@nemesis-oss/ollama-sdk 1.0.0 → 1.1.0

package/dist/index.d.cts

@@ -1,5 +1,5 @@
-import { M as Message, C as ChatResponse, G as GenerateResponse, T as ToolCall, P as ProgressResponse, a as ModelResponse, S as ShowRequestOptions, b as ShowResponse, c as PullRequestOptions, d as PushRequestOptions, e as CreateRequestOptions, D as DeleteRequestOptions, f as StatusResponse, g as CopyRequestOptions, h as PsResponse, V as VersionResponse, i as ChatRequestOptions, j as GenerateRequestOptions, E as EmbedRequestOptions, k as EmbedResponse, l as EmbeddingsRequestOptions, m as EmbeddingsResponse, W as WebSearchRequestOptions, n as WebSearchResponse, o as WebFetchRequestOptions, p as WebFetchResponse, q as ToolHandler, r as Tool, A as AnyTool, s as ToolRegistry, t as ToolExecutionResult, u as ModelOptions, v as ToolDefinition } from './compose-DC11aDJn.cjs';
-export { w as AppliedSkillResult, x as ApplySkillOptions, F as FormatOption, L as ListResponse, y as ModelDetails, z as ParsedFrontmatter, R as RequestCancellationOptions, B as Role, H as Skill, I as SkillFrontmatter, J as SkillSummary, K as ToolCallFunction, N as ToolExecutionContext, O as ToolExecutionFailure, Q as ToolExecutionSuccess, U as ToolFunctionDefinition, X as ToolProperty, Y as ToolRegistryOptions, Z as WebSearchResult, _ as applySkill, $ as parseFrontmatter } from './compose-DC11aDJn.cjs';
+import { M as Message, C as ChatResponse, G as GenerateResponse, T as ToolCall, P as ProgressResponse, a as ModelResponse, S as ShowRequestOptions, b as ShowResponse, c as PullRequestOptions, d as PushRequestOptions, e as CreateRequestOptions, D as DeleteRequestOptions, f as StatusResponse, g as CopyRequestOptions, h as PsResponse, V as VersionResponse, i as ChatRequestOptions, j as GenerateRequestOptions, E as EmbedRequestOptions, k as EmbedResponse, l as EmbeddingsRequestOptions, m as EmbeddingsResponse, W as WebSearchRequestOptions, n as WebSearchResponse, o as WebFetchRequestOptions, p as WebFetchResponse, q as ToolHandler, r as Tool, A as AnyTool, s as ToolRegistry, t as ToolExecutionResult, u as ModelOptions, v as ToolDefinition } from './compose-DJjVpDrk.cjs';
+export { w as AppliedSkillResult, x as ApplySkillOptions, F as FormatOption, L as ListResponse, y as ModelDetails, z as ParsedFrontmatter, R as RequestCancellationOptions, B as Role, H as Skill, I as SkillFrontmatter, J as SkillSummary, K as ToolCallFunction, N as ToolExecutionContext, O as ToolExecutionFailure, Q as ToolExecutionSuccess, U as ToolFunctionDefinition, X as ToolProperty, Y as ToolRegistryOptions, Z as WebSearchResult, _ as applySkill, $ as parseFrontmatter } from './compose-DJjVpDrk.cjs';
 import { z } from 'zod';
 import * as Otel from '@opentelemetry/api';
 
@@ -202,6 +202,8 @@ interface ChatStreamResult {
     readonly message: Message;
     readonly model: string;
     readonly done: boolean;
+    /** Mirrors the final chunk's `done_reason` (e.g. `"stop"`, `"load"`, `"unload"`). */
+    readonly doneReason?: string | undefined;
     readonly totalDurationMs?: number | undefined;
     readonly usage?: TokenUsage | undefined;
     readonly raw?: ChatResponse | undefined;
@@ -210,6 +212,8 @@ interface GenerateStreamResult {
     readonly response: string;
     readonly model: string;
     readonly done: boolean;
+    /** Mirrors the final chunk's `done_reason` (e.g. `"stop"`, `"load"`, `"unload"`). */
+    readonly doneReason?: string | undefined;
     readonly totalDurationMs?: number | undefined;
     readonly usage?: TokenUsage | undefined;
     readonly raw?: GenerateResponse | undefined;
@@ -568,13 +572,15 @@ declare class ModelsClient {
  *
  * @remarks
  * This is a typed pass-through to Ollama's own OpenAI-compatible `/v1` endpoints — it
- * implements the subset of the OpenAI Chat Completions API that Ollama documents as
- * supported (https://docs.ollama.com/api/openai-compatibility), not the full OpenAI
- * surface (Responses API, hosted tools like web/file search, computer use, code
- * interpreter, or server-side conversation state). Notably, `tools` is supported by
- * Ollama's compat layer but `tool_choice` and `parallel_tool_calls` are explicitly
- * **not** — deliberately omitted from {@link OpenAIChatCompletionRequest} rather than
- * typed as accepted-but-ignored.
+ * implements the subset of the OpenAI API surface that Ollama documents as supported
+ * (https://docs.ollama.com/api/openai-compatibility), not the full OpenAI surface (hosted
+ * tools like web/file search, computer use, code interpreter, or stateful server-side
+ * conversation state). Notably, `tools` is supported by Ollama's compat layer but
+ * `tool_choice` and `parallel_tool_calls` are not — both are still typed on
+ * {@link OpenAIChatCompletionRequest} (accepted-but-ignored) so consumers passing a
+ * standard OpenAI request object don't get spurious type errors; see the `@remarks` on
+ * each field. `/v1/responses` is supported non-statefully — see
+ * {@link OpenAIResponsesRequest}.
  */
 
