Inference API

LLM GATEWAY / INFERENCE V1

Call any model with the SDK you already use.
One gateway, every provider format.

The client-facing model APIs of the LLM Gateway, in each provider's own wire format: OpenAI, Azure OpenAI, Anthropic, AWS Bedrock, Gemini and Vertex AI, and more. Browse every operation and what the gateway adds to it, or import the OpenAPI document into your API gateway.

Import into Kong Open the playground OpenAPI reference

Provider wire formats Guardrails on every call Routing and failover headers

The Inference API is the part of the LLM Gateway your applications call to reach models. Each API family speaks a provider's own wire format, so an existing SDK works by pointing its base URL at the gateway. Requests are logged and checked against the app's access rules and limits; which other gateway features run (guardrails, routing, failover, response cache and more) depends on the route. See Gateway features by route.

This page documents every operation in the OpenAPI 3.1 document and everything the gateway adds on top of the provider's API: credential headers, response headers, error codes and guardrail responses. To place the gateway behind your own API gateway, see Import into your API gateway (Kong). For how one client format reaches any provider, see any-to-any formats.

Choose an API family

Base URL

The spec has one server, https://{region}.quilr.ai, with the region as a variable. Combine the base URL with an operation's path, or set it as your SDK's base URL as described in the integration guide.

Region	Base URL	region value
USA 1 (US Central West) (default)	https://guardrails-usa-1.quilr.ai	guardrails-usa-1
USA 2 (US East)	https://guardrails-usa-2.quilr.ai	guardrails-usa-2
Europe	https://guardrails-europe-1.quilr.ai	guardrails-europe-1
India 1 (Mumbai)	https://guardrails-india-1.quilr.ai	guardrails-india-1
India 2 (Mumbai)	https://guardrails-india-2.quilr.ai	guardrails-india-2
Japan (Tokyo)	https://guardrails-jp-1.quilr.ai	guardrails-jp-1
Global	https://guardrails.quilr.ai	guardrails

Configuration, keys and limits apply per region: use the regional host your organization uses. For a dedicated or on-premises gateway, replace the URL with your gateway's. Your Quilr contact can provide a copy of the spec with your gateway URL as its server, or you can edit servers yourself.

Authentication

Send a Quilr API key (sk-quilr-...) or a Quilr-verified JWT. Every family accepts Authorization: Bearer, plus the header its provider's SDK uses and the key headers common API gateways use. Each operation lists the headers it accepts under ACCEPTS.

• Send one credential header. When Authorization is present it decides: a malformed value is rejected and no other header is tried. Otherwise the protocol header is read (for example api-key, x-api-key or x-goog-api-key), then apikey, ocp-apim-subscription-key and x-apikey.
• Credentials are never forwarded. The gateway removes the credential header before calling the provider, which is authenticated with the credential configured on the app.
• Bedrock. AWS SDKs sign requests with Signature Version 4, using the Quilr API key as the access key id. A Bearer Quilr API key is accepted instead.
• Realtime. Browser clients that cannot set headers on a WebSocket handshake can send the key as a query parameter or a Sec-WebSocket-Protocol entry. Prefer a header: query strings can appear in proxy access logs.

Header	Accepted on	Details
Authorization: Bearer	Every family	Authorization: Bearer with a Quilr API key (sk-quilr-...) or a Quilr-verified JWT. When present, this header decides: a malformed value is rejected and no other header is tried.
Authorization: AWS4-HMAC-SHA256 (SigV4)	AWS Bedrock Runtime	AWS Signature Version 4 (AWS4-HMAC-SHA256 ...) computed by an AWS SDK configured with the gateway as its endpoint; the access key id is the Quilr API key. A Bearer Quilr API key is accepted instead.
api-key	Specification, Mistral, Azure OpenAI, OpenAI Assistants, OpenAI-compatible, OpenAI Realtime, OpenAI Responses, Rerank, Sarvam, TypeSafe System One, xAI	Azure OpenAI style api-key with a Quilr API key (sk-quilr-...) or a Quilr-verified JWT.
x-api-key	Every family	x-api-key with a Quilr API key (sk-quilr-...) or a Quilr-verified JWT.
apikey	Every family	Kong key-auth style apikey header with a Quilr API key (sk-quilr-...) or a Quilr-verified JWT.
x-goog-api-key	Gemini and Vertex AI	Gemini SDK style x-goog-api-key with a Quilr API key (sk-quilr-...) or a Quilr-verified JWT.
ocp-apim-subscription-key	Every family	Azure API Management style header with a Quilr API key (sk-quilr-...) or a Quilr-verified JWT.
api-subscription-key	Sarvam	Sarvam SDK style api-subscription-key with a Quilr API key (sk-quilr-...) or a Quilr-verified JWT.
x-apikey	Every family	Apigee style x-apikey header with a Quilr API key (sk-quilr-...) or a Quilr-verified JWT.

Identity, conversation, provider and prompt headers

Model-call operations also take these optional request headers. Each appears under the operation's parameters.

Header	Purpose	More
X-User-Email	The end user behind the request, for identity-aware policy, limits and logs. Required when the app enforces identity (401 identity_required). A verified JWT or a named self-service key carries its own identity, which this header never replaces.	Identity and network trust
X-Conversation-Id	Groups related requests into one conversation for logs, routing and session pinning. Required when the app enforces conversation ids (400 conversation_id_required).	Conversation grouping
X-Provider-Label	Sends the request to one provider configured on the app (by its label) instead of the app's routing. Also accepted as provider_label in the body. An explicitly selected provider is never moved by failover.	Provider support
X-Provider-Name	Sends the request to the app's provider of this type (for example openai or bedrock) instead of the app's routing. Also accepted as provider in the body. X-Provider-Label wins when both are sent; never moved by failover.	Provider support for provider type names
X-Prompt-Variables	JSON values for {{variable}} placeholders in stored prompts, keyed by prompt reference. Chat, Messages, Responses and generateContent.	Prompt Store template variables

Gateway features by route

Each operation lists the gateway features that apply on its route under GATEWAY (from x-quilr-gateway-features in the spec). Endpoint-family routes (moderations, images, OCR, completions, files, batches, videos, and the Mistral and xAI media, batch and fill-in-the-middle routes) run no guardrails and no failover. For what each guardrail covers per endpoint, see security guardrails endpoint coverage.

Route	Request guardrails	Response guardrails	Guardian	Hallucination	Attachments	Model access	Routing groups	Content routing	Failover	Limits	Token saving	Prompt Store	Universal prompt	Response cache
OpenAI Chat and Completions Create a chat completion	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓
OpenAI Chat and Completions Create a text completion	-	-	-	-	-	✓	-	-	-	✓	-	-	-	-
OpenAI Embeddings Create embeddings	✓	-	-	-	-	✓	-	-	✓	✓	-	-	-	✓
OpenAI Audio Generate speech	✓	-	-	-	-	✓	-	-	-	✓	-	-	-	-
OpenAI Audio Transcribe audio, Translate audio to English	✓	✓	-	-	-	✓	-	-	-	✓	-	-	-	-
OpenAI Images Edit images, Generate images, Create image variations	-	-	-	-	-	✓	-	-	-	✓	-	-	-	-
OpenAI Moderations and OCR Classify content, Extract text from a document	-	-	-	-	-	✓	-	-	-	✓	-	-	-	-
OpenAI Files List files, Upload a file and 3 more	-	-	-	-	-	✓	-	-	-	✓	-	-	-	-
OpenAI Batches List batches, Create a batch and 3 more	-	-	-	-	-	✓	-	-	-	✓	-	-	-	-
OpenAI Videos List videos, Create a video and 4 more	-	-	-	-	-	✓	-	-	-	✓	-	-	-	-
OpenAI Decisions Answer decision questions	✓	-	-	-	-	✓	-	-	-	✓	-	-	-	-
OpenAI Responses Create a model response, Delete a stored response and 3 more	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	-
Azure OpenAI Generate speech	✓	-	-	-	-	✓	-	-	-	✓	-	-	-	-
Azure OpenAI Transcribe audio, Translate audio to English	✓	✓	-	-	-	✓	-	-	-	✓	-	-	-	-
Azure OpenAI Create a chat completion	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓
Azure OpenAI Create embeddings	✓	-	-	-	-	✓	-	-	✓	✓	-	-	-	✓
Azure OpenAI Create a model response, Delete a stored response and 3 more	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	-
OpenAI Realtime Open a Realtime session (WebSocket)	-	-	-	-	-	✓	✓	-	-	✓	-	-	-	-
OpenAI Assistants List assistants, Create an assistant and 21 more	✓	✓	✓	-	-	✓	-	-	-	✓	-	-	-	-
Anthropic Messages Create a message	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓
AWS Bedrock Runtime Converse, Converse with a response stream and 2 more	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	-	-	✓	-
Google Gemini and Vertex AI Batch embed contents (Gemini API path), Count tokens (Gemini API path) and 13 more	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	-
Rerank Rerank documents	✓	-	-	-	-	✓	-	-	-	✓	-	-	-	-
Mistral Create a fill-in-the-middle completion, Extract text from a document	-	-	-	-	-	✓	-	-	-	✓	-	-	-	-
xAI Images and Videos Edit images, Generate images and 4 more	-	-	-	-	-	✓	-	-	-	✓	-	-	-	-
xAI Batches List batches, Create a batch and 5 more	-	-	-	-	-	✓	-	-	-	✓	-	-	-	-
xAI Responses Create a response, Delete a response, Get a response	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	-
Sarvam Transcribe speech, Transcribe and translate speech and 3 more	✓	✓	-	-	-	✓	-	-	-	✓	-	-	-	-
Sarvam Synthesize speech	✓	-	-	-	-	✓	-	-	-	✓	-	-	-	-
TypeSafe System One Answer System One decision questions	✓	-	-	-	-	✓	-	-	-	✓	-	-	-	-

No gateway features (no model call): Anthropic Messages: Count message tokens, Anthropic Messages: List models, Anthropic Messages: Retrieve a model, Anthropic Messages: Get the caller's quota usage, Specification: Get this OpenAPI document, Azure OpenAI: List models, OpenAI Models: List models, OpenAI Models: Retrieve a model, OpenAI Quota Usage: Get my quota usage, OpenAI Responses: Read the caller's quota usage, TypeSafe System One: List decision models.

The OpenAI-compatible files, batches and videos routes are off by default and answer 404 endpoint_not_enabled until they are enabled for the gateway deployment. See endpoint families.

Request and response bodies

Request and response schemas document the main fields only and allow additional fields: the gateway forwards provider options it does not know, so a new provider parameter works without a gateway change. Do not enable strict body validation that rejects unknown fields, in your API gateway or in a client-side validator.

Gateway response headers

On most model routes, provider response headers pass through unchanged; endpoint-family routes do not relay them (the operation's description says so). On top of them, the gateway adds its own headers. A provider can never set an x-quilr-* header: they are removed from upstream responses, so any you receive came from the gateway.

• Correlation. X-Quilr-Request-ID identifies the request in gateway logs. Quote it when you contact support.
• Routing. X-Quilr-Route is the opaque id of the routing rule that chose the upstream (r_ and 16 hex characters, or default), and X-Quilr-Served-Model the model actually called. See routing and failover.
• Retries and failover. X-Quilr-Attempts and X-Quilr-Failover report same-target retries and moves to another target. See failover.
• Response cache. X-Quilr-Cache and Age mark a response served from the app's response cache.
• Quotas and budgets. RateLimit-Policy and RateLimit (IETF format) describe the tightest quota, with X-Quilr-Quota-Remaining, X-Quilr-Quota-Reset, X-Quilr-Quota-Warning, X-Quilr-Budget-Overflow and X-Quilr-Quota-Scope. See rate, token and budget limits.
• Errors. X-Quilr-Error-Source and X-Quilr-Error-Code say who raised an error and its code, Retry-After when to retry, and Link an administrator's help link. A 429 adds X-RateLimit-*, X-Quilr-Quota-Id, X-Quilr-Quota-Blocked-By and X-Should-Retry.

Browsers cannot read these headers

The gateway headers are not CORS-exposed, so cross-origin JavaScript in a browser cannot read them. Server-side clients can.

Every header, with its exact format and when it is sent. Each operation's Responses panel lists the headers of each status.

