Sources
Sources are the material Woes uses to answer customer questions. They can be API specifications, docs pages, repositories, schemas, examples, or support knowledge.Source Lifecycle
1
Create
Add a URL, file, raw text, schema, GitHub repository, imported Notion page,
imported Confluence space, or synced HubSpot CRM context in Context.
2
Scan
Woes fetches, parses, normalizes, and chunks the source.
3
Inspect
Review endpoint groups, documents, generated endpoint Markdown, examples, auth, and health.
4
Attach
Attach the source to one or more agents.
5
Rescan
Rescan when source content changes. Woes avoids unnecessary re-embedding when content has not changed.
Ingestion Stages
Woes shows progress while a source is scanned. Website sources use a crawl-first flow, while structured sources use an API-shape detection flow.Detail View
The source detail view can show:- Endpoint method and path.
- Summary and description.
- Parameters.
- Request bodies.
- Response codes and schemas.
- Auth requirements.
- Examples.
- Generated Markdown.
- Documents and chunks.
- Source health and status.
Retrieval Quality
High-quality sources include:- Exact endpoint paths.
- Auth requirements.
- Required and optional fields.
- Request and response examples.
- Error examples.
- Webhook and event behavior.
- SDK snippets.
- Versioning notes.
Source Types
Woes can auto-detect or ingest a specific source type:- Website.
- OpenAPI.
- GraphQL.
- AsyncAPI.
- Postman.
- gRPC.
- WSDL / SOAP.
- URL.
- SDK docs.
- Raw Markdown or text.
- GitHub repository docs and examples.
- Notion pages shared with the Woes integration.
- Confluence spaces imported from the Woes integration.
- HubSpot CRM records synced as operator-only account context.
Safe Fetch And Crawl Boundaries
Woes source ingestion is intentionally bounded. URL and website sources are fetched through safe-fetch controls that block private, loopback, reserved, and unsafe hosts or ports, re-check redirects, and enforce request, byte, and time budgets. Website crawling stays within the allowed origin/path scope and uses bounded depth and page counts; it is not an unbounded crawl of every page on a site.Rescan And Source Limits
URL-backed and connected-context sources can be updated when the upstream material changes. Manual or pasted-content sources should be recreated when their content changes because there is no upstream URL or connector to fetch again. Workspaces have plan and add-on source capacity. When a new source would exceed the effective source limit, Woes blocks creation instead of silently dropping or partially importing the source.Public Docs Publication
Ready sources can be published as sanitized public docs under/docs/{slug}.
Publishing exposes customer-safe generated documentation, endpoint Markdown,
and source summaries; it does not expose source auth credentials, raw private
source metadata, operator debug data, or workspace secrets. Non-ready sources
must be fixed before they are published.