 interface OpenAIToolCall {
@@ -621,6 +627,34 @@ interface OpenAIChatCompletionRequest {
     readonly frequency_penalty?: number | undefined;
     readonly user?: string | undefined;
     readonly tools?: readonly OpenAITool[] | undefined;
+    /**
+     * @remarks Accepted for OpenAI compatibility but ignored by Ollama — every tool the
+     * model is given remains callable regardless of this value.
+     */
+    readonly tool_choice?: 'none' | 'auto' | 'required' | {
+        readonly type: 'function';
+        readonly function: {
+            readonly name: string;
+        };
+    } | undefined;
+    /**
+     * @remarks Accepted for OpenAI compatibility but ignored by Ollama — Ollama does not
+     * control tool-call parallelism through this flag.
+     */
+    readonly parallel_tool_calls?: boolean | undefined;
+    /**
+     * Effort level for thinking models (e.g. `deepseek-r1`, `qwen3`). Equivalent to
+     * `reasoning.effort`; only effective for models that support reasoning/thinking.
+     */
+    readonly reasoning_effort?: 'high' | 'medium' | 'low' | 'max' | 'none' | undefined;
+    /**
+     * Effort level for thinking models (e.g. `deepseek-r1`, `qwen3`), nested OpenAI-style.
+     * Equivalent to `reasoning_effort`; only effective for models that support
+     * reasoning/thinking.
+     */
+    readonly reasoning?: {
+        readonly effort?: 'high' | 'medium' | 'low' | 'max' | 'none' | undefined;
+    } | undefined;
 }
 interface OpenAIChatCompletionChoice {
     readonly index: number;
@@ -649,12 +683,60 @@ interface OpenAIListModelsResponse {
     readonly object: 'list';
     readonly data: readonly OpenAIModelItem[];
 }
+/**
+ * Request body for OpenAI's Responses API (`/v1/responses`), added in Ollama v0.13.3.
+ *
+ * @remarks
+ * Ollama implements this **non-statefully**: every call is independent, so
+ * {@link previous_response_id} and {@link conversation} — OpenAI's mechanisms for
+ * resuming server-side conversation state — are accepted for compatibility but ignored.
+ * Send the full conversation in `input` on every call, the same as `/api/chat`.
+ */
+interface OpenAIResponsesRequest {
+    readonly model: string;
+    readonly input: string | readonly OpenAIMessage[];
+    readonly instructions?: string | undefined;
+    readonly tools?: readonly OpenAITool[] | undefined;
+    readonly stream?: boolean | undefined;
+    readonly temperature?: number | undefined;
+    readonly top_p?: number | undefined;
+    readonly max_output_tokens?: number | undefined;
+    /** @remarks Not supported by Ollama, which is stateless across calls — accepted but ignored. */
+    readonly previous_response_id?: string | undefined;
+    /** @remarks Not supported by Ollama, which is stateless across calls — accepted but ignored. */
+    readonly conversation?: string | undefined;
+    readonly truncation?: string | undefined;
+}
+interface OpenAIResponsesOutputTextContent {
+    readonly type: 'output_text';
+    readonly text: string;
+}
+interface OpenAIResponsesOutputMessage {
+    readonly type: 'message';
+    readonly role?: string | undefined;
+    readonly content: readonly OpenAIResponsesOutputTextContent[];
+}
+/** Non-stateful response shape for `/v1/responses`; see {@link OpenAIResponsesRequest}. */
+interface OpenAIResponsesResponse {
+    readonly id: string;
+    readonly object: 'response';
+    readonly created: number;
+    readonly model: string;
+    readonly output: readonly OpenAIResponsesOutputMessage[];
+    readonly usage?: {
+        readonly input_tokens: number;
+        readonly output_tokens: number;
+        readonly total_tokens: number;
+    } | undefined;
+}
 declare class OpenAICompatClient {
     private readonly http;
     constructor(http: HttpClient);
     createChatCompletion(request: OpenAIChatCompletionRequest, signal?: AbortSignal): Promise<OpenAIChatCompletionResponse>;
     chatCompletions(request: OpenAIChatCompletionRequest, signal?: AbortSignal): Promise<OpenAIChatCompletionResponse>;
     listModels(signal?: AbortSignal): Promise<OpenAIListModelsResponse>;
+    createResponses(request: OpenAIResponsesRequest, signal?: AbortSignal): Promise<OpenAIResponsesResponse>;
+    responses(request: OpenAIResponsesRequest, signal?: AbortSignal): Promise<OpenAIResponsesResponse>;
 }
 
 /**
@@ -1041,4 +1123,19 @@ declare const ATTR_OLLAMA_AGENT_MAX_ITERATIONS = "ollama.agent.max_iterations";
 declare const ATTR_OLLAMA_AGENT_ITERATION = "ollama.agent.iteration";
 declare const GEN_AI_SYSTEM_OLLAMA = "ollama";
 