Header	Sent on	Meaning
Age	Success	Seconds since the cached response was stored (cache hits only).
Link	Errors	<url>; rel="help": help link an administrator attached to this error (buffered responses only).
RateLimit	Success and errors	IETF RateLimit for the tightest quota: limit=<n>, remaining=<n>, reset=<seconds> (reset omitted for lifetime windows). Present when a quota or budget applies to the request (on success, on later errors and on streams). Values are a snapshot taken before the upstream call: this request's request count is included, its tokens and spend are not.
RateLimit-Policy	Success and errors	IETF RateLimit-Policy: one item per enforced quota, <limit>;w=<window seconds>;qu="<metric>";id="<quota id>" (w omitted for lifetime windows; spend in USD). Present when a quota or budget applies to the request (on success, on later errors and on streams). Values are a snapshot taken before the upstream call: this request's request count is included, its tokens and spend are not.
Retry-After	Errors	Seconds to wait before retrying (quota and limit refusals, no available target, or relayed from the provider).
X-Quilr-Attempts	Success and errors	Number of upstream sends made, same-target retries included. Only sent when greater than 1; also on errors.
X-Quilr-Budget-Overflow	Success	true when a quota admitted this request inside its overflow band (at or over its limit but under the hard stop).
X-Quilr-Cache	Success	HIT or MISS when the app's response cache is on and the request was eligible (non-streaming Chat Completions, Anthropic Messages, Embeddings). Absent on bypass. Send Cache-Control: no-cache to skip the lookup (the response is still stored) or no-store to skip both.
X-Quilr-Error-Code	Errors	The error code (see x-quilr-error-codes). Sent on upstream failures, routing and translation errors, and Bedrock and Google-format errors that carry a gateway code; elsewhere the code is in the body.
X-Quilr-Error-Source	Errors	upstream when the provider failed (status and scrubbed reason relayed), gateway for the gateway's own routing refusals.
X-Quilr-Failover	Success and errors	Failover tier of the target that served, when it was not the first choice: route (another member of the route), fallback (an explicit fallback), same_model (the same model on another provider) or app (another target of the app). Also on errors.
X-Quilr-Quota-Blocked-By	Errors	Comma-separated ids of every quota that refused the request (429).
X-Quilr-Quota-Id	Errors	Id of the primary exhausted quota (429 quota_exceeded).
X-Quilr-Quota-Remaining	Success and errors	Remaining allowance of the tightest quota (USD for spend). Present when a quota or budget applies to the request (on success, on later errors and on streams). Values are a snapshot taken before the upstream call: this request's request count is included, its tokens and spend are not.
X-Quilr-Quota-Reset	Success and errors	Reset time of the tightest quota (ISO 8601 UTC); omitted for lifetime windows. Present when a quota or budget applies to the request (on success, on later errors and on streams). Values are a snapshot taken before the upstream call: this request's request count is included, its tokens and spend are not.
X-Quilr-Quota-Scope	Success and errors	node: quota counters are kept per gateway node; shared: counters are shared by every gateway node in the region. Sent whenever any quota header is.
X-Quilr-Quota-Warning	Success	id=<quota id>; pct=<used percent>: the most-used quota at or past one of its alert thresholds.
X-Quilr-Request-ID	Success and errors	Gateway request id (UUID). Quote it to support. Sent on every response of Chat Completions, Anthropic Messages, Embeddings and Bedrock Runtime, and on streams the gateway had to commit early; on other APIs only on errors rewritten by an administrator's friendly-error rule.
X-Quilr-Route	Success	Opaque, stable id of the routing rule that chose the upstream: r_ plus 16 hex digits, or default. Never a rule name; administrators map it to a rule in the routing summary. Sent on successful responses (status below 400) that reached routing, cached responses included.
X-Quilr-Served-Model	Success	Upstream model id the gateway actually called (on a cache hit, the model of the cached response). Same condition as X-Quilr-Route; omitted when the deployment hides model identity, and on streams the gateway had to commit early.
X-Quilr-Token-Count-Source	Success	estimate when the gateway counted tokens itself instead of the provider (the provider has no counting endpoint, or the model is served through another API).
X-RateLimit-Limit	Errors	Limit of the exhausted quota or app limit (429).
X-RateLimit-Remaining	Errors	Remaining allowance of the exhausted quota or app limit (429).
X-RateLimit-Reset	Errors	Unix time (seconds) the exhausted quota or limit resets (429).
X-Should-Retry	Errors	false when retrying cannot help (a budget is exhausted, a zero allowance, or the administrator chose so). OpenAI and Anthropic SDKs honour it.
x-amzn-ErrorType	Errors	Bedrock APIs: AWS exception name of the error, e.g. ThrottlingException.
x-amzn-RequestId	Errors	Bedrock APIs: request id (the gateway's on gateway errors, AWS's on success).

Errors and error codes

Errors come back in the calling family's own format, so your SDK raises its usual exception. Each operation's Responses panel names the error schema it uses:

Error schema	Families	Where the gateway code goes
OpenAIError	OpenAI-compatible, Responses, Azure OpenAI, Assistants, rerank, Mistral, xAI, Sarvam, TypeSafe	error.code, with error.type, error.source, error.help_url and, for quotas, error.quota / error.blocked_by
AnthropicError	Anthropic Messages	error.code (guardrail blocks keep the code but not details)
BedrockError	AWS Bedrock Runtime	reason and X-Quilr-Error-Code, appended to the message as (reason: <code>); the type is in __type and x-amzn-ErrorType
GoogleError	Gemini and Vertex AI	error.details[].reason and X-Quilr-Error-Code

• Who raised it. Read X-Quilr-Error-Source (upstream when the provider failed, gateway for the gateway's own routing refusals) or error.source (upstream). A code marked as also sent by providers does not by itself say who raised the error.
• Stable codes. Codes are only ever added, never renamed. The catalog lists the gateway's own codes (source gateway) and the gateway's names for provider failures (source provider), each with its usual status.
• Help text. Administrators may replace an error's message and attach a help link (help_url and the Link header). Status, type and code never change.
• Gateway response ids. Responses API ids starting resp_qtx_ were issued by the gateway for a translated provider and are valid only against the gateway.

The full catalog from x-quilr-error-codes in the spec:

Search error codesGroupAll groupsaccessattachmentsauthenticationconfigurationguardrailsidentitylimitsprompt_storequotarequestroutingtranslationupstream

94 of 94 codes

Code	Source	Status	Group	Description
missing_api_key	gateway	401	authentication	No gateway API key was sent.
invalid_api_key	gateway	401	authentication	The gateway API key is unknown or revoked, or its app is disabled. Also sent by providers: read X-Quilr-Error-Source.
expired_api_key	gateway	401	authentication	The self-service API key has expired.
jwt_auth_failed	gateway	401	authentication	The bearer token could not be verified.
self_service_identity_required	gateway	401	authentication	A self-service app accepts only its own Quilr API keys.
missing_provider	gateway	500	configuration	The app has no provider configured.
missing_provider_key	gateway	500	configuration	The provider has no credential configured.
unsupported_provider	gateway	400	configuration	The app's provider cannot serve this API.
provider_not_configured	gateway	400	configuration	The requested provider is not configured for this app.
sdk_mode_required	gateway	403	configuration	This endpoint requires a Quilr SDK app key.
internal_error	gateway	500	configuration	An unexpected gateway error.
policy_access_denied	gateway	403	access	A Policy Engine rule denied the request.
policy_model_denied	gateway	403	access	The model is on a rejected-models list.
policy_model_not_allowed	gateway	403	access	The model is outside the allowed-models list that applies.
model_access_denied	gateway	403	access	The model is not enabled for this app.
policy_tool_access_denied	gateway	403	access	A Policy Engine rule denied a tool call.
source_ip_not_allowed	gateway	403	access	The client IP address is not allowed.
policy_source_ip_denied	gateway	403	access	A Policy Engine rule denied the client IP address.
policy_provider_unsupported	gateway	403	access	The routed provider cannot serve this request.
policy_routing_group_wrong_surface	gateway	403	access	The routing group cannot serve this API.
policy_routing_group_not_defined	gateway	403	access	The requested routing group is not defined.
policy_configuration_error	gateway	500	access	The published policy cannot be applied.
policy_evaluation_error	gateway	500	access	The policy could not be evaluated.
identity_check_failed	gateway	401	identity	The user identity could not be verified.
identity_required	gateway	401	identity	The app requires a user identity.
conversation_id_required	gateway	400	identity	The app requires a conversation id.
system_prompt_not_from_store	gateway	400	prompt_store	The system prompt does not come from the Prompt Store.
prompt_store_not_configured	gateway	400	prompt_store	The Prompt Store is not configured for this app.
prompt_store_error	gateway	400	prompt_store	The Prompt Store prompt could not be used.
invalid_prompt_variables	gateway	400	prompt_store	The Prompt Store variables are invalid.
invalid_json	gateway	400	request	The request body is not valid JSON.
missing_required_parameter	gateway	400	request	A required request parameter is missing.
invalid_extra_body	gateway	400	request	The extra_body parameter is invalid.
invalid_request	gateway	400	request	The request is invalid for the provider adapter.
model_not_found	gateway	404	request	The app does not offer the requested model. Also sent by providers: read X-Quilr-Error-Source.
rate_limit_exceeded	gateway	429	limits	The app's request rate limit was reached. Also sent by providers: read X-Quilr-Error-Source.
concurrency_limit_exceeded	gateway	429	limits	The app's concurrency limit was reached.
input_token_limit_exceeded	gateway	429	limits	The app's input-token limit was reached.
output_token_limit_exceeded	gateway	429	limits	The app's output-token limit was reached.
max_tokens_per_request_exceeded	gateway	400	limits	The request asks for more tokens than one request may use.
max_tokens_exceeded	gateway	400	limits	The request asks for more tokens than one request may use (Messages).
model_rate_limit_exceeded	gateway	429	limits	The model's request rate limit was reached.
model_concurrency_limit_exceeded	gateway	429	limits	The model's concurrency limit was reached.
model_input_token_limit_exceeded	gateway	429	limits	The model's input-token limit was reached.
model_output_token_limit_exceeded	gateway	429	limits	The model's output-token limit was reached.
model_max_tokens_per_request_exceeded	gateway	400	limits	The request asks for more tokens than the model allows per request.
rate_or_token_limit_exceeded	gateway	429	limits	A rate or token limit was reached (Bedrock).
limit_exceeded	gateway	429	limits	A usage limit was reached.
quota_exceeded	gateway	429	quota	A usage quota or budget is exhausted.
quota_price_unavailable	gateway	403	quota	The model has no price for a spend budget.
quota_usage_unavailable	gateway	503	quota	The route cannot measure a quota's usage.
quota_accounting_error	gateway	503	quota	Quota usage could not be recorded.
sensitive_data_blocked	gateway	400	guardrails	Sensitive data was detected and blocked.
adversarial_content_blocked	gateway	400	guardrails	Adversarial content (prompt attack) was blocked.
custom_intent_blocked	gateway	400	guardrails	A custom intent was detected and blocked.
language_blocked	gateway	400	guardrails	The language of the content is blocked.
guardian_task_adherence_blocked	gateway	400	guardrails	Guardian blocked a request outside the app's task.
hallucination_blocked	gateway	400	guardrails	The response was blocked as a likely hallucination.
all_targets_unavailable	gateway	503	routing	No provider target could take the request.
embedding_dimension_mismatch	gateway	502	routing	A failover target returned embeddings of a different size.
routing_constraint_unsatisfied	gateway	403	routing	No allowed target satisfies the request's routing constraints.
routing_state_bound_conflict	gateway	409	routing	The request's provider-bound state lives on a target its routing constraints exclude.
routing_constraint_check_unavailable	gateway	503	routing	A routing-constraint check could not be completed.
translation_unconvertible_request	gateway	400	translation	The request cannot be converted to the provider's API.
translation_unsupported_state	gateway	400	translation	The request depends on state the provider cannot use.
upstream_api_unsupported	gateway	400	translation	The provider cannot serve this client API.
state_not_portable	gateway	400	translation	Conversation state cannot move to this provider.
provider_not_conversational	gateway	400	translation	The provider does not serve conversational requests.
unsupported_client_api	gateway	400	translation	The client API is not supported.
translation_internal_error	gateway	500	translation	The request could not be converted.
translation_response_failed	gateway	502	translation	The provider's response could not be converted.
translation_target_configuration_error	gateway	500	translation	The translated provider target is misconfigured.
attachment_type_not_allowed	gateway	415	attachments	An attachment's detected file type is not allowed.
attachment_too_large	gateway	413	attachments	An attachment is larger than the allowed size.
attachment_request_too_large	gateway	413	attachments	The request's attachments are larger in total than allowed.
attachment_too_many	gateway	413	attachments	The request carries more new attachments than allowed.
attachment_too_many_pages	gateway	413	attachments	A document has more pages than allowed.
attachment_unscannable	gateway	422	attachments	An attachment is encrypted, unparseable or a remote reference, and the policy blocks unscannable files.
attachment_blocked	gateway	403	attachments	An attachment was blocked: a DLP finding in its text (sent with the surface's DLP block status, 400) or macros / an attachment that cannot be removed (403).
upstream_bad_request	provider	400	upstream	The provider rejected the request.
upstream_unauthorized	provider	401	upstream	The provider rejected the gateway's credentials. Failover class: auth.
upstream_forbidden	provider	403	upstream	The provider denied access. Failover class: auth.
upstream_not_found	provider	404	upstream	The provider did not find the model or resource. Failover class: model_not_found.
upstream_timeout	provider	504	upstream	The provider timed out. Failover class: timeout.
upstream_conflict	provider	409	upstream	The provider rejected the request as a conflict.
upstream_request_too_large	provider	413	upstream	The provider rejected the request as too large.
upstream_rate_limited	provider	429	upstream	The provider rate-limited the request. Failover class: rate_limit.
upstream_overloaded	provider	529	upstream	The provider is overloaded. Failover class: server_error.
upstream_error	provider	502	upstream	The provider returned an error.
upstream_unreachable	provider	502	upstream	The provider could not be reached. Failover class: connection.
upstream_stream_truncated	provider	502	upstream	The provider's stream ended early.
upstream_stream_malformed	provider	502	upstream	The provider's stream could not be read.
upstream_stream_error	provider	502	upstream	The provider's stream reported an error.
remote_connector_unreachable	provider	503	upstream	The provider could not be reached through its remote connector. Failover class: connection.

