> ## Documentation Index
> Fetch the complete documentation index at: https://docs.argentos.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# SIS Reliability

> How SIS consolidation works, fails, and should be monitored.

## What SIS Does

<Tooltip tip="Self-Improvement System — consolidates contemplation episodes into lessons that influence future agent behavior.">SIS</Tooltip> consolidates contemplation/session episodes into lessons that influence future prompts.

```mermaid theme={null}
flowchart LR
  A["Episodes"] --> B["SIS consolidation"]
  B --> C["Lessons"]
  C --> D["Prompt injection"]
  D --> E["Improved future responses"]
```

## Failure Modes

<AccordionGroup>
  <Accordion title="Parse failure">
    **Symptoms:**

    * `sis: no parseable JSON in response`
    * `consolidation returned no patterns`

    **Impact:** Lessons stop updating.

    **Operator response:**

    1. Verify selected provider/model for SIS path.
    2. Verify latest output is current issue (not stale historical log).
    3. Validate JSON contract and fallback parser behavior.
  </Accordion>

  <Accordion title="Empty consolidation with active episodes">
    **Symptoms:** Many episodes but no new lessons.

    **Impact:** Weak adaptation over time.

    **Operator response:**

    1. Check thresholds (`episodesPerConsolidation`, lesson confidence).
    2. Verify contemplation episodes contain structured content.
  </Accordion>
</AccordionGroup>

## Reliability Targets

| Metric                      | Target              |
| --------------------------- | ------------------- |
| Parse success rate          | >= 95%              |
| Consolidation cycle success | >= 99%              |
| Fresh lesson update         | At expected cadence |

## Required Observability

Expose these in health UI:

1. Last successful consolidation time
2. Parse failures in last 24h
3. Lessons created in last 24h
4. Last lesson injected time