-export { ATTR_GEN_AI_OPERATION_NAME, ATTR_GEN_AI_REQUEST_MODEL, ATTR_GEN_AI_RESPONSE_MODEL, ATTR_GEN_AI_SYSTEM, ATTR_GEN_AI_TOOL_CALL_ID, ATTR_GEN_AI_TOOL_NAME, ATTR_GEN_AI_USAGE_INPUT_TOKENS, ATTR_GEN_AI_USAGE_OUTPUT_TOKENS, ATTR_HTTP_REQUEST_METHOD, ATTR_HTTP_RESPONSE_STATUS_CODE, ATTR_OLLAMA_AGENT_ITERATION, ATTR_OLLAMA_AGENT_MAX_ITERATIONS, ATTR_OLLAMA_ENDPOINT_ATTEMPT, ATTR_OLLAMA_ENDPOINT_NAME, ATTR_SERVER_ADDRESS, ATTR_SERVER_PORT, ATTR_URL_FULL, type AbortableAsyncIterable, Agent, type AgentChatClient, type AgentConfig, type AgentHooks, type AgentResult, type AgentRunInput, type AgentTurn, type AnthropicCacheControl, AnthropicCompatClient, type AnthropicContentBlock, type AnthropicMessage, type AnthropicMessagesRequest, type AnthropicMessagesResponse, AnyTool, type BackoffOptions, ChatRequestOptions, ChatResponse, type ChatStreamResult, CopyRequestOptions, CreateRequestOptions, DEFAULT_BACKOFF, DEFAULT_BASE_URL, DEFAULT_FAILOVER_CODES, DEFAULT_RETRY_CONFIG, DEFAULT_TIMEOUT_MS, type DefineToolOptions, DeleteRequestOptions, type DoneEvent, EmbedRequestOptions, EmbedResponse, EmbeddingsRequestOptions, EmbeddingsResponse, type EndpointHealth, type EndpointHealthCheckResult, EndpointRegistry, type EndpointRegistryOptions, type ErrorEvent, type FetchLike, GEN_AI_SYSTEM_OLLAMA, GenerateRequestOptions, GenerateResponse, type GenerateStreamResult, type HttpBody, HttpClient, type HttpClientOptions, type HttpRequestOptions, type LifecycleErrorEvent, type LifecycleRetryEvent, type LifecycleStartEvent, type LifecycleSuccessEvent, type LoadMcpToolsOptions, type LogFn, type Logger, type McpCallToolResult, type McpClientLike, type McpContentBlock, type McpListToolsResult, type McpToolDescriptor, Message, type MessageEvent, type Middleware, type MiddlewareContext, type ModelCapabilities, ModelOptions, ModelResponse, ModelsClient, NANOS_PER_MS, NANOS_PER_SECOND, NOOP_LOGGER, type NextFunction, OllamaAbortError, OllamaAgentMaxIterationsError, OllamaAuthError, OllamaClient, type OllamaClientConfig, OllamaClientError, type OllamaClientErrorOptions, type OllamaEndpoint, type OllamaErrorRequestContext, type OllamaErrorResponseContext, OllamaGenericClientError, OllamaMcpError, OllamaNetworkError, OllamaNotFoundError, OllamaRateLimitError, OllamaServerError, OllamaSkillInvalidError, OllamaSkillNotFoundError, OllamaStream, type OllamaStreamEvent, type OllamaStreamEventType, OllamaTimeoutError, OllamaToolTimeoutError, OllamaToolValidationError, OllamaUnsupportedCapabilityError, type OpenAIChatCompletionChoice, type OpenAIChatCompletionRequest, type OpenAIChatCompletionResponse, OpenAICompatClient, type OpenAIFunctionDefinition, type OpenAIListModelsResponse, type OpenAIMessage, type OpenAIModelItem, type OpenAIStreamOptions, type OpenAITool, type OpenAIToolCall, ProgressResponse, type ProgressStreamResult, PsResponse, PullRequestOptions, PushRequestOptions, type RawUsageSource, type RequestContext, type RequestLifecycleEvent, type RequestLifecycleHook, type RequestRunner, type ResponseContext, type RetryConfig, type RuntimeMode, ShowRequestOptions, ShowResponse, type SpanAttributeValue, type SpanAttributes, StatusResponse, type SupportedSchema, type ThinkingEvent, type TimeoutSignal, type TokenEvent, type TokenUsage, Tool, ToolCall, type ToolCallEvent, ToolDefinition, ToolExecutionResult, ToolHandler, ToolRegistry, VersionResponse, WebFetchRequestOptions, WebFetchResponse, WebSearchRequestOptions, WebSearchResponse, addSpanEvent, calculateBackoff, checkEndpointHealth, composeMiddleware, createConsoleLogger, createTimeoutSignal, defineTool, detectModelCapabilities, extractUsage, inferRuntimeMode, listAvailableModels, loadMcpTools, mapError, normalizeChatStream, normalizeGenerateStream, normalizeProgressStream, parseNdjsonStream, parseStructuredOutput, registerMcpTools, setSpanError, toDataStream, toResponse, toTextStream, withRetry, withSpan, zodToJsonSchema };
+/**
+ * Cross-runtime helpers with no Node.js API dependency, so they stay safe to import from
+ * the core client bundle (see ADR 0006 — Edge Runtime CI).
+ */
+/**
+ * Encodes an image to a base64 string, matching the wire format Ollama's `/api/chat` and
+ * `/api/generate` `images` arrays expect.
+ *
+ * - A `string` is assumed to already be base64-encoded and is returned unchanged.
+ * - A `Uint8Array` is base64-encoded using whichever universal primitive the runtime
+ *   exposes: `Buffer` on Node.js, `btoa` on browsers/Edge runtimes. Chunked so it doesn't
+ *   blow the engine's max call-stack/argument-count limits on large images.
+ */
+declare function encodeImage(image: Uint8Array | string): Promise<string>;
+
+export { ATTR_GEN_AI_OPERATION_NAME, ATTR_GEN_AI_REQUEST_MODEL, ATTR_GEN_AI_RESPONSE_MODEL, ATTR_GEN_AI_SYSTEM, ATTR_GEN_AI_TOOL_CALL_ID, ATTR_GEN_AI_TOOL_NAME, ATTR_GEN_AI_USAGE_INPUT_TOKENS, ATTR_GEN_AI_USAGE_OUTPUT_TOKENS, ATTR_HTTP_REQUEST_METHOD, ATTR_HTTP_RESPONSE_STATUS_CODE, ATTR_OLLAMA_AGENT_ITERATION, ATTR_OLLAMA_AGENT_MAX_ITERATIONS, ATTR_OLLAMA_ENDPOINT_ATTEMPT, ATTR_OLLAMA_ENDPOINT_NAME, ATTR_SERVER_ADDRESS, ATTR_SERVER_PORT, ATTR_URL_FULL, type AbortableAsyncIterable, Agent, type AgentChatClient, type AgentConfig, type AgentHooks, type AgentResult, type AgentRunInput, type AgentTurn, type AnthropicCacheControl, AnthropicCompatClient, type AnthropicContentBlock, type AnthropicMessage, type AnthropicMessagesRequest, type AnthropicMessagesResponse, AnyTool, type BackoffOptions, ChatRequestOptions, ChatResponse, type ChatStreamResult, CopyRequestOptions, CreateRequestOptions, DEFAULT_BACKOFF, DEFAULT_BASE_URL, DEFAULT_FAILOVER_CODES, DEFAULT_RETRY_CONFIG, DEFAULT_TIMEOUT_MS, type DefineToolOptions, DeleteRequestOptions, type DoneEvent, EmbedRequestOptions, EmbedResponse, EmbeddingsRequestOptions, EmbeddingsResponse, type EndpointHealth, type EndpointHealthCheckResult, EndpointRegistry, type EndpointRegistryOptions, type ErrorEvent, type FetchLike, GEN_AI_SYSTEM_OLLAMA, GenerateRequestOptions, GenerateResponse, type GenerateStreamResult, type HttpBody, HttpClient, type HttpClientOptions, type HttpRequestOptions, type LifecycleErrorEvent, type LifecycleRetryEvent, type LifecycleStartEvent, type LifecycleSuccessEvent, type LoadMcpToolsOptions, type LogFn, type Logger, type McpCallToolResult, type McpClientLike, type McpContentBlock, type McpListToolsResult, type McpToolDescriptor, Message, type MessageEvent, type Middleware, type MiddlewareContext, type ModelCapabilities, ModelOptions, ModelResponse, ModelsClient, NANOS_PER_MS, NANOS_PER_SECOND, NOOP_LOGGER, type NextFunction, OllamaAbortError, OllamaAgentMaxIterationsError, OllamaAuthError, OllamaClient, type OllamaClientConfig, OllamaClientError, type OllamaClientErrorOptions, type OllamaEndpoint, type OllamaErrorRequestContext, type OllamaErrorResponseContext, OllamaGenericClientError, OllamaMcpError, OllamaNetworkError, OllamaNotFoundError, OllamaRateLimitError, OllamaServerError, OllamaSkillInvalidError, OllamaSkillNotFoundError, OllamaStream, type OllamaStreamEvent, type OllamaStreamEventType, OllamaTimeoutError, OllamaToolTimeoutError, OllamaToolValidationError, OllamaUnsupportedCapabilityError, type OpenAIChatCompletionChoice, type OpenAIChatCompletionRequest, type OpenAIChatCompletionResponse, OpenAICompatClient, type OpenAIFunctionDefinition, type OpenAIListModelsResponse, type OpenAIMessage, type OpenAIModelItem, type OpenAIResponsesOutputMessage, type OpenAIResponsesOutputTextContent, type OpenAIResponsesRequest, type OpenAIResponsesResponse, type OpenAIStreamOptions, type OpenAITool, type OpenAIToolCall, ProgressResponse, type ProgressStreamResult, PsResponse, PullRequestOptions, PushRequestOptions, type RawUsageSource, type RequestContext, type RequestLifecycleEvent, type RequestLifecycleHook, type RequestRunner, type ResponseContext, type RetryConfig, type RuntimeMode, ShowRequestOptions, ShowResponse, type SpanAttributeValue, type SpanAttributes, StatusResponse, type SupportedSchema, type ThinkingEvent, type TimeoutSignal, type TokenEvent, type TokenUsage, Tool, ToolCall, type ToolCallEvent, ToolDefinition, ToolExecutionResult, ToolHandler, ToolRegistry, VersionResponse, WebFetchRequestOptions, WebFetchResponse, WebSearchRequestOptions, WebSearchResponse, addSpanEvent, calculateBackoff, checkEndpointHealth, composeMiddleware, createConsoleLogger, createTimeoutSignal, defineTool, detectModelCapabilities, encodeImage, extractUsage, inferRuntimeMode, listAvailableModels, loadMcpTools, mapError, normalizeChatStream, normalizeGenerateStream, normalizeProgressStream, parseNdjsonStream, parseStructuredOutput, registerMcpTools, setSpanError, toDataStream, toResponse, toTextStream, withRetry, withSpan, zodToJsonSchema };