Guardrail blocks and redaction

A request or response stopped by a guardrail returns 400 with error type content_policy_violation and one of these codes. Policy and access denials return 403 (policy_error).

Code	Blocked by
sensitive_data_blocked	Sensitive data. Details give only a count, never the detected values.
adversarial_content_blocked	A prompt attack or other adversarial content.
custom_intent_blocked	A custom intent detection.
language_blocked	A language the app does not allow.
guardian_task_adherence_blocked	The Guardian agent's task adherence check.
hallucination_blocked	The hallucination check.

• Anthropic-format errors keep the code but not details. Bedrock and Google-format guardrail blocks carry no code.
• Response checks run on non-streaming responses only.
• Redaction. When a guardrail redacts instead of blocking, detected values are masked in place: fully, with X characters of the same length (text to speech uses REDACTED), or partially, keeping the first and last characters. Request-side redaction changes what the model receives; response-side redaction (non-streaming) changes what the client receives. No field marks a redaction.
• Advisories. When coding-helper checks are on, a non-streaming response may end with an appended paragraph starting Guardian note: about dependency vulnerabilities or newer versions.

See security guardrails for what each check detects and how to configure it.

Streaming

Operations that stream return text/event-stream. Bedrock streaming operations return application/vnd.amazon.eventstream. Disable response buffering for these paths in any proxy in front of the gateway.

• Keep-alive. While a stream is idle for 15 seconds, the gateway writes SSE comment lines (: prefix). SSE parsers ignore them.
• Early start. If the gateway has not answered 15 seconds after admitting a request, it starts a 200 event stream early. A later failure then arrives as an in-stream error event with an added status_code field instead of an HTTP status.
• Bedrock. Bedrock streams end with an exception frame (modelStreamErrorException or internalServerException) when the upstream stream is interrupted.
• Realtime. Realtime operations upgrade to a WebSocket (101 Switching Protocols). Errors found after the upgrade arrive as one Realtime error event, followed by a close with code 1008.

Aliases

Several paths are aliases of a canonical path (marked x-quilr-alias-of in the spec), kept so every SDK's base-URL convention resolves: for example /model/{model_id}/converse and /bedrock-runtime/model/{model_id}/converse, or the Anthropic paths with and without /v1. The reference below lists each operation once, with its aliases under Also served at. An alias takes the same inputs and returns the same responses. Route all of them in your API gateway.

Import into your API gateway (Kong)

The spec is ready to convert with decK. Use the combined document, or one document per family to give each family its own Kong service. To put the Management API behind Kong as well, see Management API behind Kong.

1. Choose your region

The spec has exactly one server, https://{region}.quilr.ai, because openapi2kong pools several servers into one load-balanced upstream. Import one region per Kong setup:

• Get a pinned copy. Add ?region=<value> to the live spec route, GET /llmgateway_inference_apis/v1/openapi.json, and the document lists that region's URL as its only server (for example ?region=guardrails-europe-1 for Europe only). Combine it with ?family=<key> for one family. The route needs a Quilr API key or JWT, and an unknown region returns 400 unknown_region.
• Or set the variable. In a downloaded file, set the region variable's default to your region (or edit servers to your gateway URL) before converting.

Pick a region and family to get the matching live URL:

RegionUSA 1 (US Central West) (guardrails-usa-1)USA 2 (US East) (guardrails-usa-2)Europe (guardrails-europe-1)India 1 (Mumbai) (guardrails-india-1)India 2 (Mumbai) (guardrails-india-2)Japan (Tokyo) (guardrails-jp-1)Global (guardrails)FamilyAll familiesOpenAI-compatibleOpenAI ResponsesAzure OpenAIOpenAI RealtimeOpenAI AssistantsAnthropic MessagesAWS Bedrock RuntimeGemini and Vertex AIRerankMistralxAISarvamTypeSafe System One

Pinned to USA 1 (US Central West)

curl --header "Authorization: Bearer $QUILR_API_KEY" \
  'https://guardrails-usa-1.quilr.ai/llmgateway_inference_apis/v1/openapi.json?region=guardrails-usa-1&family=anthropic_messages' \
  -o anthropic_messages.json

Family	Download	Live, pinned to guardrails-usa-1	Paths
All families	llm-gateway-inference-v1.json	/llmgateway_inference_apis/v1/openapi.json?region=guardrails-usa-1	239
OpenAI-compatible	llm-gateway-inference-v1-openai_compatible.json	/llmgateway_inference_apis/v1/openapi.json?region=guardrails-usa-1&family=openai_compatible	61
OpenAI Responses	llm-gateway-inference-v1-openai_responses.json	/llmgateway_inference_apis/v1/openapi.json?region=guardrails-usa-1&family=openai_responses	16
Azure OpenAI	llm-gateway-inference-v1-azure_openai.json	/llmgateway_inference_apis/v1/openapi.json?region=guardrails-usa-1&family=azure_openai	27
OpenAI Realtime	llm-gateway-inference-v1-openai_realtime.json	/llmgateway_inference_apis/v1/openapi.json?region=guardrails-usa-1&family=openai_realtime	6
OpenAI Assistants	llm-gateway-inference-v1-openai_assistants.json	/llmgateway_inference_apis/v1/openapi.json?region=guardrails-usa-1&family=openai_assistants	69
Anthropic Messages	llm-gateway-inference-v1-anthropic_messages.json	/llmgateway_inference_apis/v1/openapi.json?region=guardrails-usa-1&family=anthropic_messages	7
AWS Bedrock Runtime	llm-gateway-inference-v1-bedrock_runtime.json	/llmgateway_inference_apis/v1/openapi.json?region=guardrails-usa-1&family=bedrock_runtime	8
Gemini and Vertex AI	llm-gateway-inference-v1-google_gemini.json	/llmgateway_inference_apis/v1/openapi.json?region=guardrails-usa-1&family=google_gemini	15
Rerank	llm-gateway-inference-v1-rerank.json	/llmgateway_inference_apis/v1/openapi.json?region=guardrails-usa-1&family=rerank	3
Mistral	llm-gateway-inference-v1-mistral.json	/llmgateway_inference_apis/v1/openapi.json?region=guardrails-usa-1&family=mistral	2
xAI	llm-gateway-inference-v1-xai.json	/llmgateway_inference_apis/v1/openapi.json?region=guardrails-usa-1&family=xai	16
Sarvam	llm-gateway-inference-v1-sarvam.json	/llmgateway_inference_apis/v1/openapi.json?region=guardrails-usa-1&family=sarvam	6
TypeSafe System One	llm-gateway-inference-v1-typesafe.json	/llmgateway_inference_apis/v1/openapi.json?region=guardrails-usa-1&family=typesafe	2

2. Convert and sync

# One family per Kong service: convert each per-family file, then merge.
deck file openapi2kong -s llm-gateway-inference-v1-openai_compatible.json -o openai_compatible.yaml
deck file openapi2kong -s llm-gateway-inference-v1-anthropic_messages.json -o anthropic_messages.yaml
deck file merge openai_compatible.yaml anthropic_messages.yaml -o kong.yaml

# Or the combined spec as one service:
deck file openapi2kong -s llm-gateway-inference-v1.json -o kong.yaml

deck gateway sync kong.yaml

3. Check route settings

Before you sync, check these settings:

Area	What to do
Body validation	Do not enable strict body validation (for example the request-validator plugin with unknown fields rejected). The schemas are permissive by design and the gateway forwards provider options it does not know.
Streaming	Turn off response buffering (response_buffering: false on the route) for SSE and Bedrock eventstream paths, or streams arrive all at once at the end.
Realtime	Realtime paths are WebSocket upgrades. Give their routes the ws / wss protocols (or the http / https protocols your Kong edition passes upgrades through).
Aliases	Route every alias path. The converted config includes them; keep them if you hand-edit routes.
Credentials	The gateway reads Authorization first, then the protocol header (api-key, x-api-key, x-goog-api-key, api-subscription-key), then apikey (Kong key-auth's default), ocp-apim-subscription-key and x-apikey. If you use key-auth, keep hide_credentials off when clients send their Quilr key in apikey, or inject the Quilr key upstream (for example Authorization: Bearer <key> with the request-transformer plugin).
Bedrock SigV4	SigV4 signs the host, path and body. Set preserve_host: true on Bedrock routes and do not rewrite their path or body.
Gemini and Vertex AI	Paths under /vertex_ai/ are a catch-all for Google's resource paths (including :generateContent-style suffixes). A prefix route on /vertex_ai/ with strip_path: false is simpler than the generated per-path routes.
Browser clients	The gateway does not CORS-expose x-quilr-* response headers, so browser JavaScript cannot read them unless your API gateway's CORS configuration exposes them (for example exposed_headers on Kong's CORS plugin).

Find an endpoint

Search by path, family, action or HTTP method. Open an endpoint to see the credentials it accepts, its parameters, request body, every response with its headers, and a cURL example. Operations the LLM Gateway Playground supports have a Try it in the playground link.

