Significant-Gravitas/AutoGPT · error · HTTPException
chat_transport_not_configured
Error message
chat_transport_not_configured
What it means
A 503 from _resolve_new_session_llm_route's builder branch: the request is a valid builder session (builder_graph_id set, no codex/credential override), but _is_deployment_chat_available() reports the deployment chat transport is not configured on this backend. The backend knows it cannot serve the session, so it fails fast with service-unavailable instead of creating a broken session.
Source
Thrown at autogpt_platform/backend/backend/api/features/chat/routes.py:575
async def _resolve_new_session_llm_route(
user_id: str,
request: CreateSessionRequest | None,
) -> tuple[CopilotLlmAuthProvider, str | None]:
auth_provider = request.llm_auth_provider if request else "platform"
credential_id = request.llm_credential_id if request else None
if auth_provider == "codex":
await enforce_codex_access_http(user_id)
if request is not None and request.builder_graph_id is not None:
if auth_provider == "codex" or credential_id is not None:
raise HTTPException(
status_code=422,
detail="codex_builder_session_unsupported",
)
if not _is_deployment_chat_available():
raise HTTPException(
status_code=503,
detail="chat_transport_not_configured",
)
return "platform", None
transports = await _get_chat_transports(user_id)
if request is not None:
route_was_explicit = bool(
{"llm_auth_provider", "llm_credential_id"} & request.model_fields_set
)
if route_was_explicit:
if auth_provider == "platform" and credential_id is not None:
raise HTTPException(
status_code=422,
detail="codex_credential_not_allowed",
)
if auth_provider == "codex" and credential_id is None:
raise HTTPException(View on GitHub (pinned to 9c8bb5550f)
Solutions
- Enable/configure deployment chat on the backend: set the required transport/env config (deployment chat / agent-server endpoint) and restart.
- Verify you are pointing at an environment that supports builder chat (prod/staging vs a stripped-down local stack).
- If the feature is intentionally off in this environment, gate the builder-chat entry point in the client so the request is never made.
Defensive patterns
Strategy: try-catch
Validate before calling
const caps = await getBackendCapabilities();
if (!caps.deployment_chat) showUnavailable(); else createSession({builder_graph_id}); Try / catch
try {
await createSession({builder_graph_id: g});
} catch (e) {
if (e.status === 503) showFeatureUnavailable('Builder chat is not configured here');
else throw e;
} Prevention
- Gate builder-chat UI on a capability/feature flag from the backend.
- Keep deployment-chat env config present in every environment that serves the builder UI.
When it happens
Trigger: POST /chat/sessions with builder_graph_id while the backend lacks the deployment-chat configuration (missing agent-server base URL / transport settings for deployment chat in that environment).
Common situations: Local dev backend started without the deployment/agent-server env vars; a staging environment where the builder chat feature is not enabled; version skew where the frontend deploys builder-chat UI ahead of backend configuration.
Related errors
- Payment redirect URLs cannot be validated: frontend_base_url
- Session {session_id} not found.
- Title must not be blank
- codex_builder_session_unsupported
- codex_credential_not_allowed
AI-assisted analysis of Significant-Gravitas/AutoGPT@9c8bb5550f (2026-08-14).
Data as JSON: /api/errors/a12175cfa21cb01d.
Report an issue: GitHub.