runbios-sdk 0.2.14-dev.245 → 0.2.14-dev.247

package/README.md

@@ -186,6 +186,37 @@ if (answer) {
 Streaming billing is charged server-side on completed usage; the SDK only needs
 to request `usage` where the endpoint exposes it (no client change).
 
+#### Other supported `/v1` inference tasks
+
+Serverless catalogs serve OpenAI chat and, when the model supports that dialect,
+Anthropic Messages (`inference.messages` / `streamMessages`). Serverless does
+**not** serve `/v1/completions`, `/v1/embeddings`, or `/v1/rerank`. Those three
+routes require a dedicated deployment that actually advertises the matching
+task and a credential with `deployments:read` or `deployments:write`; a chat-only
+deployment cannot embed or rerank. The server's preflight/status result, not the
+model name, determines serving mode. The SDK forwards a dedicated `inferenceKey`
+when configured, otherwise the workspace platform `apiKey`.
+
+```typescript
+import { Inference } from 'runbios-sdk';
+const reply = await client.inference.messages({
+  model: 'catalog-model-with-messages-support', max_tokens: 64,
+  messages: [{ role: 'user', content: 'Hello.' }],
+});
+
+const dedicated = new Inference();
+const text = await dedicated.completions({ model: 'completion-deployment', prompt: 'Continue' });
+const vectors = await dedicated.embeddings({ model: 'embedding-deployment', input: ['First', 'Second'] });
+const ranking = await dedicated.rerank({ model: 'rerank-deployment', query: 'Question', documents: ['A', 'B'] });
+for await (const event of dedicated.streamCompletions({ model: 'completion-deployment', prompt: 'Continue' })) {
+  console.log(event);
+}
+```
+
+`streamMessages` also yields each Anthropic SSE event in order. Both streaming
+methods close the upstream reader when iteration ends. None of the inference
+POSTs are retried automatically, and `idempotencyKey` is sent only if supplied.
+
 #### Workspace serverless usage and limits (read-only)
 
 A workspace-bound platform key or hosted OAuth grant with `analytics:read` can
@@ -251,6 +282,34 @@ const compat = await client.models.getAdapterCompatibility({
 });
 ```
 
+#### Models this credential can invoke
+
+`client.models.list()` reads the unified `GET /v1/models` roster with the SDK's
+platform API key: serverless pool models and workspace deployments appear only
+when that key's scopes permit them. `client.models.retrieve(id)` uses the same
+scope and workspace boundary. The trainable `models.search()` / `models.get()`
+registry above is separate; a registry result does not guarantee that this
+credential can invoke it.
+
+When a dedicated inference key is configured separately, use
+`client.inference.listModels()` and `client.inference.retrieveModel(id)`: they
+use the exact inference key and base URL selected for chat completions. Model
+ids containing `author/name` are encoded safely. A partially readable list
+includes `usf_unreachable_sources`; an unavailable source is not proof there
+are no models in it. Neither SDK retries a failed inference request for you.
+
+```typescript
+const available = await client.inference.listModels();
+for (const model of available.data) console.log(model.id);
+if (available.usf_unreachable_sources?.length) {
+  console.log('Some model sources were unreachable; retry discovery before concluding they are empty.');
+}
+if (available.data.length) {
+  const detail = await client.inference.retrieveModel(available.data[0].id);
+  console.log(detail.id);
+}
+```
+
 ### Datasets
 
