Attribution

Attribution maps inbound request headers to client, end-user, and session identifiers, which are persisted with each request log. Request detail surfaces this metadata, the request list can filter by it, and session IDs deep-link back into the filtered log.

Configure the header mapping on the Attribution page (or via /api/settings/attribution).

Fields

llama-dash stores three attribution fields per request:

• clientName - the application or integration making the request.
• endUserId - the user inside that application.
• sessionId - a conversation, workflow, trace, or other session identifier.

The Requests page can filter by all three. Request detail links a session ID back to the filtered request list.

Header mapping

Configure which inbound headers populate each field. Header names are normalized before lookup, and blank header values are ignored.

If no explicit client header is present, llama-dash can infer clientName from common user agents:

• claude-code
• opencode
• open-webui
• home-assistant
• curl
• python-requests

Client examples

Use stable identifiers that help you answer operational questions later:

curl http://localhost:3000/v1/chat/completions \
  -H 'Authorization: Bearer sk-...' \
  -H 'Content-Type: application/json' \
  -H 'x-client-name: open-webui' \
  -H 'x-end-user-id: alice' \
  -H 'x-session-id: chat_123' \
  -d '{"model":"default-chat","messages":[{"role":"user","content":"hi"}]}'

Attribution is metadata only. It does not authenticate the caller; use API keys and routing rules for enforcement.