Search endpointsResourceAll API familiesOpenAI-compatibleOpenAI ResponsesAzure OpenAIOpenAI RealtimeOpenAI AssistantsAnthropic MessagesAWS Bedrock RuntimeGemini and Vertex AIRerankMistralxAISarvamTypeSafe System OneSpecificationOpenAI Chat and CompletionsOpenAI EmbeddingsOpenAI AudioOpenAI ImagesOpenAI Moderations and OCROpenAI FilesOpenAI BatchesOpenAI VideosOpenAI DecisionsOpenAI ModelsOpenAI Quota UsageOpenAI ResponsesAzure OpenAIOpenAI RealtimeOpenAI AssistantsAnthropic MessagesAWS Bedrock RuntimeGoogle Gemini and Vertex AIRerankMistralxAI Images and VideosxAI BatchesxAI ResponsesSarvamTypeSafe System OneSpecification

124 of 124 operations (plus 115 alias paths, listed under each operation)

POST

/anthropic_messages/v1/messagesCreate a message

POST

/anthropic_messages/v1/messages/count_tokensCount message tokens

GET

/anthropic_messages/v1/modelsList models+1 alias

GET

/anthropic_messages/v1/models/{model_id}Retrieve a model+1 alias

GET

/anthropic_messages/v1/quilr/usageGet the caller's quota usage

GET

/llmgateway_inference_apis/v1/openapi.jsonGet this OpenAPI document

POST

/mistral/v1/fim/completionsCreate a fill-in-the-middle completion

POST

/mistral/v1/ocrExtract text from a document

POST

/model/{model_id}/converseConverse+1 alias

POST

/model/{model_id}/converse-streamConverse with a response stream+1 alias

POST

/model/{model_id}/invokeInvoke a model+1 alias

POST

/model/{model_id}/invoke-with-response-streamInvoke a model with a response stream+1 alias

POST

/openai/deployments/{deployment}/audio/speechGenerate speech+2 aliases

POST

/openai/deployments/{deployment}/audio/transcriptionsTranscribe audio+2 aliases

POST

/openai/deployments/{deployment}/audio/translationsTranslate audio to English+2 aliases

POST

/openai/deployments/{deployment}/chat/completionsCreate a chat completion+2 aliases

POST

/openai/deployments/{deployment}/embeddingsCreate embeddings+2 aliases

POST

/openai/deployments/{deployment}/responsesCreate a model response+1 alias

DELETE

/openai/deployments/{deployment}/responses/{response_id}Delete a stored response+1 alias

GET

/openai/deployments/{deployment}/responses/{response_id}Retrieve a stored response+1 alias

POST

/openai/deployments/{deployment}/responses/{response_id}/cancelCancel a background response+1 alias

GET

/openai/deployments/{deployment}/responses/{response_id}/input_itemsList a response's input items+1 alias

GET

/openai_assistants/v1/assistantsList assistants+2 aliases

POST

/openai_assistants/v1/assistantsCreate an assistant+2 aliases

DELETE

/openai_assistants/v1/assistants/{assistant_id}Delete an assistant+2 aliases

GET

/openai_assistants/v1/assistants/{assistant_id}Retrieve an assistant+2 aliases

POST

/openai_assistants/v1/assistants/{assistant_id}Modify an assistant+2 aliases

POST

/openai_assistants/v1/threadsCreate a thread+2 aliases

POST

/openai_assistants/v1/threads/runsCreate a thread and run it+2 aliases

DELETE

/openai_assistants/v1/threads/{thread_id}Delete a thread+2 aliases

GET

/openai_assistants/v1/threads/{thread_id}Retrieve a thread+2 aliases

POST

/openai_assistants/v1/threads/{thread_id}Modify a thread+2 aliases

GET

/openai_assistants/v1/threads/{thread_id}/messagesList messages+2 aliases

POST

/openai_assistants/v1/threads/{thread_id}/messagesCreate a message+2 aliases

DELETE

/openai_assistants/v1/threads/{thread_id}/messages/{message_id}Delete a message+2 aliases

GET

/openai_assistants/v1/threads/{thread_id}/messages/{message_id}Retrieve a message+2 aliases

POST

/openai_assistants/v1/threads/{thread_id}/messages/{message_id}Modify a message+2 aliases

GET

/openai_assistants/v1/threads/{thread_id}/runsList runs+2 aliases

POST

/openai_assistants/v1/threads/{thread_id}/runsCreate a run+2 aliases

GET

/openai_assistants/v1/threads/{thread_id}/runs/{run_id}Retrieve a run+2 aliases

POST

/openai_assistants/v1/threads/{thread_id}/runs/{run_id}Modify a run+2 aliases

POST

/openai_assistants/v1/threads/{thread_id}/runs/{run_id}/cancelCancel a run+2 aliases

GET

/openai_assistants/v1/threads/{thread_id}/runs/{run_id}/stepsList run steps+2 aliases

GET

/openai_assistants/v1/threads/{thread_id}/runs/{run_id}/steps/{step_id}Retrieve a run step+2 aliases

POST

/openai_assistants/v1/threads/{thread_id}/runs/{run_id}/submit_tool_outputsSubmit tool outputs+2 aliases

GET

/openai_compatible/openai/modelsList models+1 alias

POST

/openai_compatible/v1/audio/speechGenerate speech+1 alias

POST

/openai_compatible/v1/audio/transcriptionsTranscribe audio+1 alias

POST

/openai_compatible/v1/audio/translationsTranslate audio to English+1 alias

GET

/openai_compatible/v1/batchesList batches+1 alias

POST

/openai_compatible/v1/batchesCreate a batch+1 alias

GET

/openai_compatible/v1/batches/{batch_id}Retrieve a batch+1 alias

POST

/openai_compatible/v1/batches/{batch_id}/cancelCancel a batch+1 alias

GET

/openai_compatible/v1/batches/{batch_id}/resultsGet batch results+1 alias

POST

/openai_compatible/v1/chat/completionsCreate a chat completion+1 alias

POST

/openai_compatible/v1/completionsCreate a text completion+1 alias

POST

/openai_compatible/v1/decisionsAnswer decision questions

POST

/openai_compatible/v1/embeddingsCreate embeddings+1 alias

GET

/openai_compatible/v1/filesList files+1 alias

POST

/openai_compatible/v1/filesUpload a file+1 alias

DELETE

/openai_compatible/v1/files/{file_id}Delete a file+1 alias

GET

/openai_compatible/v1/files/{file_id}Retrieve a file+1 alias

GET

/openai_compatible/v1/files/{file_id}/contentDownload file content+1 alias

POST

/openai_compatible/v1/images/editsEdit images+1 alias

POST

/openai_compatible/v1/images/generationsGenerate images+1 alias

POST

/openai_compatible/v1/images/variationsCreate image variations+1 alias

GET

/openai_compatible/v1/modelsList models+1 alias

GET

/openai_compatible/v1/models/{model_id}Retrieve a model+1 alias

POST

/openai_compatible/v1/moderationsClassify content+1 alias

POST

/openai_compatible/v1/ocrExtract text from a document+1 alias

GET

/openai_compatible/v1/quilr/usageGet my quota usage+1 alias

GET

/openai_compatible/v1/videosList videos+1 alias

POST

/openai_compatible/v1/videosCreate a video+1 alias

POST

/openai_compatible/v1/videos/editsEdit a video+1 alias

DELETE

/openai_compatible/v1/videos/{video_id}Delete a video+1 alias

GET

/openai_compatible/v1/videos/{video_id}Retrieve a video job+1 alias

GET

/openai_compatible/v1/videos/{video_id}/contentDownload video content+1 alias

GET

/openai_realtime/v1/realtimeOpen a Realtime session (WebSocket)WebSocket+5 aliases

GET

/openai_responses/v1/quilr/usageRead the caller's quota usage

POST

/openai_responses/v1/responsesCreate a model response+2 aliases

DELETE

/openai_responses/v1/responses/{response_id}Delete a stored response+2 aliases

GET

/openai_responses/v1/responses/{response_id}Retrieve a stored response+2 aliases

POST

/openai_responses/v1/responses/{response_id}/cancelCancel a background response+2 aliases

GET

/openai_responses/v1/responses/{response_id}/input_itemsList a response's input items+2 aliases

POST

/rerank/v1/rerankRerank documents+2 aliases

POST

/sarvam/speech-to-textTranscribe speech

POST

/sarvam/speech-to-text-translateTranscribe and translate speech

POST

/sarvam/text-lidIdentify the language of text

POST

/sarvam/text-to-speechSynthesize speech

POST

/sarvam/translateTranslate text

POST

/sarvam/transliterateTransliterate text

GET

/typesafe/v1/modelsList decision models

POST

/typesafe/v1/systemoneAnswer System One decision questions

POST

/vertex_ai/{api_version}/models/{model}:batchEmbedContentsBatch embed contents (Gemini API path)

POST

/vertex_ai/{api_version}/models/{model}:countTokensCount tokens (Gemini API path)

POST

/vertex_ai/{api_version}/models/{model}:embedContentEmbed content (Gemini API path)

POST

/vertex_ai/{api_version}/models/{model}:generateContentGenerate content (Gemini API path)

POST

/vertex_ai/{api_version}/models/{model}:streamGenerateContentStream generated content (Gemini API path)

POST

/vertex_ai/{api_version}/projects/{project}/locations/{location}/endpoints/openapi/chat/completionsOpenAI-compatible chat completions on Vertex AI

POST

/vertex_ai/{api_version}/projects/{project}/locations/{location}/endpoints/{endpoint}:predictPredict on a Vertex AI endpoint

POST

/vertex_ai/{api_version}/projects/{project}/locations/{location}/publishers/{publisher}/models/{model}:countTokensCount tokens (Vertex AI path)

POST

/vertex_ai/{api_version}/projects/{project}/locations/{location}/publishers/{publisher}/models/{model}:embedContentEmbed content (Vertex AI path)

POST

/vertex_ai/{api_version}/projects/{project}/locations/{location}/publishers/{publisher}/models/{model}:generateContentGenerate content (Vertex AI path)

POST

/vertex_ai/{api_version}/projects/{project}/locations/{location}/publishers/{publisher}/models/{model}:predictPredict (publisher model)

POST

/vertex_ai/{api_version}/projects/{project}/locations/{location}/publishers/{publisher}/models/{model}:rawPredictRaw predict (publisher model)

POST

/vertex_ai/{api_version}/projects/{project}/locations/{location}/publishers/{publisher}/models/{model}:streamGenerateContentStream generated content (Vertex AI path)

POST

/vertex_ai/{api_version}/projects/{project}/locations/{location}/publishers/{publisher}/models/{model}:streamRawPredictStream raw predict (publisher model)

POST

/vertexai/v1/generateContentGenerate content (model in the body)

GET

/xai/v1/batchesList batches

POST

/xai/v1/batchesCreate a batch

GET

/xai/v1/batches/{batch_id}Get a batch

GET

/xai/v1/batches/{batch_id}/requestsList batch requests

POST

/xai/v1/batches/{batch_id}/requestsAdd batch requests

GET

/xai/v1/batches/{batch_id}/resultsList batch results

POST

/xai/v1/batches/{batch_id}:cancelCancel a batch

POST

/xai/v1/images/editsEdit images

POST

/xai/v1/images/generationsGenerate images

POST

/xai/v1/responsesCreate a response

DELETE

/xai/v1/responses/{response_id}Delete a response

GET

/xai/v1/responses/{response_id}Get a response

POST

/xai/v1/videos/editsStart a video edit

POST

/xai/v1/videos/extensionsStart a video extension

POST

/xai/v1/videos/generationsStart a video generation

GET

/xai/v1/videos/{video_id}Get a video job

Endpoint reference

OpenAI-compatible

Point the OpenAI SDK at <gateway>/openai_compatible/v1. Which providers serve each family beyond chat: endpoint families.

POST/openai_compatible/v1/audio/speech

Generate speech

Text to speech. Returns audio bytes in response_format, or a server-sent event stream when stream_format is sse. Body fields and query parameters are forwarded to the provider as sent.

GATEWAYRequest guardrailsModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/openai_compatible/audio/speech

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/audio/speech' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "<model>",
  "input": "Hello"
}'

Authentication·Errors & error codes· Try it in the playground

POST/openai_compatible/v1/audio/transcriptions

Transcribe audio

Speech to text from a multipart upload. Every form field is forwarded. stream=true returns a server-sent event stream when no response guardrail needs the full transcript.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Model access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/openai_compatible/audio/transcriptions

Path, query & header parameters 4

Full request body specification multipart/form-data

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/audio/transcriptions' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --form 'file=@file' \
  --form 'model=<model>'

Authentication·Errors & error codes

POST/openai_compatible/v1/audio/translations

Translate audio to English

Speech to English text from a multipart upload. Every form field is forwarded.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Model access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/openai_compatible/audio/translations

Path, query & header parameters 4

Full request body specification multipart/form-data

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/audio/translations' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --form 'file=@file' \
  --form 'model=<model>'

Authentication·Errors & error codes

GET/openai_compatible/v1/batches

List batches

Lists the provider's batch jobs. Query parameters are forwarded to the provider.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• GET/openai_compatible/batches

Path, query & header parameters 4

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/batches' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/openai_compatible/v1/batches

Create a batch

Starts a batch job from an uploaded input file (OpenAI, Azure OpenAI, Mistral or xAI). Responses use the OpenAI batch shape.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/openai_compatible/batches

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/batches' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "input_file_id": "<input_file_id>",
  "endpoint": "<endpoint>",
  "completion_window": "<completion_window>"
}'

Authentication·Errors & error codes

GET/openai_compatible/v1/batches/{batch_id}

Retrieve a batch

Returns a batch job's status.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• GET/openai_compatible/batches/{batch_id}

Path, query & header parameters 3

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/batches/<batch_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/openai_compatible/v1/batches/{batch_id}/cancel

Cancel a batch

Cancels a running batch job. No request body.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/openai_compatible/batches/{batch_id}/cancel

Path, query & header parameters 6

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/batches/<batch_id>/cancel' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

GET/openai_compatible/v1/batches/{batch_id}/results

Get batch results

Returns a batch's per-request results. xAI providers only (xAI returns results through its API instead of an output file).

