/**
 * Provider-agnostic LLM types used by the inspector.
 *
 * These replace the langchain BaseMessage / StreamEvent types that the
 * inspector previously depended on. The goal is to keep the inspector free
 * of any `@langchain/*` or langchain runtime imports so that consumers of
 * `@mcp-use/inspector` (and, transitively, of mcp-use) do not need to install
 * langchain to bundle their apps.
 */
/** LLM providers supported by the native agent runtime. */
export type ProviderName = "openai" | "openai-compatible" | "anthropic" | "google" | "openrouter" | "ollama";
/** Provider-neutral model and HTTP request configuration. */
export interface ProviderConfig {
    /** Provider implementation used for requests. */
    provider: ProviderName;
    /** Provider-specific model identifier. */
    model: string;
    /** Provider API key. Ollama may use an empty string. */
    apiKey: string;
    /** Sampling temperature forwarded to the provider. */
    temperature?: number;
    /** Maximum number of output tokens requested from the provider. */
    maxTokens?: number;
    /** Provider API base URL override. */
    baseUrl?: string;
    /** Extra HTTP headers to merge into every request (e.g. OpenRouter's HTTP-Referer). */
    extraHeaders?: Record<string, string>;
    /** Fetch credentials (e.g. `include` for session-cookie auth against a proxy). */
    credentials?: RequestCredentials;
    /** Responses API reasoning effort (direct OpenAI provider). */
    reasoningEffort?: "none" | "low" | "medium" | "high";
}
/** Image content supplied to a multimodal model. */
export interface ImageContentPart {
    /** Content discriminator. */
    type: "image";
    /** Full data URL (`data:image/png;base64,...`) or raw https URL. */
    url: string;
    /** Extracted mime type; filled in by `messageFormat` when converting. */
    mimeType?: string;
    /** Base64 payload without the data-URL prefix (when available). */
    data?: string;
}
/** Text content supplied to a model. */
export interface TextContentPart {
    /** Content discriminator. */
    type: "text";
    /** Plain-text content. */
    text: string;
}
/** Text or image content in a provider message. */
export type ContentPart = TextContentPart | ImageContentPart;
/**
 * Provider-neutral message shape. Each provider module is responsible for
 * mapping this into its wire format (OpenAI chat.completions, Anthropic
 * messages, Gemini generateContent).
 */
export interface ProviderMessage {
    /** Message role in the provider-neutral conversation. */
    role: "system" | "user" | "assistant" | "tool";
    /** Text content, OR a list of rich content parts (for multimodal input). */
    content: string | ContentPart[];
    /** Tool calls emitted by the assistant (assistant messages only). */
    toolCalls?: ProviderToolCall[];
    /** Tool result payload (tool messages only). */
    toolCallId?: string;
    /** Tool name associated with a tool result. */
    toolName?: string;
    /** Raw result returned by the MCP tool. */
    toolResult?: unknown;
    /** Whether the MCP tool reported an error. */
    toolIsError?: boolean;
}
/** Tool call requested by a model. */
export interface ProviderToolCall {
    /** Provider-generated call identifier. */
    id: string;
    /** Tool name exposed to the model. */
    name: string;
    /** Parsed tool arguments. */
    args: Record<string, unknown>;
}
/** Provider-neutral tool definition supplied to a model. */
export interface ProviderTool {
    /** Tool name exposed to the model. */
    name: string;
    /** Human-readable tool description. */
    description?: string;
    /** JSON Schema object describing the tool's input. */
    inputSchema: Record<string, unknown>;
}
/** Token counts reported by an LLM provider. */
export interface TokenUsage {
    /** Tokens read from the request. */
    inputTokens?: number;
    /** Tokens generated in the response. */
    outputTokens?: number;
    /** Total input and output tokens. */
    totalTokens?: number;
    /** Input tokens served from a provider cache. */
    cachedInputTokens?: number;
    /** Input tokens written to a provider cache. Billed, and on Anthropic above base rate. */
    cacheCreationInputTokens?: number;
    /** Tokens used for model reasoning. */
    reasoningTokens?: number;
}
/** Text emitted by the model. */
export interface LlmTextDeltaEvent {
    /** Event discriminator. */
    type: "text-delta";
    /** Incremental text. */
    delta: string;
}
/** Start of a model-requested tool call. */
export interface LlmToolCallStartEvent {
    /** Event discriminator. */
    type: "tool-call-start";
    /** Stable per-turn index for tracking parallel calls. */
    index: number;
    /** Provider-generated tool call identifier. */
    toolCallId: string;
    /** Tool name exposed to the model. */
    toolName: string;
}
/** Incremental JSON arguments for a model-requested tool call. */
export interface LlmToolCallArgsDeltaEvent {
    /** Event discriminator. */
    type: "tool-call-args-delta";
    /** Stable per-turn index for tracking parallel calls. */
    index: number;
    /** Provider-generated tool call identifier. */
    toolCallId: string;
    /** Tool name exposed to the model. */
    toolName: string;
    /** Partial JSON fragment. Concatenate all fragments before parsing. */
    argsDelta: string;
}
/** A tool call with complete, parsed arguments. */
export interface LlmToolCallReadyEvent {
    /** Event discriminator. */
    type: "tool-call-ready";
    /** Stable per-turn index for tracking parallel calls. */
    index: number;
    /** Provider-generated tool call identifier. */
    toolCallId: string;
    /** Tool name exposed to the model. */
    toolName: string;
    /** Parsed tool arguments. */
    args: Record<string, unknown>;
}
/** Result of an MCP tool invocation. */
export interface LlmToolResultEvent {
    /** Event discriminator. */
    type: "tool-result";
    /** Provider-generated tool call identifier. */
    toolCallId: string;
    /** Invoked tool name. */
    toolName: string;
    /** Raw MCP tool result. */
    result: unknown;
    /** Whether the tool reported an error. */
    isError: boolean;
}
/** Token usage reported by the provider. */
export interface LlmUsageEvent {
    /** Event discriminator. */
    type: "usage";
    /** Provider-normalized token counts. */
    usage: TokenUsage;
}
/** Error emitted during streaming. */
export interface LlmErrorEvent {
    /** Event discriminator. */
    type: "error";
    /** Human-readable error message. */
    message: string;
}
/** Indicates that the tool loop finished. */
export interface LlmDoneEvent {
    /** Event discriminator. */
    type: "done";
}
/** Provider-neutral event emitted by the native agent tool loop. */
export type LlmStreamEvent = LlmTextDeltaEvent | LlmToolCallStartEvent | LlmToolCallArgsDeltaEvent | LlmToolCallReadyEvent | LlmToolResultEvent | LlmUsageEvent | LlmErrorEvent | LlmDoneEvent;
//# sourceMappingURL=types.d.ts.map