 ```typescript

package/dist/index.d.ts

@@ -36,7 +36,7 @@ import { Loop } from './resources/loop.js';
  * SDK version. Sent as part of the User-Agent header.
  * Must match package.json "version" -- enforced by a contract test.
  */
-export declare const VERSION = "0.2.14-dev.245";
+export declare const VERSION = "0.2.14-dev.247";
 export declare class RunBiOS {
     /** Search models, fetch configs, check adapter compatibility. */
     readonly models: Models;
@@ -74,8 +74,8 @@ export { Integrations, type HuggingFaceIntegration, type IntegrationBrowseParams
 export { Training } from './resources/training.js';
 export { Wallet } from './resources/wallet.js';
 export { GPU, type GPURecommendation } from './resources/gpu.js';
-export { Inference, validateChatRequest, parseSSE, buildInferenceRequest, contextSizingBasis, type ContextSizingBasis, CONTEXT_DEFAULT_CEILING, CONTEXT_EDITABLE_FLOOR, type ChatMessage, type FunctionTool, type ChatCompletionParams, type ChatCompletionResponse, type ChatCompletionChunk, type ServerlessUsageWindow, type ServerlessTimeseriesMetric, type ServerlessWorkspaceLimits, type ServerlessUsageEnvelope, type ServerlessUsageOverviewResponse, type ServerlessUsageRowsResponse, type ServerlessUsageTimeseriesResponse, type ServerlessUsageDailyResponse, type ServerlessSavingsTotals, type ServerlessSavingsPeriod, type ServerlessUsageSavingsResponse, } from './resources/inference.js';
-export type { BiOSConfig, PaginatedResponse, ApiErrorBody, AvailableGpuAlternative, CapacityMinimumRequirement, InferenceBookingAccepted, Model, ModelSearchParams, ModelSearchResponse, ModelDetailResponse, ModelConfig, Dataset, DatasetListParams, DatasetListResponse, DatasetUploadParams, DatasetPreview, DatasetPreviewParams, DatasetImportHFParams, DatasetRegisterHFParams, DatasetHubSearchParams, DatasetHubPreviewParams, DatasetValidation, DatasetFormatVariant, DatasetFormatSpec, DatasetFormatSpecs, DatasetStorageUsage, TrainingMethod, RLHFAlgorithm, AdapterType, TrainingJobStatus, TrainingStatusFilter, TrainingCreateParams, TrainingListParams, TrainingListResponse, TrainingJob, TrainingMetrics, TrainingResourceMetrics, TrainingDeviceMetrics, MetricPoint, MetricGraphConfig, TrainingCheckpoint, TrainingLogs, TrainingLogEntry, TrainingStopResponse, TrainingResumeResponse, CanonicalTrainingRequest, TrainingPreflightDataset, TrainingPreflightWarning, TrainingPreflightResponse, TrainingCapabilityChoice, TrainingConfigFieldCapability, TrainingCapabilities, GPUChoice, WalletBalance, Transaction, TransactionListResponse, TransactionListParams, GPUInfo, GPUPricingResponse, GPUOptionsParams, TrainingRecommendParams, TrainingAdvisorVerdict, TrainingAdvisorRecommendation, TrainingAdvisorUserShape, TrainingAdvisorJustification, TrainingAdvisorBasis, GPUOption, GPUOptionSuggestion, GPUOptionsResponse, InferenceStatus, InferenceCreateParams, InferenceUpdateParams, InferenceDeployment, InferenceDeploymentSummary, InferenceCreateResponse, InferenceListResponse, InferenceUpdateResponse, InferenceLifecycleResponse, InferenceDeleteResponse, InferenceGPUOptionMarket, InferenceGPUOption, InferenceGPUAlternative, InferenceGPUOptionsParams, InferenceGPUOptionsResponse, InferenceCanonicalRequest, InferencePreflightWarning, InferencePreflightResponse, AdapterCompatibility, AdapterCompatibilityResponse, AdapterCompatibilityParams, SupportedArchitecture, ArchitectureScope, SupportedArchitecturesResponse, ApiKey, ApiKeyScope, ApiKeyIntrospection, Organization, OrgMember, OrgInvite, Workspace, Integration, IntegrationCreateParams, StorageUsage, StorageObject, PromptTokensDetails, ChatCompletionUsage, TrainingCadence, TrainingCombinator, ServingKind, TrainingRulePausedReason, LoopTrainingMethod, TrainType, GradersScope, TrainingTrigger, TrainingRecipe, TrainingRunState, TrainingVerdict, TrainingDecision, NotifyState, EvaluationStatus, EvaluationItemStatus, EvaluationWinner, EvaluationWarningCode, ConsentVia, TrainingToolCall, TrainingMessage, TrainingRuleServing, TrainingGPURung, TrainingRule, TrainingRuleConsent, TrainingRulePreflightRefusal, TrainingRuleKeyRef, TrainingRuleKeyCheck, TrainingRulePreflight, TrainingRuleBuildSpec, TrainingRuleTriggerInput, TrainingRuleTrainingInput, TrainingRuleDeployInput, TrainingRuleMoneyInput, TrainingRuleEvaluationInput, TrainingRulePromotionInput, TrainingRuleAcceptTerms, TrainingRulePreflightRequest, TrainingRuleCreateRequest, TrainingRuleUpdateRequest, TrainingRuleConsentRequest, TrainingRuleListParams, TrainingRunPromoteRequest, TrainingRunRejectRequest, TrainingRunRollbackRequest, TrainingRunCancelRequest, TrainingRunListParams, TrainingRun, TrainingRunSummary, TrainingRunEvent, TrainingRunLinks, TrainingRunActions, EvaluationRef, EvaluationDecoding, EvaluationJudgeDimension, EvaluationDimensionScore, EvaluationGraderScore, EvaluationWarning, EvaluationMargin, JudgeAgreementDimension, JudgeAgreement, Evaluation, EvaluationItemSide, EvaluationItem, EvaluationItemListParams, JudgeAgreementParams, AgentSettings, AgentSettingsRequest, BenchmarkStatus, BenchmarkSourceKind, BenchmarkRunStatus, BenchmarkScoring, BenchmarkDimensionScore, Benchmark, BenchmarkItem, BenchmarkRun, BenchmarkHistoryPoint, BenchmarkSource, BenchmarkCreateParams, BenchmarkListParams, BenchmarkItemListParams, BenchmarkHistoryParams, BenchmarkRetireParams, TrainingRuleBenchmarkRequest, InferenceAlias, InferenceAliasRequest, TrainingRuleListResponse, TrainingRuleResponse, TrainingRuleMutationResponse, TrainingRuleDeleteResponse, TrainingRunListResponse, TrainingRunResponse, TrainingRunActionResponse, EvaluationItemsResponse, AgentSettingsResponse, InferenceAliasListResponse, InferenceAliasResponse, InferenceAliasDeleteResponse, BenchmarkListResponse, BenchmarkItemsResponse, BenchmarkHistoryResponse, TrainingRuleRunRefusalCode, TrainingRuleMonthlyLimitRefusal, LoopDatasetDeleteRefusalCode, PipelineStatus, PipelineRecipeExploration, PipelineActiveRun, PipelineVersion, Pipeline, PipelineListResponse, PipelineResponse, } from './types.js';
+export { Inference, validateChatRequest, parseSSE, buildInferenceRequest, contextSizingBasis, type ContextSizingBasis, CONTEXT_DEFAULT_CEILING, CONTEXT_EDITABLE_FLOOR, type ChatMessage, type FunctionTool, type ChatCompletionParams, type ChatCompletionResponse, type ChatCompletionChunk, type InferenceVerbParams, type CompletionParams, type EmbeddingParams, type RerankParams, type AnthropicMessageParams, type ServerlessUsageWindow, type ServerlessTimeseriesMetric, type ServerlessWorkspaceLimits, type ServerlessUsageEnvelope, type ServerlessUsageOverviewResponse, type ServerlessUsageRowsResponse, type ServerlessUsageTimeseriesResponse, type ServerlessUsageDailyResponse, type ServerlessSavingsTotals, type ServerlessSavingsPeriod, type ServerlessUsageSavingsResponse, } from './resources/inference.js';
+export type { BiOSConfig, PaginatedResponse, ApiErrorBody, AvailableGpuAlternative, CapacityMinimumRequirement, InferenceBookingAccepted, Model, ModelSearchParams, ModelSearchResponse, ModelDetailResponse, InferenceModel, InferenceModelListResponse, ModelConfig, Dataset, DatasetListParams, DatasetListResponse, DatasetUploadParams, DatasetPreview, DatasetPreviewParams, DatasetImportHFParams, DatasetRegisterHFParams, DatasetHubSearchParams, DatasetHubPreviewParams, DatasetValidation, DatasetFormatVariant, DatasetFormatSpec, DatasetFormatSpecs, DatasetStorageUsage, TrainingMethod, RLHFAlgorithm, AdapterType, TrainingJobStatus, TrainingStatusFilter, TrainingCreateParams, TrainingListParams, TrainingListResponse, TrainingJob, TrainingMetrics, TrainingResourceMetrics, TrainingDeviceMetrics, MetricPoint, MetricGraphConfig, TrainingCheckpoint, TrainingLogs, TrainingLogEntry, TrainingStopResponse, TrainingResumeResponse, CanonicalTrainingRequest, TrainingPreflightDataset, TrainingPreflightWarning, TrainingPreflightResponse, TrainingCapabilityChoice, TrainingConfigFieldCapability, TrainingCapabilities, GPUChoice, WalletBalance, Transaction, TransactionListResponse, TransactionListParams, GPUInfo, GPUPricingResponse, GPUOptionsParams, TrainingRecommendParams, TrainingAdvisorVerdict, TrainingAdvisorRecommendation, TrainingAdvisorUserShape, TrainingAdvisorJustification, TrainingAdvisorBasis, GPUOption, GPUOptionSuggestion, GPUOptionsResponse, InferenceStatus, InferenceCreateParams, InferenceUpdateParams, InferenceDeployment, InferenceDeploymentSummary, InferenceCreateResponse, InferenceListResponse, InferenceUpdateResponse, InferenceLifecycleResponse, InferenceDeleteResponse, InferenceGPUOptionMarket, InferenceGPUOption, InferenceGPUAlternative, InferenceGPUOptionsParams, InferenceGPUOptionsResponse, InferenceCanonicalRequest, InferencePreflightWarning, InferencePreflightResponse, AdapterCompatibility, AdapterCompatibilityResponse, AdapterCompatibilityParams, SupportedArchitecture, ArchitectureScope, SupportedArchitecturesResponse, ApiKey, ApiKeyScope, ApiKeyIntrospection, Organization, OrgMember, OrgInvite, Workspace, Integration, IntegrationCreateParams, StorageUsage, StorageObject, PromptTokensDetails, ChatCompletionUsage, TrainingCadence, TrainingCombinator, ServingKind, TrainingRulePausedReason, LoopTrainingMethod, TrainType, GradersScope, TrainingTrigger, TrainingRecipe, TrainingRunState, TrainingVerdict, TrainingDecision, NotifyState, EvaluationStatus, EvaluationItemStatus, EvaluationWinner, EvaluationWarningCode, ConsentVia, TrainingToolCall, TrainingMessage, TrainingRuleServing, TrainingGPURung, TrainingRule, TrainingRuleConsent, TrainingRulePreflightRefusal, TrainingRuleKeyRef, TrainingRuleKeyCheck, TrainingRulePreflight, TrainingRuleBuildSpec, TrainingRuleTriggerInput, TrainingRuleTrainingInput, TrainingRuleDeployInput, TrainingRuleMoneyInput, TrainingRuleEvaluationInput, TrainingRulePromotionInput, TrainingRuleAcceptTerms, TrainingRulePreflightRequest, TrainingRuleCreateRequest, TrainingRuleUpdateRequest, TrainingRuleConsentRequest, TrainingRuleListParams, TrainingRunPromoteRequest, TrainingRunRejectRequest, TrainingRunRollbackRequest, TrainingRunCancelRequest, TrainingRunListParams, TrainingRun, TrainingRunSummary, TrainingRunEvent, TrainingRunLinks, TrainingRunActions, EvaluationRef, EvaluationDecoding, EvaluationJudgeDimension, EvaluationDimensionScore, EvaluationGraderScore, EvaluationWarning, EvaluationMargin, JudgeAgreementDimension, JudgeAgreement, Evaluation, EvaluationItemSide, EvaluationItem, EvaluationItemListParams, JudgeAgreementParams, AgentSettings, AgentSettingsRequest, BenchmarkStatus, BenchmarkSourceKind, BenchmarkRunStatus, BenchmarkScoring, BenchmarkDimensionScore, Benchmark, BenchmarkItem, BenchmarkRun, BenchmarkHistoryPoint, BenchmarkSource, BenchmarkCreateParams, BenchmarkListParams, BenchmarkItemListParams, BenchmarkHistoryParams, BenchmarkRetireParams, TrainingRuleBenchmarkRequest, InferenceAlias, InferenceAliasRequest, TrainingRuleListResponse, TrainingRuleResponse, TrainingRuleMutationResponse, TrainingRuleDeleteResponse, TrainingRunListResponse, TrainingRunResponse, TrainingRunActionResponse, EvaluationItemsResponse, AgentSettingsResponse, InferenceAliasListResponse, InferenceAliasResponse, InferenceAliasDeleteResponse, BenchmarkListResponse, BenchmarkItemsResponse, BenchmarkHistoryResponse, TrainingRuleRunRefusalCode, TrainingRuleMonthlyLimitRefusal, LoopDatasetDeleteRefusalCode, PipelineStatus, PipelineRecipeExploration, PipelineActiveRun, PipelineVersion, Pipeline, PipelineListResponse, PipelineResponse, } from './types.js';
 /** The closed set of run states a training run never leaves. */
 export { TERMINAL_RUN_STATES } from './types.js';
 /** The most versions a pipeline may be set to make. */