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• GET/openai_compatible/batches/{batch_id}/results

Path, query & header parameters 3

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/batches/<batch_id>/results' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/openai_compatible/v1/chat/completions

Create a chat completion

OpenAI Chat Completions. Guardrails scan the request and the response; routing, failover, limits and quotas apply. Returns a server-sent event stream when stream is true.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system promptResponse cache

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/openai_compatible/chat/completions

Path, query & header parameters 5

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 409 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/chat/completions' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "messages": [
    {
      "content": "Hello",
      "role": "user"
    }
  ],
  "model": "gpt-4o-mini"
}'

Authentication·Errors & error codes· Try it in the playground

POST/openai_compatible/v1/completions

Create a text completion

Legacy text completions, including fill in the middle with suffix on providers that support it. Returns a server-sent event stream when stream is true.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/openai_compatible/completions

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/completions' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "<model>",
  "prompt": "<prompt>"
}'

Authentication·Errors & error codes· Try it in the playground

POST/openai_compatible/v1/decisions

Answer decision questions

OpenAI Decisions format for decision models (Microsoft-Decision-1 on Azure, TypeSafe). Text and JSON input only; images are refused. Request bodies may be gzip or deflate encoded; zstd is refused with 415. Unknown fields inside a question are rejected, as OpenAI does.

GATEWAYRequest guardrailsModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 415 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/decisions' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "<model>",
  "input": "Hello",
  "questions": [
    {
      "type": "predicate",
      "instructions": "<instructions>"
    }
  ]
}'

Authentication·Errors & error codes

POST/openai_compatible/v1/embeddings

Create embeddings

Embeds text with the app's embedding model. The input is scanned by the app's request guardrails.

GATEWAYRequest guardrailsModel access rulesFailoverLimits, quotas and budgetsResponse cache

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/openai_compatible/embeddings

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 409 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/embeddings' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "<model>",
  "input": "Hello"
}'

Authentication·Errors & error codes· Try it in the playground

GET/openai_compatible/v1/files

List files

Lists the provider's files. Query parameters are forwarded to the provider.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• GET/openai_compatible/files

Path, query & header parameters 6

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/files' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/openai_compatible/v1/files

Upload a file

Uploads a file to the app's file provider (OpenAI, Azure OpenAI, Mistral or xAI).

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/openai_compatible/files

Path, query & header parameters 4

Full request body specification multipart/form-data

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/files' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --form 'file=@file' \
  --form 'purpose=<purpose>'

Authentication·Errors & error codes

DELETE/openai_compatible/v1/files/{file_id}

Delete a file

Deletes a file at the provider.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• DELETE/openai_compatible/files/{file_id}

Path, query & header parameters 6

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request DELETE \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/files/<file_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

GET/openai_compatible/v1/files/{file_id}

Retrieve a file

Returns a file's metadata.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• GET/openai_compatible/files/{file_id}

Path, query & header parameters 3

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/files/<file_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

GET/openai_compatible/v1/files/{file_id}/content

Download file content

Streams the file's bytes from the provider unchanged, with the provider's content type.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• GET/openai_compatible/files/{file_id}/content

Path, query & header parameters 3

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/files/<file_id>/content' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/openai_compatible/v1/images/edits

Edit images

Edits images from a prompt. Send multipart form data with image files, or JSON with image references; which one a provider accepts depends on the provider (Azure OpenAI takes multipart only, xAI JSON only).

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/openai_compatible/images/edits

Path, query & header parameters 4

Full request body specification application/jsonalso multipart/form-data

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/images/edits' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "prompt": "<prompt>"
}'

Authentication·Errors & error codes

POST/openai_compatible/v1/images/generations

Generate images

Creates images from a prompt.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/openai_compatible/images/generations

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/images/generations' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "prompt": "<prompt>"
}'

Authentication·Errors & error codes· Try it in the playground

POST/openai_compatible/v1/images/variations

Create image variations

Creates variations of an uploaded image. OpenAI providers only.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/openai_compatible/images/variations

Path, query & header parameters 4

Full request body specification multipart/form-data

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/images/variations' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --form 'image=@image'

Authentication·Errors & error codes

GET/openai_compatible/v1/models

List models

Models configured for the app that the caller may use. Makes no upstream call.

GATEWAYNo gateway features (no model call)

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• GET/openai_compatible/models

No request body.

Responses 200 · 400 · 401 · 403 · 500

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/models' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

GET/openai_compatible/v1/models/{model_id}

Retrieve a model

One model from the app's configured models. 404 when the model is not enabled for the caller.

GATEWAYNo gateway features (no model call)

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• GET/openai_compatible/models/{model_id}

Path, query & header parameters 1

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 500

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/models/<model_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/openai_compatible/v1/moderations

Classify content

Moderation in the OpenAI shape; Mistral moderation results are converted to it.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/openai_compatible/moderations

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/moderations' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "input": "Hello"
}'

Authentication·Errors & error codes· Try it in the playground

POST/openai_compatible/v1/ocr

Extract text from a document

Document OCR in Mistral's request and response shape. Mistral providers only.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/openai_compatible/ocr

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/ocr' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "<model>",
  "document": {
    "document_url": "<document_url>"
  }
}'

Authentication·Errors & error codes· Try it in the playground

GET/openai_compatible/v1/quilr/usage

Get my quota usage

The caller's own quota buckets for this app: limit, used, remaining and reset time. Quotas grouped per person are shown only to a verified identity (a verified JWT or a self-service key).

GATEWAYNo gateway features (no model call)

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• GET/v1/quilr/usage

No request body.

Responses 200 · 401 · 403 · 500 · 503

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/quilr/usage' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

GET/openai_compatible/v1/videos

List videos

Lists video jobs. OpenAI providers only. Query parameters are forwarded to the provider.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• GET/openai_compatible/videos

Path, query & header parameters 5

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/videos' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/openai_compatible/v1/videos

Create a video

Starts a video generation job (OpenAI or xAI). Send JSON, or multipart form data with a reference image (OpenAI only).

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/openai_compatible/videos

Path, query & header parameters 4

Full request body specification application/jsonalso multipart/form-data

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/videos' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "prompt": "<prompt>"
}'

Authentication·Errors & error codes

POST/openai_compatible/v1/videos/edits

Edit a video

Starts a video edit job. xAI providers only.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/openai_compatible/videos/edits

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/videos/edits' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "prompt": "<prompt>"
}'

Authentication·Errors & error codes

DELETE/openai_compatible/v1/videos/{video_id}

Delete a video

Deletes a video. OpenAI providers only.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• DELETE/openai_compatible/videos/{video_id}

Path, query & header parameters 6

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request DELETE \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/videos/<video_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

GET/openai_compatible/v1/videos/{video_id}

Retrieve a video job

Returns a video job's status (OpenAI or xAI).

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• GET/openai_compatible/videos/{video_id}

Path, query & header parameters 3

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/videos/<video_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

GET/openai_compatible/v1/videos/{video_id}/content

Download video content

Streams the generated video (or a thumbnail or spritesheet) from the provider unchanged. OpenAI providers only.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• GET/openai_compatible/videos/{video_id}/content

Path, query & header parameters 4

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/v1/videos/<video_id>/content' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

OpenAI Responses

GET/openai_responses/v1/quilr/usage

Read the caller's quota usage

Returns the quota buckets that apply to the caller's Responses requests: limit, used, remaining and reset time per bucket. Quotas grouped by user are shown only to a verified identity (a Quilr-verified JWT or a self-service key). Makes no upstream call.

GATEWAYNo gateway features (no model call)

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

No request body.

Responses 200 · 400 · 401 · 403 · 500 · 503

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_responses/v1/quilr/usage' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/openai_responses/v1/responses

Create a model response

Creates a model response. model selects the model among those configured for the app (the app's routing may choose the upstream). With stream: true the answer is a server-sent event stream of Responses events, ending with response.completed (or response.failed / response.incomplete). Guardrails can redact the input or output, or block it (400 with a guardrail code; policy and access denials are 403). Requests that depend on state held by one upstream (previous_response_id, conversation, stored prompts) stay on that upstream; when that is not possible the request fails with 400 or 409. Responses served by translation to a provider without a Responses API are not stored and cannot be retrieved later.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• POST/openai_responses/openai/responses
• POST/openai_responses/v1/openai/responses

Path, query & header parameters 5

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 409 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_responses/v1/responses' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{}'

Authentication·Errors & error codes· Try it in the playground

DELETE/openai_responses/v1/responses/{response_id}

Delete a stored response

Deletes a stored response at the upstream that created it.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• DELETE/openai_responses/openai/responses/{response_id}
• DELETE/openai_responses/v1/openai/responses/{response_id}

Path, query & header parameters 5

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 409 · 413 · 415 · 422 · 500 · 502 · 503 · 504

curl --request DELETE \
  'https://guardrails-usa-1.quilr.ai/openai_responses/v1/responses/<response_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

GET/openai_responses/v1/responses/{response_id}

Retrieve a stored response

Returns a stored response from the upstream that created it. With stream=true the upstream's event stream is returned as text/event-stream.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• GET/openai_responses/openai/responses/{response_id}
• GET/openai_responses/v1/openai/responses/{response_id}

Path, query & header parameters 4

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 409 · 413 · 415 · 422 · 500 · 502 · 503 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_responses/v1/responses/<response_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/openai_responses/v1/responses/{response_id}/cancel

Cancel a background response

Cancels a response created with background: true. No request body is needed.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• POST/openai_responses/openai/responses/{response_id}/cancel
• POST/openai_responses/v1/openai/responses/{response_id}/cancel

Path, query & header parameters 5

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 409 · 413 · 415 · 422 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_responses/v1/responses/<response_id>/cancel' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

GET/openai_responses/v1/responses/{response_id}/input_items

List a response's input items

Lists the input items of a stored response, paginated by the upstream.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• GET/openai_responses/openai/responses/{response_id}/input_items
• GET/openai_responses/v1/openai/responses/{response_id}/input_items

Path, query & header parameters 6

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 409 · 413 · 415 · 422 · 500 · 502 · 503 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_responses/v1/responses/<response_id>/input_items' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

Azure OpenAI

POST/openai/deployments/{deployment}/audio/speech

Generate speech

Text to speech for the body's model. Returns audio bytes in response_format; with stream_format: sse, server-sent audio events as they arrive.

GATEWAYRequest guardrailsModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• POST/openai_compatible/openai/deployments/{deployment}/audio/speech
• POST/openai_compatible/v1/openai/deployments/{deployment}/audio/speech

Path, query & header parameters 6

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 409 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai/deployments/<deployment>/audio/speech' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "<model>",
  "input": "Hello"
}'

Authentication·Errors & error codes

POST/openai/deployments/{deployment}/audio/transcriptions

Transcribe audio

Speech to text for the form's model. With stream=true the transcript is streamed as server-sent events, unless the app scans the transcript before delivery, in which case it is returned when complete.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Model access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• POST/openai_compatible/openai/deployments/{deployment}/audio/transcriptions
• POST/openai_compatible/v1/openai/deployments/{deployment}/audio/transcriptions

Path, query & header parameters 6

Full request body specification multipart/form-data

Responses 200 · 400 · 401 · 403 · 404 · 409 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai/deployments/<deployment>/audio/transcriptions' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --form 'file=@file' \
  --form 'model=<model>'

Authentication·Errors & error codes

POST/openai/deployments/{deployment}/audio/translations

Translate audio to English

Speech to English text for the form's model.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Model access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• POST/openai_compatible/openai/deployments/{deployment}/audio/translations
• POST/openai_compatible/v1/openai/deployments/{deployment}/audio/translations

Path, query & header parameters 6

Full request body specification multipart/form-data

Responses 200 · 400 · 401 · 403 · 404 · 409 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai/deployments/<deployment>/audio/translations' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --form 'file=@file' \
  --form 'model=<model>'

Authentication·Errors & error codes

POST/openai/deployments/{deployment}/chat/completions

Create a chat completion

Chat completions for a deployment. The deployment name is the model when the body has no model. With stream: true the answer is a server-sent event stream of chunks ending with data: [DONE]. Guardrails can redact, or block (400 with a guardrail code; policy and access denials are 403).

