@xberg-io/liter-llm 1.9.2 → 1.10.0

package/index.d.ts

@@ -1,5 +1,5 @@
 // This file is auto-generated by alef — DO NOT EDIT.
-// alef:hash:990a8effc943d1d0f9c7ef763981d4cd9b10eefb9c9b4f11e1222eac8fca8442
+// alef:hash:d954bd0ab4851664242c5d5bf76bcc6f4c84e06fe6cd42e3effba8a0bb5c3dcf
 // To regenerate: alef generate
 // To verify freshness: alef verify --exit-code
 /* eslint-disable */
@@ -18,7 +18,7 @@ export declare function allProviders(): Array<ProviderConfig>;
 /**
  * Return the capability flags for a named provider.
  *
- * Performs an O(n) linear scan over the embedded registry (143 entries).
+ * Performs an O(n) linear scan over the embedded registry (163 entries).
  * Returns an owned value so bindings can pass capability data without
  * borrowing registry internals.
  *
@@ -43,6 +43,16 @@ export declare function checkBound(context: string, currentLen: number, incoming
  */
 export declare function clear(): void;
 
+/**
+ * Clear the overlay registry, reverting `completion_cost`,
+ * `completion_cost_with_cache`, and `model_info` to the
+ * embedded catalog.
+ *
+ * Primarily a test seam (see [`install_catalog_overlay_from_str`]); also
+ * usable by long-running processes that want to abandon a runtime refresh.
+ */
+export declare function clearCatalogOverlay(): void;
+
 /**
  * Calculate the estimated cost of a completion given a model name and token
  * counts.
@@ -69,13 +79,14 @@ export declare function completionCost(model: string, promptTokens: number, comp
  *
  * Returns `None` if the model is not present in the embedded pricing
  * registry, mirroring [`completion_cost`].
+ *
+ * When the model has [`ModelPricing::tiers`], the tier whose
+ * `min_context_tokens` is the highest value `<= prompt_tokens` supplies the
+ * input/output/cache rates for the whole call; models without tiers (or
+ * when `prompt_tokens` is below every tier threshold) use the base rates
+ * unchanged, matching the original flat-rate behaviour.
  */
-export declare function completionCostWithCache(
-  model: string,
-  promptTokens: number,
-  cachedTokens: number,
-  completionTokens: number,
-): number | null;
+export declare function completionCostWithCache(model: string, promptTokens: number, cachedTokens: number, completionTokens: number): number | null;
 
 /**
  * Return the set of complex provider names.
@@ -119,13 +130,7 @@ export declare function countTokens(model: string, text: string): number;
  * @throws Returns [`LiterLlmError`] if the underlying HTTP client cannot be
  * constructed, or if the resolved provider configuration is invalid.
  */
-export declare function createClient(
-  apiKey: string,
-  baseUrl?: string | undefined | null,
-  timeoutSecs?: number | undefined | null,
-  maxRetries?: number | undefined | null,
-  modelHint?: string | undefined | null,
-): DefaultClient;
+export declare function createClient(apiKey: string, baseUrl?: string | undefined | null, timeoutSecs?: number | undefined | null, maxRetries?: number | undefined | null, modelHint?: string | undefined | null): DefaultClient;
 
 /**
  * Create a new LLM client from a JSON string.
@@ -176,6 +181,23 @@ export declare function encodeDataUrl(bytes: Uint8Array, mime?: string | undefin
  */
 export declare function ensureCryptoProvider(): void;
 
+/**
+ * Install the overlay registry from a raw catalog JSON string, bypassing
+ * the network and disk cache entirely.
+ *
+ * Parses and flattens `catalog_json` with the same
+ * [`registry_from_catalog_str`] logic used for the embedded catalog and the
+ * network refresh path, then atomically swaps it in as the active overlay.
+ * A parse failure returns [`CatalogRefreshError::Parse`] and leaves any
+ * existing overlay untouched.
+ *
+ * This is primarily a testable seam: it lets tests exercise overlay
+ * installation and the embedded/overlay fallback behavior in
+ * `completion_cost` / `model_info` without a real network
+ * call.
+ */
+export declare function installCatalogOverlayFromStr(catalogJson: string): void;
+
 /**
  * Content shape for assistant messages.
  *
@@ -198,15 +220,15 @@ export interface AssistantMessage {
    *
    * `None` is valid when the model replies with tool calls only.
    */
-  readonly content?: AssistantContent;
+  readonly content?: AssistantContent
   /** Optional name for the assistant. */
-  readonly name?: string;
+  readonly name?: string
   /** Tool calls the model wants to execute, if any. */
-  readonly toolCalls?: Array<ToolCall>;
+  readonly toolCalls?: Array<ToolCall>
   /** Refusal reason, if the model declined to respond per safety policies. */
-  readonly refusal?: string;
+  readonly refusal?: string
   /** Deprecated legacy function_call field; retained for API compatibility. */
-  readonly functionCall?: FunctionCall;
+  readonly functionCall?: FunctionCall
 }
 
 /**
@@ -216,28 +238,28 @@ export interface AssistantMessage {
  * parts-spec discriminator (`"type": "text"`, `"type": "output_image"`, …).
  */
 export type AssistantPart =
-  | { type: "text"; text: string }
-  | { type: "refusal"; refusal: string }
-  | { type: "output_image"; imageUrl: ImageUrl }
-  | { type: "output_audio"; audio: AudioContent };
+  | { type: 'text'; text: string }
+  | { type: 'refusal'; refusal: string }
+  | { type: 'output_image'; imageUrl: ImageUrl }
+  | { type: 'output_audio'; audio: AudioContent }
 
 /** Audio content part for speech-capable models. */
 export interface AudioContent {
   /** Base64-encoded audio data. */
-  readonly data?: string;
+  readonly data?: string
   /** Audio format (e.g., "wav", "mp3", "ogg"). */
-  readonly format?: string;
+  readonly format?: string
 }
 
 /** Auth configuration block. */
 export interface AuthConfig {
   /** Auth scheme classification. */
-  readonly authType: AuthType;
+  readonly authType: AuthType
   /**
    * Name of the environment variable that holds the API key (e.g. `"OPENAI_API_KEY"`).
    * Holds the variable name, never the secret value.
    */
-  readonly envVar?: string;
+  readonly envVar?: string
 }
 
 /** How the API key is sent in the HTTP request. */
@@ -265,65 +287,65 @@ export declare enum AuthType {
 /** Query parameters for listing batches. */
 export interface BatchListQuery {
   /** Maximum number of results to return. Defaults to 20. */
-  readonly limit?: number;
+  readonly limit?: number
   /** Pagination cursor: return results after this batch ID. */
-  readonly after?: string;
+  readonly after?: string
 }
 
 /** Response from listing batches. */
 export interface BatchListResponse {
   /** Object type (always `"list"`). */
-  readonly object?: string;
+  readonly object?: string
   /** List of batch objects. */
-  readonly data?: Array<BatchObject>;
+  readonly data?: Array<BatchObject>
   /** Whether more results are available. */
-  readonly hasMore?: boolean;
+  readonly hasMore?: boolean
   /** First batch ID in the result set (for pagination). */
-  readonly firstId?: string;
+  readonly firstId?: string
   /** Last batch ID in the result set (for pagination). */
-  readonly lastId?: string;
+  readonly lastId?: string
 }
 
 /** A batch job object. */
 export interface BatchObject {
   /** Unique batch ID. */
-  readonly id?: string;
+  readonly id?: string
   /** Object type (always `"batch"`). */
-  readonly object?: string;
+  readonly object?: string
   /** API endpoint (e.g., `"/v1/chat/completions"`). */
-  readonly endpoint?: string;
+  readonly endpoint?: string
   /** ID of the input file. */
-  readonly inputFileId?: string;
+  readonly inputFileId?: string
   /** Completion window (e.g., `"24h"`). */
-  readonly completionWindow?: string;
+  readonly completionWindow?: string
   /** Current job status. */
-  readonly status?: BatchStatus;
+  readonly status?: BatchStatus
   /** ID of the output file (present when completed). */
-  readonly outputFileId?: string;
+  readonly outputFileId?: string
   /** ID of the error file (present if some requests failed). */
-  readonly errorFileId?: string;
+  readonly errorFileId?: string
   /** Unix timestamp of batch creation. */
-  readonly createdAt?: number;
+  readonly createdAt?: number
   /** Unix timestamp of completion (if completed). */
-  readonly completedAt?: number;
+  readonly completedAt?: number
   /** Unix timestamp of failure (if failed). */
-  readonly failedAt?: number;
+  readonly failedAt?: number
   /** Unix timestamp of expiration (if expired). */
-  readonly expiredAt?: number;
+  readonly expiredAt?: number
   /** Request processing counts. */
-  readonly requestCounts?: BatchRequestCounts;
+  readonly requestCounts?: BatchRequestCounts
   /** Metadata attached to the batch. */
-  readonly metadata?: JsonValue;
+  readonly metadata?: JsonValue
 }
 
