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

# Knowledge Base Overview

> Build the API context layer Woes agents use to answer developer questions.

# Knowledge Base Overview

The Woes knowledge base is the source of truth for grounded AI support. It is
not just a collection of articles; it is an API context layer built from docs,
specs, repositories, examples, auth rules, schemas, and endpoint-level
Markdown.

## What You Can Add

| Source                        | Use it for                                                                                                       |
| ----------------------------- | ---------------------------------------------------------------------------------------------------------------- |
| OpenAPI or Swagger            | REST endpoint paths, methods, schemas, auth, and examples.                                                       |
| Postman                       | Collection folders, requests, examples, and auth hints.                                                          |
| Documentation sites           | Human-written guides, concepts, setup flows, and troubleshooting.                                                |
| Markdown or pasted text       | Runbooks, known issues, policy notes, and launch docs.                                                           |
| GraphQL, AsyncAPI, SOAP, gRPC | Non-REST API surfaces and protocol-specific docs.                                                                |
| GitHub repositories           | Default-branch READMEs, docs folders, specs, SDK examples, and code-adjacent support context.                    |
| Notion pages                  | Product notes, FAQs, setup guides, release notes, and support playbooks shared with the Woes Notion integration. |
| Confluence spaces             | Support runbooks, implementation notes, release guidance, and troubleshooting pages imported from Confluence.    |
| HubSpot CRM records           | Operator-only account, contact, deal, and owner context for support triage.                                      |

## Source Lifecycle

<Steps>
  <Step title="Add a source">
    Use [Add API Context](/quickstart/add-context) or [Sources](/context/sources) to submit a URL, spec, repository, imported Notion page, imported Confluence space, or pasted content.
  </Step>

  <Step title="Watch ingestion">
    Woes scans the source, extracts usable content, builds Markdown, chunks it for retrieval, and publishes searchable context.
  </Step>

  <Step title="Attach it to agents">
    Use [Agents](/agents/overview) to decide which agents can answer from the source.
  </Step>

  <Step title="Test answers">
    Ask realistic customer questions and verify the agent uses the right evidence.
  </Step>

  <Step title="Maintain coverage">
    Rescan sources when docs change and review [Knowledge Base Gaps](/knowledge-base/gaps).
  </Step>
</Steps>

## What Makes Good Context

* Specific endpoint descriptions.
* Authentication rules and examples.
* Request and response examples.
* Error codes and troubleshooting details.
* SDK examples with language names.
* Product concepts customers commonly ask about.
* Versioning and deprecation notes.

## Safety Rules

Do not use the knowledge base to store raw customer secrets, private tokens, or
operator-only credentials. Source auth credentials used for testing are stored
separately from source content and should never be copied into public docs.

URL ingestion and website crawling are bounded by safe-fetch, same-origin, size,
time, and request-count controls. Public docs publication is sanitized and
available only for ready source material.

HubSpot CRM records are different from normal knowledge-base sources: they are
operator-only account context and are not used as customer-facing widget answer
evidence by default.