Loop

Overview

The Loop block signals the pipeline to jump back to a previous block, allowing you to repeat portions of your test. Useful for retry logic, polling, or testing iterative agent behaviors.

{
  "block": "Loop",
  "config": {
    "target": "checkStatus",
    "maxIterations": 5
  }
}

Configuration

Parameter	Type	Default	Description
`target`	string/number	required	Block ID or index to loop back to
`maxIterations`	number	`10`	Maximum iterations to prevent infinite loops
`condition`	object	optional	Condition object to evaluate. Loop continues while condition is true. If omitted, loops unconditionally.

Input Parameters

Loop accepts any inputs and passes them through. This allows you to carry data through loop iterations.

Output Fields

Loop outputs special control fields:

Field	Type	Description
`_loopTo`	string/number	Target block to loop back to (only when looping)
`_maxLoops`	number	Maximum iterations allowed (only when looping)
`_shouldLoop`	boolean	Whether the loop condition was met and looping should occur

Plus all input fields are passed through.

How It Works

Pipeline executes blocks in sequence
When Loop executes, it evaluates the condition (if provided)
If condition is true (or no condition), returns _loopTo, _maxLoops, and _shouldLoop: true
Pipeline jumps back to the target block
Blocks between target and Loop execute again
Loop counter increments each iteration
When condition becomes false or maxIterations reached, loop stops

Example flow:

Block 1 → Block 2 → Block 3 (Loop target="Block 2") → back to Block 2
         ↑                                           │
         └───────────────────────────────────────────┘

Conditional Logic

Loop supports conditional evaluation to control when looping occurs. The condition parameter accepts an object with:

Field	Type	Description
`path`	string	Path to value in DataBus (e.g., `"response.status"`)
`operator`	string	Comparison operator (see below)
`value`	any	Expected value to compare against

Available Operators

Equality: equals, notEquals
Numeric: gt, gte, lt, lte
String/Array: contains, notContains, minLength, maxLength
Pattern: matches, notMatches
Type: isNull, isNotNull, isDefined, isUndefined
Boolean: isTrue, isFalse
Empty: isEmpty, isNotEmpty

Condition Examples

{
  "condition": {
    "path": "response.status",
    "operator": "notEquals",
    "value": 200
  }
}

Loops while response.status is not 200.

{
  "condition": {
    "path": "job.progress",
    "operator": "lt",
    "value": 100
  }
}

Loops while job.progress is less than 100.

{
  "condition": {
    "path": "data.ready",
    "operator": "isFalse"
  }
}

Loops while data.ready is false.

Examples

Basic Loop

{
  "pipeline": [
    {
      "id": "attempt",
      "block": "HttpRequest",
      "input": {
        "url": "${BASE_URL}/status",
        "method": "GET"
      },
      "output": "response"
    },
    {
      "id": "check",
      "block": "Loop",
      "config": {
        "target": "attempt",
        "maxIterations": 5
      }
    }
  ]
}

This loops back to “attempt” up to 5 times.

Conditional Loop (Polling)

{
  "pipeline": [
    {
      "id": "checkStatus",
      "block": "HttpRequest",
      "input": {
        "url": "${BASE_URL}/job/${jobId}/status",
        "method": "GET"
      },
      "output": "statusResponse"
    },
    {
      "id": "parse",
      "block": "JsonParser",
      "input": "${statusResponse.body}",
      "output": {
        "parsed": "status"
      }
    },
    {
      "id": "conditionalLoop",
      "block": "Loop",
      "config": {
        "target": "checkStatus",
        "maxIterations": 10,
        "condition": {
          "path": "status.state",
          "operator": "notEquals",
          "value": "completed"
        }
      }
    }
  ],
  "assertions": {
    "status.state": "completed"
  }
}

How it works: The loop checks status.state each iteration. If it’s not “completed”, it loops back to checkStatus. When the status becomes “completed” (or maxIterations is reached), the loop stops and assertions verify the final state.

Retry Until Success

{
  "pipeline": [
    {
      "id": "attempt",
      "block": "HttpRequest",
      "input": {
        "url": "${API_URL}/flaky-endpoint",
        "method": "GET"
      },
      "output": "response"
    },
    {
      "id": "retry",
      "block": "Loop",
      "config": {
        "target": "attempt",
        "maxIterations": 3,
        "condition": {
          "path": "response.status",
          "operator": "notEquals",
          "value": 200
        }
      }
    }
  ],
  "assertions": {
    "response.status": 200
  }
}

This retries the request up to 3 times, stopping as soon as it gets a 200 status.

Threshold Checking

{
  "pipeline": [
    {
      "id": "measure",
      "block": "HttpRequest",
      "input": {
        "url": "${API_URL}/metrics",
        "method": "GET"
      },
      "output": "metricsResponse"
    },
    {
      "id": "parse",
      "block": "JsonParser",
      "input": "${metricsResponse.body}",
      "output": {
        "parsed": "metrics"
      }
    },
    {
      "id": "checkThreshold",
      "block": "Loop",
      "config": {
        "target": "measure",
        "maxIterations": 5,
        "condition": {
          "path": "metrics.score",
          "operator": "lt",
          "value": 0.8
        }
      }
    }
  ],
  "assertions": {
    "metrics.score": { "gte": 0.8 }
  }
}

Keeps checking until the score reaches at least 0.8, or gives up after 5 attempts.

