Skip to main content
GET
Error
Required Permissions How permissions work

Authorizations

Authorization
string
header
required

Management API authentication for /api/* endpoints. Use the Authorization header with Bearer <token>, where <token> is one of:

  • a Bifrost management API key,
  • a dashboard session token issued by POST /api/session/login,
  • base64 of <admin-username>:<admin-password> (legacy equivalent of BasicAuth).

Virtual keys (sk-bf-*) and the x-api-key header are not accepted on management APIs - the sole exception is GET /api/governance/virtual-keys/quota, which is virtual-key-only.

Authentication alone is not sufficient in Bifrost Enterprise: each operation page shows a Required Permissions table (Resource:Operation, for example Dashboard:View) above its Authorizations section, and the caller's RBAC role or management API key scopes must include what it lists, otherwise the request is rejected with 403 Forbidden.

A local admin — authenticated with the admin password, or any caller on a deployment with dashboard auth disabled — bypasses these checks and can call every management endpoint.

OSS setup lock. On Bifrost OSS, while dashboard auth is not active (no admin account, or auth disabled), every management endpoint except the public ones (/health, /api/version, /api/session/is-auth-enabled, /api/session/login, ...) requires the operator's setup token in the X-Bifrost-Setup-Token header, in place of Authorization. The token is set with setup_token in config.json or the BIFROST_SETUP_TOKEN environment variable. A missing header returns 401, a wrong token 403. The header stops working once dashboard auth is enabled. The dashboard instead trades the token once for an HttpOnly bifrost_setup_session cookie via POST /api/session/setup. See Required permissions for how permissions are derived and which endpoints are exempt.

Path Parameters

router_id
string
required

Model router id.

Response

Model router status retrieved successfully

Runtime status of the semantic complexity classifier and, when configured, the LLM classifier. Never contains phrases, embeddings, or provider secrets.

state
enum<string>

disabled = no embedding configuration; warming = reference phrases are being embedded; ready = serving the current configuration; failed = the desired configuration failed to warm

Available options:
disabled,
warming,
ready,
failed
loaded
integer

Reference phrases embedded so far in the current warmup

total
integer

Total reference phrases to embed in the current warmup

serving_previous
boolean

When true with state=failed, the previous generation is still serving while the new one failed to warm

failure_reason
enum<string>

Category of a warmup failure when state=failed. Provider and backend error bodies stay in server logs; this bounded vocabulary is what status clients receive.

Available options:
authentication,
model_unavailable,
rate_limited,
timeout,
provider_unavailable,
vector_store_unavailable,
invalid_response,
unknown
error
string

Warmup failure detail when state=failed

cached_phrases
integer

How many phrase vectors the gateway currently holds in process for the configured provider and model. The cache is in-process only, since vectors cannot be read back out of a vector store, so a restart empties it while the saved phrases look unchanged. Zero means the next save re-embeds every phrase regardless of what changed.

storage_mode
enum<string>

Where exemplar vectors are actually kept, which is not always what was configured. semantic.vector_store=vector_store degrades to embedded whenever no top-level vector store is configured, so this reports what is in force rather than what was asked for. Absent until a store has been resolved.

Available options:
embedded,
vector_store
namespace
string

The fingerprinted namespace the serving generation queries, for example BifrostComplexityRouter_. The backend holds no phrase text, so this is the only handle on the records the classifier owns in a shared vector store. Absent while nothing is serving.

capture_available
boolean

Whether there is a vector store to keep captured request vectors in, for phrase suggestions. When false, capture_unavailable_reason says why.

capture_unavailable_reason
enum<string>

Why requests are not captured: no_vector_store = no top-level vector store is configured; embedded_vector_store = the configured store is the embedded one (Chromem), which capture does not use. Absent when capture_available is true.

Available options:
no_vector_store,
embedded_vector_store
llm
object

Present on status reads, where state is disabled when no llm block is configured; left out when the llm status could not be read, and in the warmup retry response.

suggestion_default_instructions
string

The shipped phrase-suggestion instructions, served so a configuration client can seed its editor and offer a reset. The tier definitions and fixed output and safety rules are not exposed.

suggestion_default_tier_instructions
string

The shipped tier-step instructions for suggestions, served so a configuration client can seed its editor and offer a reset. The tier definitions, examples and fixed output and safety rules are not exposed.

suggestion_default_tier_prompt
string

Model router status only: the llm tier prompt this router's suggestion runs send while it is not edited, rendered from the router (its tier definitions and its own phrases as examples).

suggestion_default_tier_criteria
object

Model router status only: the per-tier criteria a decision tier step sends while they are not edited, rendered from the router.

llm_default_prompt
string

The shipped classification guidance, served so a configuration client can seed its prompt editor and offer a reset without holding a copy that drifts from the gateway's. Present on status reads whether or not an llm block is configured; left out with llm, and in the warmup retry response.

decision_defaults
object

The shipped decision-model per-tier criteria, served so a configuration client can seed its editors and offer resets. Filled on status reads; in the warmup retry response criteria is null. The fixed question, decision rule, and context rule are not exposed.