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system promptResponse cache

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• POST/openai_compatible/openai/deployments/{deployment}/chat/completions
• POST/openai_compatible/v1/openai/deployments/{deployment}/chat/completions

Path, query & header parameters 7

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 409 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai/deployments/<deployment>/chat/completions' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "messages": [
    {
      "role": "<role>"
    }
  ]
}'

Authentication·Errors & error codes

POST/openai/deployments/{deployment}/embeddings

Create embeddings

Embeddings for the body's model (the Azure SDK sends the deployment name there).

GATEWAYRequest guardrailsModel access rulesFailoverLimits, quotas and budgetsResponse cache

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• POST/openai_compatible/openai/deployments/{deployment}/embeddings
• POST/openai_compatible/v1/openai/deployments/{deployment}/embeddings

Path, query & header parameters 6

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 409 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai/deployments/<deployment>/embeddings' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "<model>",
  "input": "Hello"
}'

Authentication·Errors & error codes

POST/openai/deployments/{deployment}/responses

Create a model response

Creates a model response. model selects the model among those configured for the app (the app's routing may choose the upstream). With stream: true the answer is a server-sent event stream of Responses events, ending with response.completed (or response.failed / response.incomplete). Guardrails can redact the input or output, or block it (400 with a guardrail code; policy and access denials are 403). Requests that depend on state held by one upstream (previous_response_id, conversation, stored prompts) stay on that upstream; when that is not possible the request fails with 400 or 409. Responses served by translation to a provider without a Responses API are not stored and cannot be retrieved later. Deployment-style path built by older Azure OpenAI SDKs; for create, the deployment name is the model when the body has none, and it is otherwise ignored.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/openai_responses/openai/deployments/{deployment}/responses

Path, query & header parameters 7

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 409 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai/deployments/<deployment>/responses' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{}'

Authentication·Errors & error codes

DELETE/openai/deployments/{deployment}/responses/{response_id}

Delete a stored response

Deletes a stored response at the upstream that created it. Deployment-style path built by older Azure OpenAI SDKs; for create, the deployment name is the model when the body has none, and it is otherwise ignored.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• DELETE/openai_responses/openai/deployments/{deployment}/responses/{response_id}

Path, query & header parameters 7

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 409 · 413 · 415 · 422 · 500 · 502 · 503 · 504

curl --request DELETE \
  'https://guardrails-usa-1.quilr.ai/openai/deployments/<deployment>/responses/<response_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

GET/openai/deployments/{deployment}/responses/{response_id}

Retrieve a stored response

Returns a stored response from the upstream that created it. With stream=true the upstream's event stream is returned as text/event-stream. Deployment-style path built by older Azure OpenAI SDKs; for create, the deployment name is the model when the body has none, and it is otherwise ignored.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• GET/openai_responses/openai/deployments/{deployment}/responses/{response_id}

Path, query & header parameters 6

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 409 · 413 · 415 · 422 · 500 · 502 · 503 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai/deployments/<deployment>/responses/<response_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/openai/deployments/{deployment}/responses/{response_id}/cancel

Cancel a background response

Cancels a response created with background: true. No request body is needed. Deployment-style path built by older Azure OpenAI SDKs; for create, the deployment name is the model when the body has none, and it is otherwise ignored.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/openai_responses/openai/deployments/{deployment}/responses/{response_id}/cancel

Path, query & header parameters 7

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 409 · 413 · 415 · 422 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai/deployments/<deployment>/responses/<response_id>/cancel' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

GET/openai/deployments/{deployment}/responses/{response_id}/input_items

List a response's input items

Lists the input items of a stored response, paginated by the upstream. Deployment-style path built by older Azure OpenAI SDKs; for create, the deployment name is the model when the body has none, and it is otherwise ignored.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• GET/openai_responses/openai/deployments/{deployment}/responses/{response_id}/input_items

Path, query & header parameters 8

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 409 · 413 · 415 · 422 · 500 · 502 · 503 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai/deployments/<deployment>/responses/<response_id>/input_items' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

GET/openai_compatible/openai/models

List models

Models configured for the app and visible to the caller, as the Azure OpenAI SDK's models.list() requests them (not deployment-scoped). Makes no upstream call.

GATEWAYNo gateway features (no model call)

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• GET/openai_compatible/v1/openai/models

Path, query & header parameters 1

No request body.

Responses 200 · 400 · 401 · 403 · 500

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_compatible/openai/models' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

OpenAI Realtime

GET/openai_realtime/v1/realtime

Open a Realtime session (WebSocket)

Upgrades to a WebSocket and relays Realtime client and server events between the caller and the upstream session. model (or the Azure preview deployment) selects the model among those configured for the app; a transcription session (intent=transcription) may omit it and set its model in the session configuration. Other query parameters are forwarded upstream.

Credentials. A credential header is read first: Authorization: Bearer, api-key, x-api-key, then the API-gateway key headers. Only when the handshake carries none of them, the gateway also accepts the key in the api-key, api_key or authorization query parameter, or in a Sec-WebSocket-Protocol entry openai-insecure-api-key.<key> (browser clients). Credentials are never forwarded upstream. Prefer a header: query-string credentials can appear in proxy access logs.

Subprotocols. openai-beta.realtime-v1 and realtime are accepted when offered.

Guardrails do not scan this route.

GATEWAYModel access rulesRouting groupsLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikeyPROTOCOLWebSocket (101)

Also served at 5 alias paths

• GET/openai/realtime
• GET/openai/v1/realtime
• GET/openai_realtime/openai/realtime
• GET/openai_realtime/openai/v1/realtime
• GET/openai_realtime/v1/openai/realtime

Path, query & header parameters 10

No request body.

Responses 101 · 400 · 401 · 403 · 429 · 500 · 502 · 504

# WebSocket upgrade (any client that sends headers on the handshake)
websocat \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  'wss://guardrails-usa-1.quilr.ai/openai_realtime/v1/realtime?model=%3Cmodel%3E'

Authentication·Errors & error codes

OpenAI Assistants

GET/openai_assistants/v1/assistants

List assistants

Lists assistants.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• GET/openai_assistants/openai/assistants
• GET/openai_assistants/v1/openai/assistants

Path, query & header parameters 5

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/assistants' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/openai_assistants/v1/assistants

Create an assistant

Creates an assistant.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• POST/openai_assistants/openai/assistants
• POST/openai_assistants/v1/openai/assistants

Path, query & header parameters 5

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/assistants' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "<model>"
}'

Authentication·Errors & error codes

DELETE/openai_assistants/v1/assistants/{assistant_id}

Delete an assistant

Deletes an assistant.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• DELETE/openai_assistants/openai/assistants/{assistant_id}
• DELETE/openai_assistants/v1/openai/assistants/{assistant_id}

Path, query & header parameters 6

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request DELETE \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/assistants/<assistant_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

GET/openai_assistants/v1/assistants/{assistant_id}

Retrieve an assistant

Retrieves an assistant.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• GET/openai_assistants/openai/assistants/{assistant_id}
• GET/openai_assistants/v1/openai/assistants/{assistant_id}

Path, query & header parameters 2

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/assistants/<assistant_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/openai_assistants/v1/assistants/{assistant_id}

Modify an assistant

Modifies an assistant.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• POST/openai_assistants/openai/assistants/{assistant_id}
• POST/openai_assistants/v1/openai/assistants/{assistant_id}

Path, query & header parameters 6

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/assistants/<assistant_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{}'

Authentication·Errors & error codes

POST/openai_assistants/v1/threads

Create a thread

Creates a thread, optionally with initial messages. Message contents are scanned by guardrails before forwarding.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• POST/openai_assistants/openai/threads
• POST/openai_assistants/v1/openai/threads

Path, query & header parameters 5

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/threads' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{}'

Authentication·Errors & error codes

POST/openai_assistants/v1/threads/runs

Create a thread and run it

Creates a thread and starts a run on it in one request. Message contents are scanned by guardrails before forwarding. With stream: true the response is a server-sent event stream. The streamed model output is scanned when the stream ends; the stored messages are scanned when read.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• POST/openai_assistants/openai/threads/runs
• POST/openai_assistants/v1/openai/threads/runs

Path, query & header parameters 5

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/threads/runs' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "assistant_id": "<assistant_id>"
}'

Authentication·Errors & error codes

DELETE/openai_assistants/v1/threads/{thread_id}

Delete a thread

Deletes a thread.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• DELETE/openai_assistants/openai/threads/{thread_id}
• DELETE/openai_assistants/v1/openai/threads/{thread_id}

Path, query & header parameters 6

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request DELETE \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/threads/<thread_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

GET/openai_assistants/v1/threads/{thread_id}

Retrieve a thread

Retrieves a thread.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• GET/openai_assistants/openai/threads/{thread_id}
• GET/openai_assistants/v1/openai/threads/{thread_id}

Path, query & header parameters 2

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/threads/<thread_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/openai_assistants/v1/threads/{thread_id}

Modify a thread

Modifies a thread's metadata or tool resources.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• POST/openai_assistants/openai/threads/{thread_id}
• POST/openai_assistants/v1/openai/threads/{thread_id}

Path, query & header parameters 6

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/threads/<thread_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{}'

Authentication·Errors & error codes

GET/openai_assistants/v1/threads/{thread_id}/messages

List messages

Lists a thread's messages. Message contents are scanned by guardrails before they are returned.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• GET/openai_assistants/openai/threads/{thread_id}/messages
• GET/openai_assistants/v1/openai/threads/{thread_id}/messages

Path, query & header parameters 7

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/threads/<thread_id>/messages' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/openai_assistants/v1/threads/{thread_id}/messages

Create a message

Adds a message to a thread. The content is scanned by guardrails before forwarding.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• POST/openai_assistants/openai/threads/{thread_id}/messages
• POST/openai_assistants/v1/openai/threads/{thread_id}/messages

Path, query & header parameters 6

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/threads/<thread_id>/messages' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "role": "user",
  "content": "Hello"
}'

Authentication·Errors & error codes

DELETE/openai_assistants/v1/threads/{thread_id}/messages/{message_id}

Delete a message

Deletes a message.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• DELETE/openai_assistants/openai/threads/{thread_id}/messages/{message_id}
• DELETE/openai_assistants/v1/openai/threads/{thread_id}/messages/{message_id}

Path, query & header parameters 7

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request DELETE \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/threads/<thread_id>/messages/<message_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

GET/openai_assistants/v1/threads/{thread_id}/messages/{message_id}

Retrieve a message

Retrieves a message. Its content is scanned by guardrails before it is returned.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• GET/openai_assistants/openai/threads/{thread_id}/messages/{message_id}
• GET/openai_assistants/v1/openai/threads/{thread_id}/messages/{message_id}

Path, query & header parameters 3

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/threads/<thread_id>/messages/<message_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/openai_assistants/v1/threads/{thread_id}/messages/{message_id}

Modify a message

Modifies a message's metadata.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• POST/openai_assistants/openai/threads/{thread_id}/messages/{message_id}
• POST/openai_assistants/v1/openai/threads/{thread_id}/messages/{message_id}

Path, query & header parameters 7

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/threads/<thread_id>/messages/<message_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{}'

Authentication·Errors & error codes

GET/openai_assistants/v1/threads/{thread_id}/runs

List runs

Lists a thread's runs. Tool-call arguments of runs awaiting tool outputs are scanned by guardrails before they are returned.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• GET/openai_assistants/openai/threads/{thread_id}/runs
• GET/openai_assistants/v1/openai/threads/{thread_id}/runs

Path, query & header parameters 6

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/threads/<thread_id>/runs' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/openai_assistants/v1/threads/{thread_id}/runs

Create a run

Starts a run of an assistant on a thread. additional_messages contents are scanned by guardrails before forwarding. With stream: true the response is a server-sent event stream. The streamed model output is scanned when the stream ends; the stored messages are scanned when read.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• POST/openai_assistants/openai/threads/{thread_id}/runs
• POST/openai_assistants/v1/openai/threads/{thread_id}/runs

Path, query & header parameters 7

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/threads/<thread_id>/runs' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "assistant_id": "<assistant_id>"
}'

Authentication·Errors & error codes· Try it in the playground

GET/openai_assistants/v1/threads/{thread_id}/runs/{run_id}

Retrieve a run

Retrieves a run. Tool-call arguments of a run awaiting tool outputs are scanned by guardrails before they are returned.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• GET/openai_assistants/openai/threads/{thread_id}/runs/{run_id}
• GET/openai_assistants/v1/openai/threads/{thread_id}/runs/{run_id}