Multi-Step Loop

{
  "pipeline": [
    {
      "id": "fetch",
      "block": "HttpRequest",
      "input": {
        "url": "${API_URL}/data?page=${pageNum}",
        "method": "GET"
      },
      "output": "pageResponse"
    },
    {
      "id": "parse",
      "block": "JsonParser",
      "input": "${pageResponse.body}",
      "output": {
        "parsed": "page"
      }
    },
    {
      "id": "process",
      "block": "ValidateContent",
      "input": {
        "from": "page.data",
        "as": "text"
      },
      "config": {
        "contains": "expected"
      },
      "output": "validation"
    },
    {
      "id": "loop",
      "block": "Loop",
      "config": {
        "target": "fetch",
        "maxIterations": 5
      }
    }
  ]
}

Loop Counter

The pipeline tracks loop iterations in context:

_loopCount:targetBlockId = 0, 1, 2, ...

This counter is per-target, so multiple loops can coexist.

Common Patterns

Polling Until Ready

{
  "name": "Job Polling Test",
  "context": {
    "API_URL": "${env.API_URL}",
    "JOB_ID": "job-123"
  },
  "tests": [{
    "id": "poll-until-complete",
    "pipeline": [
      {
        "id": "checkJobStatus",
        "block": "HttpRequest",
        "input": {
          "url": "${API_URL}/jobs/${JOB_ID}",
          "method": "GET"
        },
        "output": "jobResponse"
      },
      {
        "id": "parseStatus",
        "block": "JsonParser",
        "input": "${jobResponse.body}",
        "output": {
          "parsed": "job"
        }
      },
      {
        "id": "pollLoop",
        "block": "Loop",
        "config": {
          "target": "checkJobStatus",
          "maxIterations": 20
        }
      }
    ],
    "assertions": {
      "job.status": "completed"
    }
  }]
}

How it works:

Checks job status
Parses response
Loops back if not complete
Stops after 20 attempts or when complete

Retry Failed Requests

{
  "pipeline": [
    {
      "id": "sendRequest",
      "block": "HttpRequest",
      "input": {
        "url": "${API_URL}/unstable",
        "method": "POST",
        "body": { "data": "test" }
      },
      "output": "response"
    },
    {
      "id": "retryLoop",
      "block": "Loop",
      "config": {
        "target": "sendRequest",
        "maxIterations": 3
      }
    }
  ],
  "assertions": {
    "response.status": { "lt": 400 }
  }
}

Agent Iteration Testing

{
  "pipeline": [
    {
      "id": "callAgent",
      "block": "HttpRequest",
      "input": {
        "url": "${AI_URL}/agent",
        "method": "POST",
        "body": {
          "task": "Find and process data"
        }
      },
      "output": "agentResponse"
    },
    {
      "id": "parseResponse",
      "block": "StreamParser",
      "input": "${agentResponse.body}",
      "config": {
        "format": "sse-openai"
      },
      "output": {
        "text": "agentMessage",
        "toolCalls": "tools"
      }
    },
    {
      "id": "validateTools",
      "block": "ValidateTools",
      "input": {
        "from": "tools",
        "as": "toolCalls"
      },
      "config": {
        "expected": ["search", "process"]
      },
      "output": "toolValidation"
    },
    {
      "id": "iterateLoop",
      "block": "Loop",
      "config": {
        "target": "callAgent",
        "maxIterations": 5
      }
    }
  ],
  "assertions": {
    "toolValidation.passed": true
  }
}

Loop Limits

Default limit: 10 iterations per loop target Maximum safe limit: 100 iterations Why limits exist:

Prevent infinite loops
Avoid test timeouts
Protect against configuration errors

Override the limit:

{
  "config": {
    "target": "myBlock",
    "maxIterations": 50
  }
}

Best Practices

Always Set Reasonable maxIterations

Don’t rely on defaults. Set appropriate limits:

{
  "config": {
    "target": "checkStatus",
    "maxIterations": 30  // 30 seconds if polling every second
  }
}

Use Loop for Polling

Perfect for checking async job status:

{
  "pipeline": [
    { "id": "check", "block": "HttpRequest", ... },
    { "id": "parse", "block": "JsonParser", ... },
    {
      "block": "Loop",
      "config": { "target": "check", "maxIterations": 20 }
    }
  ]
}

Combine with Assertions

Use assertions to verify final state after looping:

{
  "pipeline": [
    { "id": "poll", "block": "HttpRequest", ... },
    { "block": "Loop", "config": { "target": "poll" } }
  ],
  "assertions": {
    "response.data.status": "ready"
  }
}