package/dist/index.d.ts

@@ -1,5 +1,5 @@
-import { M as Message, C as ChatResponse, G as GenerateResponse, T as ToolCall, P as ProgressResponse, a as ModelResponse, S as ShowRequestOptions, b as ShowResponse, c as PullRequestOptions, d as PushRequestOptions, e as CreateRequestOptions, D as DeleteRequestOptions, f as StatusResponse, g as CopyRequestOptions, h as PsResponse, V as VersionResponse, i as ChatRequestOptions, j as GenerateRequestOptions, E as EmbedRequestOptions, k as EmbedResponse, l as EmbeddingsRequestOptions, m as EmbeddingsResponse, W as WebSearchRequestOptions, n as WebSearchResponse, o as WebFetchRequestOptions, p as WebFetchResponse, q as ToolHandler, r as Tool, A as AnyTool, s as ToolRegistry, t as ToolExecutionResult, u as ModelOptions, v as ToolDefinition } from './compose-DC11aDJn.js';
-export { w as AppliedSkillResult, x as ApplySkillOptions, F as FormatOption, L as ListResponse, y as ModelDetails, z as ParsedFrontmatter, R as RequestCancellationOptions, B as Role, H as Skill, I as SkillFrontmatter, J as SkillSummary, K as ToolCallFunction, N as ToolExecutionContext, O as ToolExecutionFailure, Q as ToolExecutionSuccess, U as ToolFunctionDefinition, X as ToolProperty, Y as ToolRegistryOptions, Z as WebSearchResult, _ as applySkill, $ as parseFrontmatter } from './compose-DC11aDJn.js';
+import { M as Message, C as ChatResponse, G as GenerateResponse, T as ToolCall, P as ProgressResponse, a as ModelResponse, S as ShowRequestOptions, b as ShowResponse, c as PullRequestOptions, d as PushRequestOptions, e as CreateRequestOptions, D as DeleteRequestOptions, f as StatusResponse, g as CopyRequestOptions, h as PsResponse, V as VersionResponse, i as ChatRequestOptions, j as GenerateRequestOptions, E as EmbedRequestOptions, k as EmbedResponse, l as EmbeddingsRequestOptions, m as EmbeddingsResponse, W as WebSearchRequestOptions, n as WebSearchResponse, o as WebFetchRequestOptions, p as WebFetchResponse, q as ToolHandler, r as Tool, A as AnyTool, s as ToolRegistry, t as ToolExecutionResult, u as ModelOptions, v as ToolDefinition } from './compose-DJjVpDrk.js';
+export { w as AppliedSkillResult, x as ApplySkillOptions, F as FormatOption, L as ListResponse, y as ModelDetails, z as ParsedFrontmatter, R as RequestCancellationOptions, B as Role, H as Skill, I as SkillFrontmatter, J as SkillSummary, K as ToolCallFunction, N as ToolExecutionContext, O as ToolExecutionFailure, Q as ToolExecutionSuccess, U as ToolFunctionDefinition, X as ToolProperty, Y as ToolRegistryOptions, Z as WebSearchResult, _ as applySkill, $ as parseFrontmatter } from './compose-DJjVpDrk.js';
 import { z } from 'zod';
 import * as Otel from '@opentelemetry/api';
 
@@ -202,6 +202,8 @@ interface ChatStreamResult {
     readonly message: Message;
     readonly model: string;
     readonly done: boolean;
+    /** Mirrors the final chunk's `done_reason` (e.g. `"stop"`, `"load"`, `"unload"`). */
+    readonly doneReason?: string | undefined;
     readonly totalDurationMs?: number | undefined;
     readonly usage?: TokenUsage | undefined;
     readonly raw?: ChatResponse | undefined;
@@ -210,6 +212,8 @@ interface GenerateStreamResult {
     readonly response: string;
     readonly model: string;
     readonly done: boolean;
+    /** Mirrors the final chunk's `done_reason` (e.g. `"stop"`, `"load"`, `"unload"`). */
+    readonly doneReason?: string | undefined;
     readonly totalDurationMs?: number | undefined;
     readonly usage?: TokenUsage | undefined;
     readonly raw?: GenerateResponse | undefined;
@@ -568,13 +572,15 @@ declare class ModelsClient {
  *
  * @remarks
  * This is a typed pass-through to Ollama's own OpenAI-compatible `/v1` endpoints — it
- * implements the subset of the OpenAI Chat Completions API that Ollama documents as
- * supported (https://docs.ollama.com/api/openai-compatibility), not the full OpenAI
- * surface (Responses API, hosted tools like web/file search, computer use, code
- * interpreter, or server-side conversation state). Notably, `tools` is supported by
- * Ollama's compat layer but `tool_choice` and `parallel_tool_calls` are explicitly
- * **not** — deliberately omitted from {@link OpenAIChatCompletionRequest} rather than
- * typed as accepted-but-ignored.
+ * implements the subset of the OpenAI API surface that Ollama documents as supported
+ * (https://docs.ollama.com/api/openai-compatibility), not the full OpenAI surface (hosted
+ * tools like web/file search, computer use, code interpreter, or stateful server-side
+ * conversation state). Notably, `tools` is supported by Ollama's compat layer but
+ * `tool_choice` and `parallel_tool_calls` are not — both are still typed on
+ * {@link OpenAIChatCompletionRequest} (accepted-but-ignored) so consumers passing a
+ * standard OpenAI request object don't get spurious type errors; see the `@remarks` on
+ * each field. `/v1/responses` is supported non-statefully — see
+ * {@link OpenAIResponsesRequest}.
  */
 
