{
  "fix_id": "fix-abc123",
  "type": "retry",
  "target": {
    "error_codes": ["TOOL.EXECUTION.TIMEOUT"],
    "agents": ["researcher"],
    "tools": ["search_api"]
  },
  "config": {
    // Type-specific configuration
  },
  "rollback_strategy": {
    "type": "immediate",
    "trigger": "error_rate > baseline * 1.1"
  },
  "metadata": {
    "created_at": "2024-01-15T10:00:00Z",
    "diagnosis_id": "diag-xyz789",
    "confidence": 0.85
  }
}

Targeting

Fixes can target specific:

Target	Description
`error_codes`	Specific error codes
`agents`	Named agents
`tools`	Specific tools
`models`	LLM models
`environments`	dev/staging/prod

Rollback Strategies

Strategy	Description
`immediate`	Instant rollback on trigger
`gradual`	Ramp down over time
`manual`	Require human approval

Fix Precedence

When multiple fixes match:

Most specific target wins
Higher confidence wins
Most recent wins

SDK Integration

The SDK automatically applies fixes:

import risicare
 
risicare.init()
 
# Fixes loaded from API and cached locally
# Applied automatically to matching operations
 
response = client.chat.completions.create(...)

Managing Fixes

Via Dashboard

Navigate to Healing → Fixes
View active, testing, graduated fixes
Edit, pause, or rollback fixes

Via API

# List fixes
curl -X GET "https://app.risicare.ai/v1/fixes" \
  -H "Authorization: Bearer rsk-..."
 
# Create fix
curl -X POST "https://app.risicare.ai/v1/fixes" \
  -H "Authorization: Bearer rsk-..." \
  -d '{"type": "retry", "config": {...}}'
 
# Rollback fix
curl -X DELETE "https://app.risicare.ai/v1/fixes/{fix_id}" \
  -H "Authorization: Bearer rsk-..."

Next Steps

Select a fix type to learn more:

Prompt Fixes

Modify prompts and examples

Learn more

Retry Fixes

Add retry with backoff

Learn more

PreviousHypothesis Testing NextPrompt