Don't Nest Loops

Avoid loops within loops - use separate tests instead:

// Bad - nested loops (confusing)
{
  "pipeline": [
    { "id": "outer", ... },
    { "id": "inner", ... },
    { "block": "Loop", "config": { "target": "inner" } },
    { "block": "Loop", "config": { "target": "outer" } }
  ]
}

// Good - separate tests
{
  "tests": [
    { "id": "test1", "pipeline": [ ... ] },
    { "id": "test2", "pipeline": [ ... ] }
  ]
}

Limitations

Current limitations:

Single target - Can only loop to one block at a time
Forward loops not allowed - Can only loop backwards to previous blocks
No loop nesting - Cannot reliably nest loops within loops

Best practices:

Use conditional logic to control loop execution
Set reasonable maxIterations to prevent timeouts
Use assertions to verify final state after looping
For complex scenarios, split into separate tests

Full Example

{
  "name": "Async Job Polling Test",
  "context": {
    "API_URL": "${env.API_URL}",
    "API_KEY": "${env.API_KEY}"
  },
  "tests": [{
    "id": "test-job-completion",
    "pipeline": [
      {
        "id": "startJob",
        "block": "HttpRequest",
        "input": {
          "url": "${API_URL}/jobs",
          "method": "POST",
          "headers": {
            "Authorization": "Bearer ${API_KEY}"
          },
          "body": {
            "task": "process_data"
          }
        },
        "output": "startResponse"
      },
      {
        "id": "parseJobId",
        "block": "JsonParser",
        "input": "${startResponse.body}",
        "output": {
          "parsed": "jobInfo"
        }
      },
      {
        "id": "pollStatus",
        "block": "HttpRequest",
        "input": {
          "url": "${API_URL}/jobs/${jobInfo.id}",
          "method": "GET",
          "headers": {
            "Authorization": "Bearer ${API_KEY}"
          }
        },
        "output": "statusResponse"
      },
      {
        "id": "parseStatus",
        "block": "JsonParser",
        "input": "${statusResponse.body}",
        "output": {
          "parsed": "status"
        }
      },
      {
        "id": "waitLoop",
        "block": "Loop",
        "config": {
          "target": "pollStatus",
          "maxIterations": 30
        }
      }
    ],
    "assertions": {
      "startResponse.status": 201,
      "jobInfo.id": { "matches": ".+" },
      "status.state": "completed",
      "status.result": { "matches": ".+" }
    }
  }]
}

Tips

Add Delays Between Iterations

Currently, Loop doesn’t have built-in delays. For polling, consider test structure or external delays.

Monitor Loop Counter

Check context for loop count if needed:

context._loopCount:targetBlock = 5

Use for Flaky Tests

Retry flaky operations automatically:

{
  "pipeline": [
    { "id": "flakyOp", "block": "HttpRequest", ... },
    {
      "block": "Loop",
      "config": { "target": "flakyOp", "maxIterations": 3 }
    }
  ]
}

When to Use

Use Loop when:

Polling for async operation completion
Retrying failed requests
Testing iterative agent behaviors
Checking status until ready

Don’t use when:

Simple sequential flow is sufficient
Need complex conditional logic (use separate tests)
Testing requires precise timing control

Next Steps

Pipelines

Learn about pipeline execution

Data Flow

Understand data passing between blocks

Get Started

Core Concepts

Blocks

Testing AI Systems

Advanced

Overview

Configuration

Input Parameters

Output Fields

How It Works

Conditional Logic

Available Operators

Condition Examples

Examples

Basic Loop

Conditional Loop (Polling)

Retry Until Success

Threshold Checking

Multi-Step Loop

Loop Counter

Common Patterns

Polling Until Ready

Retry Failed Requests

Agent Iteration Testing

Loop Limits

Best Practices

Limitations

Full Example

Tips

When to Use

Next Steps

Pipelines

Data Flow

Get Started

Core Concepts

Blocks

Testing AI Systems

Advanced

​Overview

​Configuration

​Input Parameters

​Output Fields

​How It Works

​Conditional Logic

​Available Operators

​Condition Examples

​Examples

​Basic Loop

​Conditional Loop (Polling)

​Retry Until Success

​Threshold Checking

​Multi-Step Loop

​Loop Counter

​Common Patterns

​Polling Until Ready

​Retry Failed Requests

​Agent Iteration Testing

​Loop Limits

​Best Practices

​Limitations

​Full Example

​Tips

​When to Use

​Next Steps

Pipelines

Data Flow

Overview

Configuration

Input Parameters

Output Fields

How It Works

Conditional Logic

Available Operators

Condition Examples

Examples

Basic Loop

Conditional Loop (Polling)

Retry Until Success

Threshold Checking

Multi-Step Loop

Loop Counter

Common Patterns

Polling Until Ready

Retry Failed Requests

Agent Iteration Testing

Loop Limits

Best Practices

Limitations

Full Example

Tips

When to Use

Next Steps