 interface OpenAIToolCall {
@@ -621,6 +627,34 @@ interface OpenAIChatCompletionRequest {
     readonly frequency_penalty?: number | undefined;
     readonly user?: string | undefined;
     readonly tools?: readonly OpenAITool[] | undefined;
+    /**
+     * @remarks Accepted for OpenAI compatibility but ignored by Ollama — every tool the
+     * model is given remains callable regardless of this value.
+     */
+    readonly tool_choice?: 'none' | 'auto' | 'required' | {
+        readonly type: 'function';
+        readonly function: {
+            readonly name: string;
+        };
+    } | undefined;
+    /**
+     * @remarks Accepted for OpenAI compatibility but ignored by Ollama — Ollama does not
+     * control tool-call parallelism through this flag.
+     */
+    readonly parallel_tool_calls?: boolean | undefined;
+    /**
+     * Effort level for thinking models (e.g. `deepseek-r1`, `qwen3`). Equivalent to
+     * `reasoning.effort`; only effective for models that support reasoning/thinking.
+     */
+    readonly reasoning_effort?: 'high' | 'medium' | 'low' | 'max' | 'none' | undefined;
+    /**
+     * Effort level for thinking models (e.g. `deepseek-r1`, `qwen3`), nested OpenAI-style.
+     * Equivalent to `reasoning_effort`; only effective for models that support
+     * reasoning/thinking.
+     */
+    readonly reasoning?: {
+        readonly effort?: 'high' | 'medium' | 'low' | 'max' | 'none' | undefined;
+    } | undefined;
 }
 interface OpenAIChatCompletionChoice {
     readonly index: number;
@@ -649,12 +683,60 @@ interface OpenAIListModelsResponse {
     readonly object: 'list';
     readonly data: readonly OpenAIModelItem[];
 }
+/**
+ * Request body for OpenAI's Responses API (`/v1/responses`), added in Ollama v0.13.3.
+ *
+ * @remarks
+ * Ollama implements this **non-statefully**: every call is independent, so
+ * {@link previous_response_id} and {@link conversation} — OpenAI's mechanisms for
+ * resuming server-side conversation state — are accepted for compatibility but ignored.
+ * Send the full conversation in `input` on every call, the same as `/api/chat`.
+ */
+interface OpenAIResponsesRequest {
+    readonly model: string;
+    readonly input: string | readonly OpenAIMessage[];
+    readonly instructions?: string | undefined;
+    readonly tools?: readonly OpenAITool[] | undefined;
+    readonly stream?: boolean | undefined;
+    readonly temperature?: number | undefined;
+    readonly top_p?: number | undefined;
+    readonly max_output_tokens?: number | undefined;
+    /** @remarks Not supported by Ollama, which is stateless across calls — accepted but ignored. */
+    readonly previous_response_id?: string | undefined;
+    /** @remarks Not supported by Ollama, which is stateless across calls — accepted but ignored. */
+    readonly conversation?: string | undefined;
+    readonly truncation?: string | undefined;
+}
+interface OpenAIResponsesOutputTextContent {
+    readonly type: 'output_text';
+    readonly text: string;
+}
+interface OpenAIResponsesOutputMessage {
+    readonly type: 'message';
+    readonly role?: string | undefined;
+    readonly content: readonly OpenAIResponsesOutputTextContent[];
+}
+/** Non-stateful response shape for `/v1/responses`; see {@link OpenAIResponsesRequest}. */
+interface OpenAIResponsesResponse {
+    readonly id: string;
+    readonly object: 'response';
+    readonly created: number;
+    readonly model: string;
+    readonly output: readonly OpenAIResponsesOutputMessage[];
+    readonly usage?: {
+        readonly input_tokens: number;
+        readonly output_tokens: number;
+        readonly total_tokens: number;
+    } | undefined;
+}
 declare class OpenAICompatClient {
     private readonly http;
     constructor(http: HttpClient);
     createChatCompletion(request: OpenAIChatCompletionRequest, signal?: AbortSignal): Promise<OpenAIChatCompletionResponse>;
     chatCompletions(request: OpenAIChatCompletionRequest, signal?: AbortSignal): Promise<OpenAIChatCompletionResponse>;
     listModels(signal?: AbortSignal): Promise<OpenAIListModelsResponse>;
+    createResponses(request: OpenAIResponsesRequest, signal?: AbortSignal): Promise<OpenAIResponsesResponse>;
+    responses(request: OpenAIResponsesRequest, signal?: AbortSignal): Promise<OpenAIResponsesResponse>;
 }
 
 /**
@@ -1041,4 +1123,19 @@ declare const ATTR_OLLAMA_AGENT_MAX_ITERATIONS = "ollama.agent.max_iterations";
 declare const ATTR_OLLAMA_AGENT_ITERATION = "ollama.agent.iteration";
 declare const GEN_AI_SYSTEM_OLLAMA = "ollama";
 