package/dist/index.js

@@ -36,7 +36,7 @@ import { Loop } from './resources/loop.js';
  * SDK version. Sent as part of the User-Agent header.
  * Must match package.json "version" -- enforced by a contract test.
  */
-export const VERSION = '0.2.14-dev.245';
+export const VERSION = '0.2.14-dev.247';
 export class RunBiOS {
     /** Search models, fetch configs, check adapter compatibility. */
     models;

package/dist/resources/inference.d.ts

@@ -1,5 +1,5 @@
 import type { HttpClient } from '../client.js';
-import type { InferenceDeployment, InferenceDeploymentSummary, InferenceBookingAccepted, InferenceCreateParams, InferenceCreateResponse, InferenceDeleteResponse, InferenceGPUOptionsParams, InferenceGPUOptionsResponse, InferenceLifecycleResponse, InferenceListParams, InferenceListResponse, InferenceNotificationListResponse, InferencePreflightResponse, InferenceUpdateResponse, InferenceUpdateParams, InferenceAlias, InferenceAliasRequest, InferenceAliasDeleteResponse } from '../types.js';
+import type { InferenceDeployment, InferenceDeploymentSummary, InferenceModel, InferenceModelListResponse, InferenceBookingAccepted, InferenceCreateParams, InferenceCreateResponse, InferenceDeleteResponse, InferenceGPUOptionsParams, InferenceGPUOptionsResponse, InferenceLifecycleResponse, InferenceListParams, InferenceListResponse, InferenceNotificationListResponse, InferencePreflightResponse, InferenceUpdateResponse, InferenceUpdateParams, InferenceAlias, InferenceAliasRequest, InferenceAliasDeleteResponse } from '../types.js';
 /**
  * Serving context-length policy, owned and enforced by the server. Mirrored
  * here for documentation only -- never to pre-empt a server verdict.
@@ -78,6 +78,28 @@ export interface ChatCompletionParams extends Record<string, unknown> {
 }
 export type ChatCompletionResponse = Record<string, unknown>;
 export type ChatCompletionChunk = Record<string, unknown>;
+export interface InferenceVerbParams extends Record<string, unknown> {
+    model?: string;
+    inferenceKey?: string;
+    idempotencyKey?: string;
+    requestId?: string;
+    signal?: AbortSignal;
+}
+export interface CompletionParams extends InferenceVerbParams {
+    prompt: string | string[] | number[] | number[][];
+}
+export interface EmbeddingParams extends InferenceVerbParams {
+    input: string | string[] | number[] | number[][];
+}
+export interface RerankParams extends InferenceVerbParams {
+    query: string;
+    documents: Array<string | Record<string, unknown>>;
+}
+export interface AnthropicMessageParams extends InferenceVerbParams {
+    messages: Array<Record<string, unknown>>;
+    max_tokens: number;
+    anthropicVersion?: string;
+}
 export type ServerlessUsageWindow = '1h' | '24h' | '7d' | '30d' | '90d';
 export type ServerlessTimeseriesMetric = 'requests' | 'tokens' | 'spend' | 'ttft_p50' | 'ttft_p95' | 'tps';
 export interface ServerlessWorkspaceLimits {
@@ -151,8 +173,8 @@ export declare function validateChatRequest(body: Record<string, unknown>): void
 export declare function parseSSE(body: ReadableStream<Uint8Array>): AsyncGenerator<string>;
 /**
  * Inference surface. Combines control-plane management of model-serving
- * deployments (`/api/inference*`) with OpenAI-compatible key-scoped inference
- * (`/v1/chat/completions`). Requests are dispatched once; an idempotency header
+ * deployments (`/api/inference*`) with key-scoped inference on the supported
+ * `/v1` task routes. Requests are dispatched once; an idempotency header
  * is forwarded but server-side replay is not assumed.
  */
 export declare class Inference {
@@ -330,6 +352,18 @@ export declare class Inference {
     getGPUOptions(params: InferenceGPUOptionsParams): Promise<InferenceGPUOptionsResponse>;
     private prepare;
     private abortContext;
+    private prepareVerb;
+    private sendVerb;
+    private modelRead;
+    private streamVerb;
+    completions(params: CompletionParams): Promise<Record<string, unknown>>;
+    embeddings(params: EmbeddingParams): Promise<Record<string, unknown>>;
+    rerank(params: RerankParams): Promise<Record<string, unknown>>;
+    messages(params: AnthropicMessageParams): Promise<Record<string, unknown>>;
+    streamCompletions(params: CompletionParams): AsyncGenerator<Record<string, unknown>>;
+    streamMessages(params: AnthropicMessageParams): AsyncGenerator<Record<string, unknown>>;
+    listModels(inferenceKey?: string): Promise<InferenceModelListResponse>;
+    retrieveModel(modelId: string, inferenceKey?: string): Promise<InferenceModel>;
     chatCompletions(params: ChatCompletionParams): Promise<ChatCompletionResponse>;
     streamChatCompletions(params: ChatCompletionParams): AsyncGenerator<ChatCompletionChunk>;
 }