Path, query & header parameters 3

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/threads/<thread_id>/runs/<run_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/openai_assistants/v1/threads/{thread_id}/runs/{run_id}

Modify a run

Modifies a run's metadata.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• POST/openai_assistants/openai/threads/{thread_id}/runs/{run_id}
• POST/openai_assistants/v1/openai/threads/{thread_id}/runs/{run_id}

Path, query & header parameters 7

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/threads/<thread_id>/runs/<run_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{}'

Authentication·Errors & error codes

POST/openai_assistants/v1/threads/{thread_id}/runs/{run_id}/cancel

Cancel a run

Cancels an in-progress run. No request body.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• POST/openai_assistants/openai/threads/{thread_id}/runs/{run_id}/cancel
• POST/openai_assistants/v1/openai/threads/{thread_id}/runs/{run_id}/cancel

Path, query & header parameters 7

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/threads/<thread_id>/runs/<run_id>/cancel' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

GET/openai_assistants/v1/threads/{thread_id}/runs/{run_id}/steps

List run steps

Lists a run's steps. Tool-call inputs and outputs are scanned by guardrails before they are returned.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• GET/openai_assistants/openai/threads/{thread_id}/runs/{run_id}/steps
• GET/openai_assistants/v1/openai/threads/{thread_id}/runs/{run_id}/steps

Path, query & header parameters 8

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/threads/<thread_id>/runs/<run_id>/steps' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

GET/openai_assistants/v1/threads/{thread_id}/runs/{run_id}/steps/{step_id}

Retrieve a run step

Retrieves a run step. Tool-call inputs and outputs are scanned by guardrails before they are returned.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• GET/openai_assistants/openai/threads/{thread_id}/runs/{run_id}/steps/{step_id}
• GET/openai_assistants/v1/openai/threads/{thread_id}/runs/{run_id}/steps/{step_id}

Path, query & header parameters 5

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/threads/<thread_id>/runs/<run_id>/steps/<step_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/openai_assistants/v1/threads/{thread_id}/runs/{run_id}/submit_tool_outputs

Submit tool outputs

Submits function tool outputs to a run awaiting them. The outputs are scanned by guardrails before forwarding. With stream: true the response is a server-sent event stream. The streamed model output is scanned when the stream ends; the stored messages are scanned when read.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• POST/openai_assistants/openai/threads/{thread_id}/runs/{run_id}/submit_tool_outputs
• POST/openai_assistants/v1/openai/threads/{thread_id}/runs/{run_id}/submit_tool_outputs

Path, query & header parameters 7

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/openai_assistants/v1/threads/<thread_id>/runs/<run_id>/submit_tool_outputs' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "tool_outputs": [
    {
      "output": "<output>"
    }
  ]
}'

Authentication·Errors & error codes

Anthropic Messages

POST/anthropic_messages/v1/messages

Create a message

Anthropic Messages API. Runs the app's guardrails, routing, failover and limits, then calls the provider serving model. With stream: true the response is a server-sent event stream in Anthropic's event format. Other client headers are forwarded to the provider.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system promptResponse cache

ACCEPTSAuthorization: Bearerx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 7

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 409 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/anthropic_messages/v1/messages' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'anthropic-version: 2023-06-01' \
  --header 'Content-Type: application/json' \
  --data '{
  "max_tokens": 1024,
  "messages": [
    {
      "content": "Hello",
      "role": "user"
    }
  ],
  "model": "claude-sonnet-4-5"
}'

Authentication·Errors & error codes· Try it in the playground

POST/anthropic_messages/v1/messages/count_tokens

Count message tokens

Counts input tokens with the provider serving model. Provider errors are returned as the provider sent them. When the provider has no counting endpoint the gateway returns a local estimate and sets the response header x-quilr-token-count-source: estimate.

GATEWAYNo gateway features (no model call)

ACCEPTSAuthorization: Bearerx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/anthropic_messages/v1/messages/count_tokens' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'anthropic-version: 2023-06-01' \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "<model>",
  "messages": [
    {
      "role": "user",
      "content": "Hello"
    }
  ]
}'

Authentication·Errors & error codes

GET/anthropic_messages/v1/models

List models

Models enabled for the app on its Anthropic-family providers, in Anthropic's list format. Makes no upstream call. Models the app hides from listings are left out but stay callable.

GATEWAYNo gateway features (no model call)

ACCEPTSAuthorization: Bearerx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• GET/anthropic_messages/models

Path, query & header parameters 5

No request body.

Responses 200 · 400 · 401 · 403 · 500

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/anthropic_messages/v1/models' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'anthropic-version: 2023-06-01'

Authentication·Errors & error codes

GET/anthropic_messages/v1/models/{model_id}

Retrieve a model

One model enabled for the app on its Anthropic-family providers. Makes no upstream call.

GATEWAYNo gateway features (no model call)

ACCEPTSAuthorization: Bearerx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• GET/anthropic_messages/models/{model_id}

Path, query & header parameters 3

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 500

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/anthropic_messages/v1/models/<model_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'anthropic-version: 2023-06-01'

Authentication·Errors & error codes

GET/anthropic_messages/v1/quilr/usage

Get the caller's quota usage

The caller's own quota buckets for this app: limit, used, remaining and reset time. Quotas grouped per person are included only when the caller's identity is verified (a Quilr-verified JWT or a per-user key). Makes no upstream call.

GATEWAYNo gateway features (no model call)

ACCEPTSAuthorization: Bearerx-api-keyapikeyocp-apim-subscription-keyx-apikey

No request body.

Responses 200 · 401 · 403 · 500

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/anthropic_messages/v1/quilr/usage' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

AWS Bedrock Runtime

POST/model/{model_id}/converse

Converse

Bedrock Converse. Runs the app's guardrails, routing, failover and limits, then calls the model. The app may serve the request from another provider in Converse format.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsUniversal system prompt

ACCEPTSAuthorization: AWS4-HMAC-SHA256 (SigV4)Authorization: Bearerx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/bedrock-runtime/model/{model_id}/converse

Path, query & header parameters 8

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 408 · 409 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/model/<model_id>/converse' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "inferenceConfig": {
    "maxTokens": 512
  },
  "messages": [
    {
      "content": [
        {
          "text": "Hello"
        }
      ],
      "role": "user"
    }
  ]
}'

Authentication·Errors & error codes

POST/model/{model_id}/converse-stream

Converse with a response stream

Bedrock ConverseStream. Same request as Converse; the response is an AWS event stream (messageStart, contentBlockDelta, contentBlockStop, messageStop, metadata, and exception events). Errors before the stream starts are JSON.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsUniversal system prompt

ACCEPTSAuthorization: AWS4-HMAC-SHA256 (SigV4)Authorization: Bearerx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/bedrock-runtime/model/{model_id}/converse-stream

Path, query & header parameters 8

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 408 · 409 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST --no-buffer \
  'https://guardrails-usa-1.quilr.ai/model/<model_id>/converse-stream' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "text": "Hello"
        }
      ]
    }
  ]
}'

Authentication·Errors & error codes

POST/model/{model_id}/invoke

Invoke a model

Bedrock InvokeModel with the model's own request body. Recognisable text is checked by the app's guardrails; the body is otherwise forwarded unchanged, as are X-Amzn-Bedrock-* headers. Content-Type (default application/json) and Accept are forwarded to the model.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsUniversal system prompt

ACCEPTSAuthorization: AWS4-HMAC-SHA256 (SigV4)Authorization: Bearerx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/bedrock-runtime/model/{model_id}/invoke

Path, query & header parameters 13

Full request body specification application/jsonalso */*

Responses 200 · 400 · 401 · 403 · 404 · 408 · 409 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/model/<model_id>/invoke' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{}'

Authentication·Errors & error codes

POST/model/{model_id}/invoke-with-response-stream

Invoke a model with a response stream

Bedrock InvokeModelWithResponseStream. The response is an AWS event stream of chunk events whose bytes carry the model's own streaming payload. Errors before the stream starts are JSON.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsUniversal system prompt

ACCEPTSAuthorization: AWS4-HMAC-SHA256 (SigV4)Authorization: Bearerx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 1 alias path

• POST/bedrock-runtime/model/{model_id}/invoke-with-response-stream

Path, query & header parameters 14

Full request body specification application/jsonalso */*

Responses 200 · 400 · 401 · 403 · 404 · 408 · 409 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST --no-buffer \
  'https://guardrails-usa-1.quilr.ai/model/<model_id>/invoke-with-response-stream' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{}'

Authentication·Errors & error codes

Gemini and Vertex AI

POST/vertex_ai/{api_version}/models/{model}:batchEmbedContents

Batch embed contents (Gemini API path)

Gemini API batchEmbedContents, forwarded to the provider.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerx-api-keyx-goog-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 6

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/vertex_ai/v1/models/<model>:batchEmbedContents' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "requests": [
    {
      "content": {
        "parts": [
          {}
        ]
      }
    }
  ]
}'

Authentication·Errors & error codes

POST/vertex_ai/{api_version}/models/{model}:countTokens

Count tokens (Gemini API path)

Gemini API countTokens, forwarded to the provider. Makes no generation call.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerx-api-keyx-goog-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 6

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/vertex_ai/v1/models/<model>:countTokens' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{}'

Authentication·Errors & error codes

POST/vertex_ai/{api_version}/models/{model}:embedContent

Embed content (Gemini API path)

Gemini API embedContent, forwarded to the provider.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerx-api-keyx-goog-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 6

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/vertex_ai/v1/models/<model>:embedContent' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{}'

Authentication·Errors & error codes

POST/vertex_ai/{api_version}/models/{model}:generateContent

Generate content (Gemini API path)

Gemini API generateContent. Runs the app's guardrails, routing and failover, then calls the provider serving the model. Query parameters other than credentials are forwarded.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerx-api-keyx-goog-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 7

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/vertex_ai/v1/models/<model>:generateContent' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "contents": [
    {
      "parts": [
        {
          "text": "Hello"
        }
      ],
      "role": "user"
    }
  ]
}'

Authentication·Errors & error codes

POST/vertex_ai/{api_version}/models/{model}:streamGenerateContent

Stream generated content (Gemini API path)

Gemini API streamGenerateContent. Runs the app's guardrails, routing and failover, then calls the provider serving the model. Query parameters other than credentials are forwarded. Send alt=sse for server-sent events.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerx-api-keyx-goog-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 8

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/vertex_ai/v1/models/<model>:streamGenerateContent' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "contents": [
    {
      "parts": [
        {
          "text": "Hello"
        }
      ]
    }
  ]
}'

Authentication·Errors & error codes

POST/vertex_ai/{api_version}/projects/{project}/locations/{location}/endpoints/openapi/chat/completions

OpenAI-compatible chat completions on Vertex AI

Vertex AI's OpenAI-compatible chat completions endpoint. The body is an OpenAI chat completions request whose model (for example google/gemini-2.5-flash) is checked against the app's models. Forwarded to the provider; with stream: true the provider's event stream is returned unchanged.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerx-api-keyx-goog-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 8

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/vertex_ai/v1/projects/<project>/locations/<location>/endpoints/openapi/chat/completions' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "<model>",
  "messages": [
    {}
  ]
}'

Authentication·Errors & error codes

POST/vertex_ai/{api_version}/projects/{project}/locations/{location}/endpoints/{endpoint}:predict

Predict on a Vertex AI endpoint

Vertex AI endpoint predict. The model is the one the app maps to this endpoint. Forwarded to the provider; the response comes back unchanged.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerx-api-keyx-goog-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 8

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/vertex_ai/v1/projects/<project>/locations/<location>/endpoints/<endpoint>:predict' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{}'

Authentication·Errors & error codes

POST/vertex_ai/{api_version}/projects/{project}/locations/{location}/publishers/{publisher}/models/{model}:countTokens

Count tokens (Vertex AI path)

Vertex AI countTokens, forwarded to the provider. Makes no generation call.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerx-api-keyx-goog-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 9

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/vertex_ai/v1/projects/<project>/locations/<location>/publishers/google/models/<model>:countTokens' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{}'

Authentication·Errors & error codes

POST/vertex_ai/{api_version}/projects/{project}/locations/{location}/publishers/{publisher}/models/{model}:embedContent

Embed content (Vertex AI path)

Vertex AI embedContent, forwarded to the provider.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerx-api-keyx-goog-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 9