-export { ATTR_GEN_AI_OPERATION_NAME, ATTR_GEN_AI_REQUEST_MODEL, ATTR_GEN_AI_RESPONSE_MODEL, ATTR_GEN_AI_SYSTEM, ATTR_GEN_AI_TOOL_CALL_ID, ATTR_GEN_AI_TOOL_NAME, ATTR_GEN_AI_USAGE_INPUT_TOKENS, ATTR_GEN_AI_USAGE_OUTPUT_TOKENS, ATTR_HTTP_REQUEST_METHOD, ATTR_HTTP_RESPONSE_STATUS_CODE, ATTR_OLLAMA_AGENT_ITERATION, ATTR_OLLAMA_AGENT_MAX_ITERATIONS, ATTR_OLLAMA_ENDPOINT_ATTEMPT, ATTR_OLLAMA_ENDPOINT_NAME, ATTR_SERVER_ADDRESS, ATTR_SERVER_PORT, ATTR_URL_FULL, type AbortableAsyncIterable, Agent, type AgentChatClient, type AgentConfig, type AgentHooks, type AgentResult, type AgentRunInput, type AgentTurn, type AnthropicCacheControl, AnthropicCompatClient, type AnthropicContentBlock, type AnthropicMessage, type AnthropicMessagesRequest, type AnthropicMessagesResponse, AnyTool, type BackoffOptions, ChatRequestOptions, ChatResponse, type ChatStreamResult, CopyRequestOptions, CreateRequestOptions, DEFAULT_BACKOFF, DEFAULT_BASE_URL, DEFAULT_FAILOVER_CODES, DEFAULT_RETRY_CONFIG, DEFAULT_TIMEOUT_MS, type DefineToolOptions, DeleteRequestOptions, type DoneEvent, EmbedRequestOptions, EmbedResponse, EmbeddingsRequestOptions, EmbeddingsResponse, type EndpointHealth, type EndpointHealthCheckResult, EndpointRegistry, type EndpointRegistryOptions, type ErrorEvent, type FetchLike, GEN_AI_SYSTEM_OLLAMA, GenerateRequestOptions, GenerateResponse, type GenerateStreamResult, type HttpBody, HttpClient, type HttpClientOptions, type HttpRequestOptions, type LifecycleErrorEvent, type LifecycleRetryEvent, type LifecycleStartEvent, type LifecycleSuccessEvent, type LoadMcpToolsOptions, type LogFn, type Logger, type McpCallToolResult, type McpClientLike, type McpContentBlock, type McpListToolsResult, type McpToolDescriptor, Message, type MessageEvent, type Middleware, type MiddlewareContext, type ModelCapabilities, ModelOptions, ModelResponse, ModelsClient, NANOS_PER_MS, NANOS_PER_SECOND, NOOP_LOGGER, type NextFunction, OllamaAbortError, OllamaAgentMaxIterationsError, OllamaAuthError, OllamaClient, type OllamaClientConfig, OllamaClientError, type OllamaClientErrorOptions, type OllamaEndpoint, type OllamaErrorRequestContext, type OllamaErrorResponseContext, OllamaGenericClientError, OllamaMcpError, OllamaNetworkError, OllamaNotFoundError, OllamaRateLimitError, OllamaServerError, OllamaSkillInvalidError, OllamaSkillNotFoundError, OllamaStream, type OllamaStreamEvent, type OllamaStreamEventType, OllamaTimeoutError, OllamaToolTimeoutError, OllamaToolValidationError, OllamaUnsupportedCapabilityError, type OpenAIChatCompletionChoice, type OpenAIChatCompletionRequest, type OpenAIChatCompletionResponse, OpenAICompatClient, type OpenAIFunctionDefinition, type OpenAIListModelsResponse, type OpenAIMessage, type OpenAIModelItem, type OpenAIStreamOptions, type OpenAITool, type OpenAIToolCall, ProgressResponse, type ProgressStreamResult, PsResponse, PullRequestOptions, PushRequestOptions, type RawUsageSource, type RequestContext, type RequestLifecycleEvent, type RequestLifecycleHook, type RequestRunner, type ResponseContext, type RetryConfig, type RuntimeMode, ShowRequestOptions, ShowResponse, type SpanAttributeValue, type SpanAttributes, StatusResponse, type SupportedSchema, type ThinkingEvent, type TimeoutSignal, type TokenEvent, type TokenUsage, Tool, ToolCall, type ToolCallEvent, ToolDefinition, ToolExecutionResult, ToolHandler, ToolRegistry, VersionResponse, WebFetchRequestOptions, WebFetchResponse, WebSearchRequestOptions, WebSearchResponse, addSpanEvent, calculateBackoff, checkEndpointHealth, composeMiddleware, createConsoleLogger, createTimeoutSignal, defineTool, detectModelCapabilities, extractUsage, inferRuntimeMode, listAvailableModels, loadMcpTools, mapError, normalizeChatStream, normalizeGenerateStream, normalizeProgressStream, parseNdjsonStream, parseStructuredOutput, registerMcpTools, setSpanError, toDataStream, toResponse, toTextStream, withRetry, withSpan, zodToJsonSchema };
+/**
+ * Cross-runtime helpers with no Node.js API dependency, so they stay safe to import from
+ * the core client bundle (see ADR 0006 — Edge Runtime CI).
+ */
+/**
+ * Encodes an image to a base64 string, matching the wire format Ollama's `/api/chat` and
+ * `/api/generate` `images` arrays expect.
+ *
+ * - A `string` is assumed to already be base64-encoded and is returned unchanged.
+ * - A `Uint8Array` is base64-encoded using whichever universal primitive the runtime
+ *   exposes: `Buffer` on Node.js, `btoa` on browsers/Edge runtimes. Chunked so it doesn't
+ *   blow the engine's max call-stack/argument-count limits on large images.
+ */
+declare function encodeImage(image: Uint8Array | string): Promise<string>;
+
+export { ATTR_GEN_AI_OPERATION_NAME, ATTR_GEN_AI_REQUEST_MODEL, ATTR_GEN_AI_RESPONSE_MODEL, ATTR_GEN_AI_SYSTEM, ATTR_GEN_AI_TOOL_CALL_ID, ATTR_GEN_AI_TOOL_NAME, ATTR_GEN_AI_USAGE_INPUT_TOKENS, ATTR_GEN_AI_USAGE_OUTPUT_TOKENS, ATTR_HTTP_REQUEST_METHOD, ATTR_HTTP_RESPONSE_STATUS_CODE, ATTR_OLLAMA_AGENT_ITERATION, ATTR_OLLAMA_AGENT_MAX_ITERATIONS, ATTR_OLLAMA_ENDPOINT_ATTEMPT, ATTR_OLLAMA_ENDPOINT_NAME, ATTR_SERVER_ADDRESS, ATTR_SERVER_PORT, ATTR_URL_FULL, type AbortableAsyncIterable, Agent, type AgentChatClient, type AgentConfig, type AgentHooks, type AgentResult, type AgentRunInput, type AgentTurn, type AnthropicCacheControl, AnthropicCompatClient, type AnthropicContentBlock, type AnthropicMessage, type AnthropicMessagesRequest, type AnthropicMessagesResponse, AnyTool, type BackoffOptions, ChatRequestOptions, ChatResponse, type ChatStreamResult, CopyRequestOptions, CreateRequestOptions, DEFAULT_BACKOFF, DEFAULT_BASE_URL, DEFAULT_FAILOVER_CODES, DEFAULT_RETRY_CONFIG, DEFAULT_TIMEOUT_MS, type DefineToolOptions, DeleteRequestOptions, type DoneEvent, EmbedRequestOptions, EmbedResponse, EmbeddingsRequestOptions, EmbeddingsResponse, type EndpointHealth, type EndpointHealthCheckResult, EndpointRegistry, type EndpointRegistryOptions, type ErrorEvent, type FetchLike, GEN_AI_SYSTEM_OLLAMA, GenerateRequestOptions, GenerateResponse, type GenerateStreamResult, type HttpBody, HttpClient, type HttpClientOptions, type HttpRequestOptions, type LifecycleErrorEvent, type LifecycleRetryEvent, type LifecycleStartEvent, type LifecycleSuccessEvent, type LoadMcpToolsOptions, type LogFn, type Logger, type McpCallToolResult, type McpClientLike, type McpContentBlock, type McpListToolsResult, type McpToolDescriptor, Message, type MessageEvent, type Middleware, type MiddlewareContext, type ModelCapabilities, ModelOptions, ModelResponse, ModelsClient, NANOS_PER_MS, NANOS_PER_SECOND, NOOP_LOGGER, type NextFunction, OllamaAbortError, OllamaAgentMaxIterationsError, OllamaAuthError, OllamaClient, type OllamaClientConfig, OllamaClientError, type OllamaClientErrorOptions, type OllamaEndpoint, type OllamaErrorRequestContext, type OllamaErrorResponseContext, OllamaGenericClientError, OllamaMcpError, OllamaNetworkError, OllamaNotFoundError, OllamaRateLimitError, OllamaServerError, OllamaSkillInvalidError, OllamaSkillNotFoundError, OllamaStream, type OllamaStreamEvent, type OllamaStreamEventType, OllamaTimeoutError, OllamaToolTimeoutError, OllamaToolValidationError, OllamaUnsupportedCapabilityError, type OpenAIChatCompletionChoice, type OpenAIChatCompletionRequest, type OpenAIChatCompletionResponse, OpenAICompatClient, type OpenAIFunctionDefinition, type OpenAIListModelsResponse, type OpenAIMessage, type OpenAIModelItem, type OpenAIResponsesOutputMessage, type OpenAIResponsesOutputTextContent, type OpenAIResponsesRequest, type OpenAIResponsesResponse, type OpenAIStreamOptions, type OpenAITool, type OpenAIToolCall, ProgressResponse, type ProgressStreamResult, PsResponse, PullRequestOptions, PushRequestOptions, type RawUsageSource, type RequestContext, type RequestLifecycleEvent, type RequestLifecycleHook, type RequestRunner, type ResponseContext, type RetryConfig, type RuntimeMode, ShowRequestOptions, ShowResponse, type SpanAttributeValue, type SpanAttributes, StatusResponse, type SupportedSchema, type ThinkingEvent, type TimeoutSignal, type TokenEvent, type TokenUsage, Tool, ToolCall, type ToolCallEvent, ToolDefinition, ToolExecutionResult, ToolHandler, ToolRegistry, VersionResponse, WebFetchRequestOptions, WebFetchResponse, WebSearchRequestOptions, WebSearchResponse, addSpanEvent, calculateBackoff, checkEndpointHealth, composeMiddleware, createConsoleLogger, createTimeoutSignal, defineTool, detectModelCapabilities, encodeImage, extractUsage, inferRuntimeMode, listAvailableModels, loadMcpTools, mapError, normalizeChatStream, normalizeGenerateStream, normalizeProgressStream, parseNdjsonStream, parseStructuredOutput, registerMcpTools, setSpanError, toDataStream, toResponse, toTextStream, withRetry, withSpan, zodToJsonSchema };