 /** Request processing counts for a batch. */
 export interface BatchRequestCounts {
   /** Total requests in the batch. */
-  readonly total?: number;
+  readonly total?: number
   /** Completed requests. */
-  readonly completed?: number;
+  readonly completed?: number
   /** Failed requests. */
-  readonly failed?: number;
+  readonly failed?: number
 }
 
 /** Status of a batch job. */
@@ -349,152 +371,185 @@ export declare enum BatchStatus {
 /** Configuration for budget enforcement. */
 export interface BudgetConfig {
   /** Maximum total spend across all models, in USD.  `None` means unlimited. */
-  readonly globalLimit?: number;
+  readonly globalLimit?: number
   /**
    * Per-model spending limits in USD.  Models not listed here are only
    * constrained by `global_limit`.
    */
-  readonly modelLimits?: Record<string, number>;
+  readonly modelLimits?: Record<string, number>
   /** Whether to reject requests or merely warn when a limit is exceeded. */
-  readonly enforcement?: Enforcement;
+  readonly enforcement?: Enforcement
 }
 
 /** Storage backend for the response cache. */
-export type CacheBackend = { type: "memory" } | { type: "open_dal"; scheme: string; config: Record<string, string> };
+export type CacheBackend =
+  | { type: 'memory' }
+  | { type: 'open_dal'; scheme: string; config: Record<string, string> }
 
 /** Configuration for the response cache. */
 export interface CacheConfig {
   /** Maximum number of cached entries. */
-  readonly maxEntries?: number;
+  readonly maxEntries?: number
   /** Time-to-live for each cached entry. */
-  readonly ttl?: number;
+  readonly ttl?: number
   /** Storage backend to use. */
-  readonly backend?: CacheBackend;
+  readonly backend?: CacheBackend
+}
+
+/**
+ * Plain-data configuration for [`refresh_catalog`].
+ *
+ * Deliberately FFI/binding-friendly: no `Duration` or `PathBuf`, just
+ * primitives that translate directly across language boundaries.
+ */
+export interface CatalogRefreshConfig {
+  /**
+   * Runtime catalog refresh is entirely opt-in: when `false`,
+   * [`refresh_catalog`] is a no-op that returns
+   * `Ok(`[`RefreshOutcome::Disabled`]`)` without touching the network,
+   * the filesystem, or the overlay registry.
+   */
+  readonly enabled?: boolean
+  /**
+   * Source URL to fetch `catalog.json` from. Must be `https`. Defaults to
+   * [`DEFAULT_CATALOG_URL`]; configurable so self-hosted mirrors work.
+   */
+  readonly sourceUrl?: string
+  /**
+   * How long a cached `catalog.json` remains valid before a network
+   * refetch is attempted, in seconds.
+   */
+  readonly ttlSeconds?: number
+  /**
+   * Filesystem path for the on-disk cache. `None` uses a default path
+   * under `std::env::temp_dir()`.
+   */
+  readonly cachePath?: string
 }
 
 /** A streamed chunk of a chat completion response. */
 export interface ChatCompletionChunk {
   /** Unique identifier for this stream. */
-  readonly id?: string;
+  readonly id?: string
   /**
    * Always `"chat.completion.chunk"` from OpenAI-compatible APIs.  Stored
    * as a plain `String` so non-standard provider values do not fail parsing.
    */
-  readonly object?: string;
+  readonly object?: string
   /** Unix timestamp of chunk creation. */
-  readonly created?: number;
+  readonly created?: number
   /** Model used to generate the chunk. */
-  readonly model?: string;
+  readonly model?: string
   /** Streaming choices (delta updates). */
-  readonly choices?: Array<StreamChoice>;
+  readonly choices?: Array<StreamChoice>
   /** Token usage (typically only in the final chunk). */
-  readonly usage?: Usage;
+  readonly usage?: Usage
   /** Fingerprint of the system configuration (OpenAI-specific). */
-  readonly systemFingerprint?: string;
+  readonly systemFingerprint?: string
   /** Service tier used (OpenAI-specific). */
-  readonly serviceTier?: string;
+  readonly serviceTier?: string
 }
 