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/vertex_ai/v1/projects/<project>/locations/<location>/publishers/google/models/<model>:embedContent' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{}'

Authentication·Errors & error codes

POST/vertex_ai/{api_version}/projects/{project}/locations/{location}/publishers/{publisher}/models/{model}:generateContent

Generate content (Vertex AI path)

Vertex AI generateContent. Runs the app's guardrails, routing and failover, then calls the provider serving the model. Query parameters other than credentials are forwarded.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerx-api-keyx-goog-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 10

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/vertex_ai/v1/projects/<project>/locations/<location>/publishers/google/models/<model>:generateContent' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "contents": [
    {
      "parts": [
        {
          "text": "Hello"
        }
      ],
      "role": "user"
    }
  ]
}'

Authentication·Errors & error codes

POST/vertex_ai/{api_version}/projects/{project}/locations/{location}/publishers/{publisher}/models/{model}:predict

Predict (publisher model)

Vertex AI predict (for example text embedding and image models). Forwarded to the provider with the app's access rules and limits.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerx-api-keyx-goog-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 9

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/vertex_ai/v1/projects/<project>/locations/<location>/publishers/google/models/<model>:predict' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{}'

Authentication·Errors & error codes

POST/vertex_ai/{api_version}/projects/{project}/locations/{location}/publishers/{publisher}/models/{model}:rawPredict

Raw predict (publisher model)

Vertex AI rawPredict with the publisher's own body (for example Anthropic Messages for Claude on Vertex AI). Forwarded to the provider with the app's access rules and limits.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerx-api-keyx-goog-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 9

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/vertex_ai/v1/projects/<project>/locations/<location>/publishers/google/models/<model>:rawPredict' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{}'

Authentication·Errors & error codes

POST/vertex_ai/{api_version}/projects/{project}/locations/{location}/publishers/{publisher}/models/{model}:streamGenerateContent

Stream generated content (Vertex AI path)

Vertex AI streamGenerateContent. Runs the app's guardrails, routing and failover, then calls the provider serving the model. Query parameters other than credentials are forwarded. Send alt=sse for server-sent events.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerx-api-keyx-goog-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 11

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/vertex_ai/v1/projects/<project>/locations/<location>/publishers/google/models/<model>:streamGenerateContent' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "contents": [
    {
      "parts": [
        {
          "text": "Hello"
        }
      ]
    }
  ]
}'

Authentication·Errors & error codes

POST/vertex_ai/{api_version}/projects/{project}/locations/{location}/publishers/{publisher}/models/{model}:streamRawPredict

Stream raw predict (publisher model)

Vertex AI streamRawPredict with the publisher's own body; the provider's stream is returned unchanged. Forwarded to the provider with the app's access rules and limits.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerx-api-keyx-goog-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 9

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/vertex_ai/v1/projects/<project>/locations/<location>/publishers/google/models/<model>:streamRawPredict' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{}'

Authentication·Errors & error codes

POST/vertexai/v1/generateContent

Generate content (model in the body)

Gemini generateContent with model in the body instead of the path. Runs the app's guardrails, routing and failover, then calls the provider serving the model. Query parameters other than credentials are forwarded. With stream: true the response is server-sent events of GenerateContentResponse objects.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerx-api-keyx-goog-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 5

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/vertexai/v1/generateContent' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "contents": [
    {
      "parts": [
        {
          "text": "Hello"
        }
      ],
      "role": "user"
    }
  ],
  "model": "gemini-2.5-flash"
}'

Authentication·Errors & error codes

Rerank

POST/rerank/v1/rerank

Rerank documents

Orders documents by relevance to query. /rerank/rerank and /rerank/v2/rerank accept the same body and return the same response. Provider errors are returned as the provider sent them.

GATEWAYRequest guardrailsModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Also served at 2 alias paths

• POST/rerank/rerank
• POST/rerank/v2/rerank

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/rerank/v1/rerank' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "documents": [
    "Paris is the capital of France.",
    "Berlin is in Germany."
  ],
  "model": "rerank-v3.5",
  "query": "What is the capital of France?",
  "top_n": 1
}'

Authentication·Errors & error codes· Try it in the playground

Mistral

POST/mistral/v1/fim/completions

Create a fill-in-the-middle completion

Completes code between prompt and suffix. With stream: true the response is a server-sent event stream of Mistral completion chunks ending with data: [DONE].

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/mistral/v1/fim/completions' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "codestral-latest",
  "prompt": "def fib(n):",
  "suffix": "\nprint(fib(10))"
}'

Authentication·Errors & error codes

POST/mistral/v1/ocr

Extract text from a document

Runs Mistral OCR on a document or image and returns each page as Markdown.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/mistral/v1/ocr' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "document": {
    "document_url": "https://example.com/invoice.pdf",
    "type": "document_url"
  },
  "model": "mistral-ocr-latest"
}'

Authentication·Errors & error codes

xAI

GET/xai/v1/batches

List batches

Lists batch jobs, a page at a time.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 4

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/xai/v1/batches' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/xai/v1/batches

Create a batch

Creates a batch job. Add requests with the add batch requests operation, or pass an uploaded input_file_id.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 5

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/xai/v1/batches' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{}'

Authentication·Errors & error codes

GET/xai/v1/batches/{batch_id}

Get a batch

Returns a batch job and its request counters.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 3

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/xai/v1/batches/<batch_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

GET/xai/v1/batches/{batch_id}/requests

List batch requests

Returns a page of a batch's requests and their states.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 5

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/xai/v1/batches/<batch_id>/requests' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/xai/v1/batches/{batch_id}/requests

Add batch requests

Adds requests to a batch job.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 6

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/xai/v1/batches/<batch_id>/requests' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "batch_requests": [
    {
      "batch_request": {}
    }
  ]
}'

Authentication·Errors & error codes

GET/xai/v1/batches/{batch_id}/results

List batch results

Returns a page of a batch's completed results.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 5

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/xai/v1/batches/<batch_id>/results' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/xai/v1/batches/{batch_id}:cancel

Cancel a batch

Cancels a batch job's pending requests. The path ends in the literal :cancel. No request body.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 6

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/xai/v1/batches/<batch_id>:cancel' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/xai/v1/images/edits

Edit images

Edits one or more images from a prompt. JSON bodies only, with images given by URL or data: URI.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/xai/v1/images/edits' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "<model>",
  "prompt": "<prompt>"
}'

Authentication·Errors & error codes

POST/xai/v1/images/generations

Generate images

Generates images from a prompt.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/xai/v1/images/generations' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "<model>",
  "prompt": "<prompt>"
}'

Authentication·Errors & error codes

POST/xai/v1/responses

Create a response

Creates a model response with xAI's Responses API, including xAI server-side tools. Input is scanned by guardrails before forwarding. With stream: true the response is a server-sent event stream of Responses events.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 409 · 413 · 415 · 422 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/xai/v1/responses' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "grok-4",
  "input": "Hello"
}'

Authentication·Errors & error codes

DELETE/xai/v1/responses/{response_id}

Delete a response

Deletes a stored response.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 5

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 413 · 415 · 422 · 429 · 500 · 502 · 504

curl --request DELETE \
  'https://guardrails-usa-1.quilr.ai/xai/v1/responses/<response_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

GET/xai/v1/responses/{response_id}

Get a response

Retrieves a stored response.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Guardian checksHallucination check (non-streaming)Attachment scanningModel access rulesRouting groupsContent routingFailoverLimits, quotas and budgetsToken savingPrompt StoreUniversal system prompt

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 1

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 413 · 415 · 422 · 429 · 500 · 502 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/xai/v1/responses/<response_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/xai/v1/videos/edits

Start a video edit

Starts an asynchronous edit of a video and returns its request_id.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/xai/v1/videos/edits' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "<model>",
  "prompt": "<prompt>"
}'

Authentication·Errors & error codes

POST/xai/v1/videos/extensions

Start a video extension

Starts an asynchronous extension of a video and returns its request_id.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/xai/v1/videos/extensions' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "<model>",
  "prompt": "<prompt>"
}'

Authentication·Errors & error codes

POST/xai/v1/videos/generations

Start a video generation

Starts an asynchronous video generation job and returns its request_id.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/xai/v1/videos/generations' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "<model>",
  "prompt": "<prompt>"
}'

Authentication·Errors & error codes

GET/xai/v1/videos/{video_id}

Get a video job

Returns a video job's status and, when done, the video URL.

Guardrails do not scan this route. Provider response headers are not relayed on this route.

GATEWAYModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 3

No request body.

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 504

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/xai/v1/videos/<video_id>' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

Sarvam

POST/sarvam/speech-to-text

Transcribe speech

Transcribes an uploaded audio file (multipart/form-data). Every form field is forwarded to Sarvam; the transcript is scanned by guardrails before it is returned.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Model access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapi-subscription-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 4

Full request body specification multipart/form-data

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/sarvam/speech-to-text' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --form 'file=@file'

Authentication·Errors & error codes

POST/sarvam/speech-to-text-translate

Transcribe and translate speech

Transcribes an uploaded audio file (multipart/form-data) and translates it to English. The transcript is scanned by guardrails before it is returned.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Model access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapi-subscription-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 4

Full request body specification multipart/form-data

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/sarvam/speech-to-text-translate' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --form 'file=@file'

Authentication·Errors & error codes

POST/sarvam/text-lid

Identify the language of text

Returns the language and script of input. The input is scanned by guardrails before forwarding.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Model access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapi-subscription-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/sarvam/text-lid' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "input": "Hello"
}'

Authentication·Errors & error codes

POST/sarvam/text-to-speech

Synthesize speech

Converts text to speech and returns Sarvam's JSON with base64 audio. The text is scanned by guardrails before forwarding.

GATEWAYRequest guardrailsModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapi-subscription-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/sarvam/text-to-speech' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "text": "Hello",
  "target_language_code": "<target_language_code>"
}'

Authentication·Errors & error codes

POST/sarvam/translate

Translate text

Translates input between languages. The input and the translation are scanned by guardrails.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Model access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapi-subscription-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/sarvam/translate' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "input": "Hello",
  "source_language_code": "<source_language_code>",
  "target_language_code": "<target_language_code>"
}'

Authentication·Errors & error codes

POST/sarvam/transliterate

Transliterate text

Converts input to another script. The input and the result are scanned by guardrails.

GATEWAYRequest guardrailsResponse guardrails (non-streaming)Model access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapi-subscription-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/sarvam/transliterate' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "input": "Hello",
  "source_language_code": "<source_language_code>",
  "target_language_code": "<target_language_code>"
}'

Authentication·Errors & error codes

TypeSafe System One

GET/typesafe/v1/models

List decision models

Decision models configured for the app and allowed for the caller. Makes no upstream call.

GATEWAYNo gateway features (no model call)

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

No request body.

Responses 200 · 401 · 403 · 500

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/typesafe/v1/models' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

POST/typesafe/v1/systemone

Answer System One decision questions

Answers typed questions about state with probabilities. Text and JSON input only. Request bodies may be gzip or deflate encoded; zstd is refused with 415.

GATEWAYRequest guardrailsModel access rulesLimits, quotas and budgets

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 4

Full request body specification application/json

Responses 200 · 400 · 401 · 403 · 404 · 413 · 415 · 429 · 500 · 502 · 503 · 504

curl --request POST \
  'https://guardrails-usa-1.quilr.ai/typesafe/v1/systemone' \
  --header "Authorization: Bearer $QUILR_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "state": "<state>",
  "questions": {}
}'

Authentication·Errors & error codes

Specification

GET/llmgateway_inference_apis/v1/openapi.json

Get this OpenAPI document

Returns the inference OpenAPI document. Pass family for one API family only.

GATEWAYNo gateway features (no model call)

ACCEPTSAuthorization: Bearerapi-keyx-api-keyapikeyocp-apim-subscription-keyx-apikey

Path, query & header parameters 2

No request body.

Responses 200 · 400 · 401

curl --request GET \
  'https://guardrails-usa-1.quilr.ai/llmgateway_inference_apis/v1/openapi.json' \
  --header "Authorization: Bearer $QUILR_API_KEY"

Authentication·Errors & error codes

Related

• Integration guide: base URLs and SDK examples for each format
• Any-to-any formats: call any provider from any client format
• Endpoint families: moderations, images, completions, files, batches, videos and vendor routes
• Routing and failover: what X-Quilr-Route, X-Quilr-Attempts and X-Quilr-Failover report
• Management API: configure apps, providers and controls from scripts