Cognotik | AbductiveReasoningTask

⚙️ ExecutionConfig.json


{
  "observations": [
    "Memory usage spikes every 4 hours",
    "No cron jobs scheduled at those times",
    "Spikes correlate with high disk I/O"
  ],
  "domain_context": "JVM Microservice",
  "max_hypotheses": 3,
  "suggest_tests": true
}

→

👁️ Session UI (TabbedDisplay)

Overview

Hypotheses

Best Explanation

Hypothesis: Periodic Log Rotation Buffer Flush

Overall Score: 0.88

The application uses an internal async appender that flushes large buffers based on a 4-hour policy...

✔ Explanatory Power: 0.95
✔ Simplicity: 0.80

Suggested Validation:

Check logback.xml for 'flushInterval'
Monitor FileDescriptor count during spikes

Test Workspace Browser

Explore actual artifacts, reports, and logs generated by the AbductiveReasoningTask in the test environment.

Configuration Parameters

Field	Type	Default	Description
`observations` *	List<String>	-	The set of facts, anomalies, or data points that require an explanation.
`generate_hypotheses`	Boolean	true	Whether to generate new hypotheses or only evaluate provided ones.
`max_hypotheses`	Int	5	Maximum number of candidate explanations to generate (1-10).
`evaluate_criteria`	List<String>	[power, simplicity, testability, probability]	Metrics used to score and rank hypotheses.
`input_files`	List<String>	null	Glob patterns (e.g. `*/.kt`) for source code context.
`suggest_tests`	Boolean	true	Whether to suggest concrete validation/falsification tests.
`domain_context`	String	null	Constraints or background knowledge (e.g., "JVM Microservice").
`existing_hypotheses`	List<String>	null	Pre-existing hypotheses to be evaluated against observations.

Embedded Execution

To run this task programmatically using the UnifiedHarness (Headless mode):


import com.simiacryptus.cognotik.apps.general.UnifiedHarness
import com.simiacryptus.cognotik.plan.tools.reasoning.AbductiveReasoningTask
import com.simiacryptus.cognotik.plan.tools.reasoning.AbductiveReasoningTask.AbductiveReasoningTaskExecutionConfigData
import java.io.File

val harness = UnifiedHarness(serverless = true, openBrowser = false).apply { start() }

val executionConfig = AbductiveReasoningTaskExecutionConfigData(
    observations = listOf(
        "API latency increased from 50ms to 450ms",
        "Database CPU is at 95%",
        "No new deployments in the last 24 hours"
    ),
    domain_context = "E-commerce Backend (Spring Boot/PostgreSQL)",
    input_files = listOf("src/main/resources/application.yml", "src/main/kotlin/db/**"),
    suggest_tests = true
)

harness.runTask(
    taskType = AbductiveReasoningTask.AbductiveReasoning,
    executionConfig = executionConfig,
    workspace = File("./project-dir"),
    autoFix = true
)

Direct Instantiation

If you are manually constructing a plan or using the OrchestrationConfig directly:


val task = AbductiveReasoningTask(
    orchestrationConfig = config,
    planTask = AbductiveReasoningTaskExecutionConfigData(
        observations = listOf("System latency increased by 200ms"),
        domain_context = "PostgreSQL Database",
        suggest_tests = true
    )
)

Prompt Segment

The following logic is injected into the agent's context:

AbductiveReasoning - Generate and evaluate explanatory hypotheses
  ** Specify observations that need explanation
  ** Configure hypothesis generation (max_hypotheses: ${executionConfig?.max_hypotheses ?: 5})
  ** Select evaluation criteria: ${executionConfig?.evaluate_criteria?.joinToString(", ") ?: "all"}
  ** Optionally provide existing hypotheses to evaluate
  ** Optionally suggest tests to validate hypotheses
  ** Useful for: Root cause analysis, Bug investigation