package/dist/resources/inference.js

@@ -1,5 +1,5 @@
 import { ApiError, GpuRejectionError, gpuRejectionCodeForReason, envApiKey, envBaseUrl, envInferenceKey } from '../client.js';
-import { readNativeMaxContext } from './models.js';
+import { asInferenceModel, asInferenceModelList, readNativeMaxContext } from './models.js';
 import { normalizeGPUPlacement } from './gpu-priorities.js';
 const FUNCTION_NAME = /^[A-Za-z0-9_-]{1,64}$/;
 const ROLES = new Set(['system', 'developer', 'user', 'assistant', 'tool', 'function']);
@@ -318,8 +318,8 @@ async function apiError(response) {
 }
 /**
  * Inference surface. Combines control-plane management of model-serving
- * deployments (`/api/inference*`) with OpenAI-compatible key-scoped inference
- * (`/v1/chat/completions`). Requests are dispatched once; an idempotency header
+ * deployments (`/api/inference*`) with key-scoped inference on the supported
+ * `/v1` task routes. Requests are dispatched once; an idempotency header
  * is forwarded but server-side replay is not assumed.
  */
 export class Inference {
@@ -830,6 +830,170 @@ export class Inference {
             remove: () => signal?.removeEventListener('abort', relay),
         };
     }
+    prepareVerb(params, stream, anthropic) {
+        const { inferenceKey, idempotencyKey, requestId, signal, anthropicVersion, stream: requestedStream, ...payload } = params;
+        if (requestedStream !== undefined)
+            throw new Error('Choose the streaming method instead of setting stream in a request body');
+        const key = inferenceKey || this.key;
+        if (!key)
+            throw new Error('an inferenceKey is required');
+        const headers = {
+            Authorization: `Bearer ${key}`,
+            Accept: stream ? 'text/event-stream' : 'application/json',
+            'Content-Type': 'application/json',
+            'X-Request-ID': requestId || crypto.randomUUID(),
+        };
+        if (idempotencyKey)
+            headers['Idempotency-Key'] = idempotencyKey;
+        if (anthropic) {
+            if (anthropicVersion !== undefined && typeof anthropicVersion !== 'string') {
+                throw new Error('anthropicVersion must be a string');
+            }
+            headers['Anthropic-Version'] = anthropicVersion || '2023-06-01';
+        }
+        return { body: stream ? { ...payload, stream: true } : payload, headers, signal };
+    }
+    async sendVerb(path, params, anthropic = false) {
+        const prepared = this.prepareVerb(params, false, anthropic);
+        const abort = this.abortContext(prepared.signal);
+        try {
+            const response = await fetch(`${this.baseUrl}${path}`, {
+                method: 'POST', headers: prepared.headers, body: JSON.stringify(prepared.body), signal: abort.controller.signal,
+            });
+            if (!response.ok)
+                throw await apiError(response);
+            let payload;
+            try {
+                payload = await response.json();
+            }
+            catch {
+                throw new ApiError(response.status, { error: 'Inference endpoint returned invalid JSON' });
+            }
+            if (!payload || typeof payload !== 'object' || Array.isArray(payload)) {
+                throw new ApiError(response.status, { error: 'Inference endpoint returned no response object' });
+            }
+            return payload;
+        }
+        catch (error) {
+            if (abort.controller.signal.aborted && !(error instanceof ApiError)) {
+                throw new ApiError(0, { error: String(abort.controller.signal.reason || 'Inference request aborted') });
+            }
+            throw error;
+        }
+        finally {
+            clearTimeout(abort.timeoutId);
+            abort.remove();
+        }
+    }
+    async modelRead(path, inferenceKey) {
+        const key = inferenceKey || this.key;
+        if (!key)
+            throw new Error('an inferenceKey is required');
+        const abort = this.abortContext();
+        try {
+            const response = await fetch(`${this.baseUrl}${path}`, {
+                method: 'GET',
+                headers: { Authorization: `Bearer ${key}`, Accept: 'application/json', 'X-Request-ID': crypto.randomUUID() },
+                signal: abort.controller.signal,
+            });
+            if (!response.ok)
+                throw await apiError(response);
+            try {
+                return await response.json();
+            }
+            catch {