package/dist/index.js

@@ -820,6 +820,7 @@ function aggregateChat(accumulated, chunk) {
     },
     model: chunk.model,
     done: chunk.done,
+    doneReason: chunk.done ? chunk.done_reason : accumulated.doneReason,
     totalDurationMs: chunk.done && chunk.total_duration !== void 0 ? chunk.total_duration / NANOS_PER_MS : accumulated.totalDurationMs,
     usage: chunk.done ? extractUsage(chunk) : accumulated.usage,
     raw: chunk
@@ -869,6 +870,7 @@ function aggregateGenerate(accumulated, chunk) {
     response: accumulated.response + (chunk.response ?? ""),
     model: chunk.model,
     done: chunk.done,
+    doneReason: chunk.done ? chunk.done_reason : accumulated.doneReason,
     totalDurationMs: chunk.done && chunk.total_duration !== void 0 ? chunk.total_duration / NANOS_PER_MS : accumulated.totalDurationMs,
     usage: chunk.done ? extractUsage(chunk) : accumulated.usage,
     raw: chunk
@@ -1157,6 +1159,16 @@ var OpenAICompatClient = class {
       signal
     });
   }
+  async createResponses(request, signal) {
+    return this.http.request({
+      path: "/v1/responses",
+      body: request,
+      signal
+    });
+  }
+  async responses(request, signal) {
+    return this.createResponses(request, signal);
+  }
 };
 
 // src/integrations/anthropic.ts
@@ -1177,6 +1189,32 @@ var AnthropicCompatClient = class {
   }
 };
 
+// src/utils.ts
+async function encodeImage(image) {
+  if (typeof image === "string") return image;
+  const bufferCtor = globalThis.Buffer;
+  if (bufferCtor !== void 0) return bufferCtor.from(image).toString("base64");
+  const CHUNK_SIZE = 32768;
+  let binary = "";
+  for (let i = 0; i < image.length; i += CHUNK_SIZE) {
+    binary += String.fromCharCode(...image.subarray(i, i + CHUNK_SIZE));
+  }
+  return btoa(binary);
+}
+async function encodeImages(images) {
+  if (images === void 0) return void 0;
+  return Promise.all(images.map((image) => encodeImage(image)));
+}
+async function withEncodedImages(req) {
+  if (!req.images?.some((image) => image instanceof Uint8Array)) return req;
+  return { ...req, images: await encodeImages(req.images) };
+}
+async function withEncodedMessageImages(messages) {
+  if (!messages.some((message) => message.images?.some((image) => image instanceof Uint8Array)))
+    return messages;
+  return Promise.all(messages.map((message) => withEncodedImages(message)));
+}
+
 // src/client.ts
 var OllamaClient = class {
   registry;
@@ -1262,12 +1300,13 @@ var OllamaClient = class {
     }
   }
   async chat(req) {
+    const messages = await withEncodedMessageImages(req.messages);
     if (req.stream) {
       return this.executeWithFailover(async (http, signal) => {
         if (req.format !== void 0) this.assertStructuredOutputSupported(http.baseUrl, req.model);
         const stream = await http.requestStream({
           path: "/api/chat",
-          body: { ...req, stream: true },
+          body: { ...req, messages, stream: true },
           signal
         });
         return normalizeChatStream(stream);
@@ -1286,7 +1325,7 @@ var OllamaClient = class {
             this.assertStructuredOutputSupported(http.baseUrl, req.model);
           return http.request({
             path: "/api/chat",
-            body: { ...req, stream: false },
+            body: { ...req, messages, stream: false },
             signal
           });
         }, req);
@@ -1313,34 +1352,36 @@ var OllamaClient = class {
     return parseStructuredOutput(res.message.content, schema);
   }
   async generate(req) {
-    if (req.stream) {
+    const encodedReq = await withEncodedImages(req);
+    if (encodedReq.stream) {
       return this.executeWithFailover(async (http, signal) => {
-        if (req.format !== void 0) this.assertStructuredOutputSupported(http.baseUrl, req.model);
+        if (encodedReq.format !== void 0)
+          this.assertStructuredOutputSupported(http.baseUrl, encodedReq.model);
         const stream = await http.requestStream({
           path: "/api/generate",
-          body: { ...req, stream: true },
+          body: { ...encodedReq, stream: true },
           signal
         });
         return normalizeGenerateStream(stream);
-      }, req);
+      }, encodedReq);
     }
     return withSpan(
-      `text_completion ${req.model}`,
+      `text_completion ${encodedReq.model}`,
       {
         [ATTR_GEN_AI_SYSTEM]: GEN_AI_SYSTEM_OLLAMA,
         [ATTR_GEN_AI_OPERATION_NAME]: "text_completion",
-        [ATTR_GEN_AI_REQUEST_MODEL]: req.model
+        [ATTR_GEN_AI_REQUEST_MODEL]: encodedReq.model
       },
       async (span) => {
         const res = await this.executeWithFailover((http, signal) => {
-          if (req.format !== void 0)
-            this.assertStructuredOutputSupported(http.baseUrl, req.model);
+          if (encodedReq.format !== void 0)
+            this.assertStructuredOutputSupported(http.baseUrl, encodedReq.model);
           return http.request({
             path: "/api/generate",
-            body: { ...req, stream: false },
+            body: { ...encodedReq, stream: false },
             signal
           });
-        }, req);
+        }, encodedReq);
         span?.setAttributes({
           [ATTR_GEN_AI_RESPONSE_MODEL]: res.model,
           ...res.prompt_eval_count !== void 0 ? { [ATTR_GEN_AI_USAGE_INPUT_TOKENS]: res.prompt_eval_count } : {},
@@ -1964,6 +2005,6 @@ var Agent = class {
   }
 };
 
