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

# Knowledge Bank

> Encode your team's runbooks, heuristics, and tribal knowledge to sharpen Traversal's investigations.

Knowledge Bank lets you teach Traversal the operational context that lives only in your team's heads — which services are business-critical, which alerts are usually noise, and how your team actually debugs an issue. This doesn't replace Traversal's intelligence; it sharpens it with your experience.

## Why Knowledge Bank?

Even the best AI can miss context that isn't in your observability data:

* Which services are business-critical
* Which alerts are usually noise
* Where your team actually starts when debugging an issue

Knowledge Bank closes this gap by letting you encode company-specific operational and tribal knowledge directly into the investigation process.

## What to upload

You can upload any specific pieces of tribal knowledge, instructions, or context about your system. Start with runbooks, then add broader system context.

**Examples of effective knowledge items:**

* **Prioritization rules** — "Staging alerts are usually low priority unless near a release."
* **Debugging workflows** — "For database issues, start with connection pool metrics before logs."
* **External dependencies** — "Payments alerts often correlate with PSP outages — check their status page first."
* **Service context** — "This cluster supports checkout and is customer-facing."

## How Knowledge Bank works

Knowledge Bank provides three channels for capturing knowledge.

<Steps>
  <Step title="Upload runbooks and custom context">
    Upload a PDF, paste text, or write instructions directly in the form. Each knowledge item includes:

    * **Title** (required)
    * **Instruction** (required, free-form text)
    * **Use when** (optional, applicability hint)

    <Tip>
      Start with runbooks. They are the most effective type of knowledge item and give Traversal immediate, actionable context.
    </Tip>
  </Step>

  <Step title="Automated learning">
    Traversal learns implicitly from how your team interacts with it. When you redirect an investigation — for example, switching from logs to traces or metrics — Traversal learns from those choices automatically, without requiring intentional feedback.

    After each investigation concludes:

    * Traversal extracts customer-specific insights
    * Generic or obvious observations are filtered out
    * New memories appear in your Knowledge Bank for review

    You can edit, approve, or delete anything Traversal learns.
  </Step>

  <Step title="In-session feedback">
    After any investigation, you can explicitly tell Traversal what it got right, what it missed, and what should happen differently next time. This feedback is intentional and specific to the investigation that just occurred, giving Traversal clear guidance for future investigations.
  </Step>
</Steps>

## Getting started

Go to **Settings → Knowledge Bank** in the Traversal web app to upload your first knowledge item.

<Note>
  Knowledge Bank is cumulative. Even a few well-written runbooks improve investigation quality immediately, and the system continues to improve as your team uses it.
</Note>