+                throw new ApiError(response.status, { error: { code: 'INVALID_MODEL_RESPONSE', message: 'Inference model discovery returned invalid JSON.' } });
+            }
+        }
+        catch (error) {
+            if (abort.controller.signal.aborted && !(error instanceof ApiError)) {
+                throw new ApiError(0, { error: String(abort.controller.signal.reason || 'Inference model discovery aborted') });
+            }
+            throw error;
+        }
+        finally {
+            clearTimeout(abort.timeoutId);
+            abort.remove();
+        }
+    }
+    async *streamVerb(path, params, anthropic = false) {
+        const prepared = this.prepareVerb(params, true, anthropic);
+        const abort = this.abortContext(prepared.signal);
+        try {
+            const response = await fetch(`${this.baseUrl}${path}`, {
+                method: 'POST', headers: prepared.headers, body: JSON.stringify(prepared.body), signal: abort.controller.signal,
+            });
+            if (!response.ok)
+                throw await apiError(response);
+            const contentType = response.headers.get('content-type')?.toLowerCase() ?? '';
+            if (!contentType.includes('text/event-stream')) {
+                throw new ApiError(response.status, { error: `Expected text/event-stream, received ${contentType || 'no content type'}` });
+            }
+            if (!response.body)
+                throw new ApiError(0, { error: 'Inference stream returned no body' });
+            for await (const data of parseSSE(response.body)) {
+                if (data === '[DONE]') {
+                    if (anthropic)
+                        throw new ApiError(0, { error: 'Anthropic stream ended without message_stop' });
+                    return;
+                }
+                let event;
+                try {
+                    event = JSON.parse(data);
+                }
+                catch {
+                    throw new ApiError(0, { error: 'Invalid JSON SSE event' });
+                }
+                if (!event || typeof event !== 'object' || Array.isArray(event)) {
+                    throw new ApiError(0, { error: 'Inference stream returned an invalid event' });
+                }
+                if ('error' in event)
+                    throw new ApiError(0, event);
+                yield event;
+                if (anthropic && event.type === 'message_stop')
+                    return;
+            }
+            throw new ApiError(0, { error: 'Inference stream ended before completion marker' });
+        }
+        catch (error) {
+            if (abort.controller.signal.aborted && !(error instanceof ApiError)) {
+                throw new ApiError(0, { error: String(abort.controller.signal.reason || 'Inference request aborted') });
+            }
+            throw error;
+        }
+        finally {
+            abort.controller.abort('stream closed');
+            clearTimeout(abort.timeoutId);
+            abort.remove();
+        }
+    }
+    completions(params) {
+        return this.sendVerb('/v1/completions', params);
+    }
+    embeddings(params) {
+        return this.sendVerb('/v1/embeddings', params);
+    }
+    rerank(params) {
+        return this.sendVerb('/v1/rerank', params);
+    }
+    messages(params) {
+        return this.sendVerb('/v1/messages', params, true);
+    }
+    async *streamCompletions(params) {
+        yield* this.streamVerb('/v1/completions', params);
+    }
+    async *streamMessages(params) {
+        yield* this.streamVerb('/v1/messages', params, true);
+    }
+    async listModels(inferenceKey) {
+        return asInferenceModelList(await this.modelRead('/v1/models', inferenceKey));
+    }
+    async retrieveModel(modelId, inferenceKey) {
+        const id = typeof modelId === 'string' ? modelId.trim() : '';
+        if (!id)
+            throw new Error('RunBiOS: modelId is required to retrieve an inference model');
+        return asInferenceModel(await this.modelRead(`/v1/models/${encodeURIComponent(id)}`, inferenceKey));
+    }
     async chatCompletions(params) {
         const prepared = this.prepare(params, false);
         const abort = this.abortContext(prepared.signal);
@@ -882,6 +1046,7 @@ export class Inference {
                 }
                 yield event;
             }
+            throw new ApiError(0, { error: 'Inference stream ended before completion marker [DONE]' });
         }
         catch (error) {
             if (abort.controller.signal.aborted && !(error instanceof ApiError)) {