-export { ATTR_GEN_AI_OPERATION_NAME, ATTR_GEN_AI_REQUEST_MODEL, ATTR_GEN_AI_RESPONSE_MODEL, ATTR_GEN_AI_SYSTEM, ATTR_GEN_AI_TOOL_CALL_ID, ATTR_GEN_AI_TOOL_NAME, ATTR_GEN_AI_USAGE_INPUT_TOKENS, ATTR_GEN_AI_USAGE_OUTPUT_TOKENS, ATTR_HTTP_REQUEST_METHOD, ATTR_HTTP_RESPONSE_STATUS_CODE, ATTR_OLLAMA_AGENT_ITERATION, ATTR_OLLAMA_AGENT_MAX_ITERATIONS, ATTR_OLLAMA_ENDPOINT_ATTEMPT, ATTR_OLLAMA_ENDPOINT_NAME, ATTR_SERVER_ADDRESS, ATTR_SERVER_PORT, ATTR_URL_FULL, Agent, AnthropicCompatClient, DEFAULT_BACKOFF, DEFAULT_BASE_URL, DEFAULT_FAILOVER_CODES, DEFAULT_RETRY_CONFIG, DEFAULT_TIMEOUT_MS, EndpointRegistry, GEN_AI_SYSTEM_OLLAMA, HttpClient, ModelsClient, NANOS_PER_MS, NANOS_PER_SECOND, NOOP_LOGGER, OllamaAbortError, OllamaAgentMaxIterationsError, OllamaAuthError, OllamaClient, OllamaClientError, OllamaGenericClientError, OllamaMcpError, OllamaNetworkError, OllamaNotFoundError, OllamaRateLimitError, OllamaServerError, OllamaSkillInvalidError, OllamaSkillNotFoundError, OllamaStream, OllamaTimeoutError, OllamaToolTimeoutError, OllamaToolValidationError, OllamaUnsupportedCapabilityError, OpenAICompatClient, ToolRegistry, addSpanEvent, applySkill, calculateBackoff, checkEndpointHealth, composeMiddleware, createConsoleLogger, createTimeoutSignal, defineTool, detectModelCapabilities, extractUsage, inferRuntimeMode, listAvailableModels, loadMcpTools, mapError, normalizeChatStream, normalizeGenerateStream, normalizeProgressStream, parseFrontmatter, parseNdjsonStream, parseStructuredOutput, registerMcpTools, setSpanError, toDataStream, toResponse, toTextStream, withRetry, withSpan, zodToJsonSchema };
+export { ATTR_GEN_AI_OPERATION_NAME, ATTR_GEN_AI_REQUEST_MODEL, ATTR_GEN_AI_RESPONSE_MODEL, ATTR_GEN_AI_SYSTEM, ATTR_GEN_AI_TOOL_CALL_ID, ATTR_GEN_AI_TOOL_NAME, ATTR_GEN_AI_USAGE_INPUT_TOKENS, ATTR_GEN_AI_USAGE_OUTPUT_TOKENS, ATTR_HTTP_REQUEST_METHOD, ATTR_HTTP_RESPONSE_STATUS_CODE, ATTR_OLLAMA_AGENT_ITERATION, ATTR_OLLAMA_AGENT_MAX_ITERATIONS, ATTR_OLLAMA_ENDPOINT_ATTEMPT, ATTR_OLLAMA_ENDPOINT_NAME, ATTR_SERVER_ADDRESS, ATTR_SERVER_PORT, ATTR_URL_FULL, Agent, AnthropicCompatClient, DEFAULT_BACKOFF, DEFAULT_BASE_URL, DEFAULT_FAILOVER_CODES, DEFAULT_RETRY_CONFIG, DEFAULT_TIMEOUT_MS, EndpointRegistry, GEN_AI_SYSTEM_OLLAMA, HttpClient, ModelsClient, NANOS_PER_MS, NANOS_PER_SECOND, NOOP_LOGGER, OllamaAbortError, OllamaAgentMaxIterationsError, OllamaAuthError, OllamaClient, OllamaClientError, OllamaGenericClientError, OllamaMcpError, OllamaNetworkError, OllamaNotFoundError, OllamaRateLimitError, OllamaServerError, OllamaSkillInvalidError, OllamaSkillNotFoundError, OllamaStream, OllamaTimeoutError, OllamaToolTimeoutError, OllamaToolValidationError, OllamaUnsupportedCapabilityError, OpenAICompatClient, ToolRegistry, addSpanEvent, applySkill, calculateBackoff, checkEndpointHealth, composeMiddleware, createConsoleLogger, createTimeoutSignal, defineTool, detectModelCapabilities, encodeImage, extractUsage, inferRuntimeMode, listAvailableModels, loadMcpTools, mapError, normalizeChatStream, normalizeGenerateStream, normalizeProgressStream, parseFrontmatter, parseNdjsonStream, parseStructuredOutput, registerMcpTools, setSpanError, toDataStream, toResponse, toTextStream, withRetry, withSpan, zodToJsonSchema };
 //# sourceMappingURL=index.js.map
 //# sourceMappingURL=index.js.map