yes | Pattern ID |
actionTaken | object | No | Details about the action taken |
pattern_dismiss
Dismiss a detected pattern.
{
"name": "pattern_dismiss",
"arguments": {
"patternId": "pat_001"
}
}
pattern_run
Manually trigger pattern detection for a workspace. Runs all configured pattern evaluators immediately.
{
"name": "pattern_run",
"arguments": {
"workspaceId": "ws_123"
}
}
Response:
{
"patternsDetected": 3,
"message": "Pattern detection completed"
}
Pattern detection also runs automatically every 6 hours for workspaces with intelligence enabled.
Pattern Types
| Type | Severity | Description |
|---|---|---|
convergence | medium | Multiple experiments validate the same hypothesis (default threshold: 3+) |
contradiction | high | Two pages have a CONTRADICTS relationship in the graph |
staleness | low | Open questions that haven't been updated in a configurable period |
cross_domain | medium | Connections detected between pages with different domain tags |
untested_implication | medium | A validated hypothesis EXTENDS a proposed hypothesis with no testing |
intake_gate | high | Hypothesis marked as "proved" without completing its intake gate checklist |
evidence_gap | medium | Hypothesis with citations or formalizations but fewer than N experiments |
reproduction_failure | high | Experiment with incoming FAILS_TO_REPRODUCE edges from other experiments |
Pattern Statuses
| Status | Description |
|---|---|
detected | Pattern was found by an evaluator |
acknowledged | A user has reviewed and acknowledged the pattern |
dismissed | Pattern was dismissed as not relevant |
How Detection Works
Pattern detection runs on a schedule (every 6 hours) or can be triggered manually:
Evaluators
Convergence — Finds hypotheses with 3+ validating experiments, suggesting strong evidence.
Contradiction — Scans the research graph for CONTRADICTS edges between pages.
Staleness — Finds open questions (tasks labeled "open-question") not updated within a configurable period.
Cross-domain — Detects connections between pages with different domain tags, revealing unexpected relationships.
Untested implications — Finds validated hypotheses that EXTEND proposed hypotheses with no experiments testing them.
Intake gate violation — Flags hypotheses whose claim label is "proved" but whose intake gate checklist hasn't been completed.
Evidence gap — Identifies hypotheses that are cited or formalized by other pages but have fewer than the configured number of experiments testing them.
Reproduction failure — Detects experiments that have incoming FAILS_TO_REPRODUCE edges, indicating another experiment could not reproduce the results.
Actions
When patterns are detected, configurable actions can be taken:
| Action | Description |
|---|---|
notify | Emit an event for notification systems |
flag | Record the pattern (default behavior) |
surface | Prominently surface the pattern in the dashboard |
create_task | Automatically create a task to address the pattern |
Configuration
Pattern detection is configured per-workspace via intelligence settings:
{
"intelligence": {
"enabled": true,
"patternRules": [
{
"type": "convergence",
"enabled": true,
"threshold": 3,
"action": "surface"
},
{
"type": "contradiction",
"enabled": true,
"action": "notify"
},
{
"type": "staleness",
"enabled": true,
"maxAgeDays": 14,
"action": "create_task"
}
]
}
}
Related
- Hypothesis Tools - Hypotheses that patterns analyze
- Relationship Tools - Graph edges patterns detect
- Intelligence Tools - Query context including patterns
- Research Intelligence Guide - Full system overview