package/dist/resources/models.d.ts

@@ -1,5 +1,5 @@
 import type { HttpClient } from '../client.js';
-import type { ModelDetailResponse, ModelSearchParams, ModelSearchResponse, ModelConfig, AdapterCompatibilityParams, AdapterCompatibilityResponse, ArchitectureScope, SupportedArchitecturesResponse } from '../types.js';
+import type { InferenceModel, InferenceModelListResponse, ModelDetailResponse, ModelSearchParams, ModelSearchResponse, ModelConfig, AdapterCompatibilityParams, AdapterCompatibilityResponse, ArchitectureScope, SupportedArchitecturesResponse } from '../types.js';
 /**
  * Access the Run BiOS model catalog -- search models, fetch
  * training-relevant configuration, and check adapter compatibility.
@@ -8,6 +8,8 @@ export declare class Models {
     private readonly _http;
     /** @internal */
     constructor(_http: HttpClient);
+    list(): Promise<InferenceModelListResponse>;
+    retrieve(modelId: string): Promise<InferenceModel>;
     /**
      * Search the Run BiOS model catalog -- the platform's own hosted, verified
      * models. Every result is mirrored in Run BiOS storage and can be trained and
@@ -98,6 +100,8 @@ export declare class Models {
         scope?: ArchitectureScope;
     }): Promise<SupportedArchitecturesResponse>;
 }
+export declare function asInferenceModelList(value: unknown): InferenceModelListResponse;
+export declare function asInferenceModel(value: unknown): InferenceModel;
 /** Registry detail path for an `author/name` catalog id, else undefined. @internal */
 export declare function modelDetailPath(modelId: string): string | undefined;
 /**

package/dist/resources/models.js

@@ -1,3 +1,4 @@
+import { ApiError } from '../client.js';
 /**
  * Access the Run BiOS model catalog -- search models, fetch
  * training-relevant configuration, and check adapter compatibility.
@@ -8,6 +9,15 @@ export class Models {
     constructor(_http) {
         this._http = _http;
     }
+    async list() {
+        return asInferenceModelList(await this._http.fetchGet('/v1/models'));
+    }
+    async retrieve(modelId) {
+        const id = typeof modelId === 'string' ? modelId.trim() : '';
+        if (!id)
+            throw new Error('RunBiOS: modelId is required to retrieve an inference model');
+        return asInferenceModel(await this._http.fetchGet(`/v1/models/${encodeURIComponent(id)}`));
+    }
     /**
      * Search the Run BiOS model catalog -- the platform's own hosted, verified
      * models. Every result is mirrored in Run BiOS storage and can be trained and
@@ -135,6 +145,22 @@ export class Models {
         return this._http.fetchGet(`/api/public/serving-architectures?${q}`);
     }
 }
+export function asInferenceModelList(value) {
+    const response = value && typeof value === 'object' ? value : null;
+    const data = response?.data;
+    if (response?.object !== 'list' || !Array.isArray(data) ||
+        !data.every(row => row && typeof row.id === 'string' && row.id.trim())) {
+        throw new ApiError(0, { error: { code: 'INVALID_MODELS_RESPONSE', message: 'Inference model discovery returned an invalid list; retry instead of treating it as empty.' } });
+    }
+    return value;
+}
+export function asInferenceModel(value) {
+    const model = value && typeof value === 'object' ? value : null;
+    if (model?.object !== 'model' || typeof model?.id !== 'string' || !model.id.trim()) {
+        throw new ApiError(0, { error: { code: 'INVALID_MODEL_RESPONSE', message: 'Inference model retrieval returned an invalid model; retry or list available models.' } });
+    }
+    return value;
+}
 /** Registry detail path for an `author/name` catalog id, else undefined. @internal */
 export function modelDetailPath(modelId) {
     const repo = (modelId || '').trim().replace(/^\/+|\/+$/g, '');

package/dist/types.d.ts

@@ -215,6 +215,22 @@ export interface ModelDetailResponse {
      */
     primary_revision?: string;
 }
+export interface InferenceModel {
+    id: string;
+    object: string;
+    owned_by?: string;
+    created?: number;
+    [field: string]: unknown;
+}
+export interface InferenceModelListResponse {
+    object: string;
+    data: InferenceModel[];
+    usf_unreachable_sources?: Array<{
+        source: string;
+        reason: string;
+    }>;
+    [field: string]: unknown;
+}
 export interface ModelSearchParams {
     /**
      * Search text. Becomes the registry's `q` filter -- the ONLY search

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "runbios-sdk",
-  "version": "0.2.14-dev.245",
+  "version": "0.2.14-dev.247",
   "description": "Official TypeScript SDK for the Run BiOS training and deployment platform API",
   "type": "module",
   "main": "./dist/index.js",