 /** Chat completion request (compatible with OpenAI and similar APIs). */
 export interface ChatCompletionRequest {
   /** Model ID (e.g., `"gpt-4o-mini"`, `"claude-3-5-sonnet"`). */
-  readonly model?: string;
+  readonly model?: string
   /** Conversation history from oldest to newest. */
-  readonly messages?: Array<Message>;
+  readonly messages?: Array<Message>
   /** Sampling temperature in `[0.0, 2.0]`. Higher increases randomness. Defaults to 1.0. */
-  readonly temperature?: number;
+  readonly temperature?: number
   /** Nucleus sampling parameter in `[0.0, 1.0]`. Lower is more focused. */
-  readonly topP?: number;
+  readonly topP?: number
   /** Number of chat completions to generate. Defaults to 1. */
-  readonly n?: number;
+  readonly n?: number
   /**
    * Whether to stream the response.
    *
    * Managed by the client layer — do not set directly.
    */
-  readonly stream?: boolean;
+  readonly stream?: boolean
   /** Stop sequence(s) that halt token generation. */
-  readonly stop?: StopSequence;
+  readonly stop?: StopSequence
   /** Max output tokens. Different from max_completion_tokens in some providers. */
-  readonly maxTokens?: number;
+  readonly maxTokens?: number
   /** Presence penalty in `[-2.0, 2.0]`. Positive discourages repeated topics. */
-  readonly presencePenalty?: number;
+  readonly presencePenalty?: number
   /** Frequency penalty in `[-2.0, 2.0]`. Positive discourages repeated tokens. */
-  readonly frequencyPenalty?: number;
+  readonly frequencyPenalty?: number
   /**
    * Token bias map.  Uses `BTreeMap` (sorted keys) for deterministic
    * serialization order — important when hashing or signing requests.
    */
-  readonly logitBias?: Record<string, number>;
+  readonly logitBias?: Record<string, number>
   /** User identifier for request tracking and abuse detection. */
-  readonly user?: string;
+  readonly user?: string
   /** Tools the model can invoke. */
-  readonly tools?: Array<ChatCompletionTool>;
+  readonly tools?: Array<ChatCompletionTool>
   /** Tool usage mode (auto, required, none, or specific tool). */
-  readonly toolChoice?: ToolChoice;
+  readonly toolChoice?: ToolChoice
   /** Whether the model can call multiple tools in parallel. Defaults to true. */
-  readonly parallelToolCalls?: boolean;
+  readonly parallelToolCalls?: boolean
   /** Output format constraint (text, JSON, JSON schema). */
-  readonly responseFormat?: ResponseFormat;
+  readonly responseFormat?: ResponseFormat
   /** Streaming options (e.g., include_usage). */
-  readonly streamOptions?: StreamOptions;
+  readonly streamOptions?: StreamOptions
   /** Random seed for reproducible outputs. Provider support varies. */
-  readonly seed?: number;
+  readonly seed?: number
   /** Reasoning effort level (low, medium, high) for extended-thinking models. */
-  readonly reasoningEffort?: ReasoningEffort;
+  readonly reasoningEffort?: ReasoningEffort
   /**
    * Output modalities to request from the model.
    *
    * For OpenAI audio models, pass `["text", "audio"]`. Vertex AI / Gemini
    * translates these to `generationConfig.responseModalities` (uppercase).
    */
-  readonly modalities?: Array<Modality>;
+  readonly modalities?: Array<Modality>
   /**
    * Provider-specific extra parameters merged into the request body.
    * Use for guardrails, safety settings, grounding config, etc.
    */
-  readonly extraBody?: JsonValue;
+  readonly extraBody?: JsonValue
 }
 
 /** Chat completion response from the API. */
 export interface ChatCompletionResponse {
   /** Unique identifier for this response. */
-  readonly id?: string;
+  readonly id?: string
   /**
    * Always `"chat.completion"` from OpenAI-compatible APIs.  Stored as a
    * plain `String` so non-standard provider values do not break deserialization.
    */
-  readonly object?: string;
+  readonly object?: string
   /** Unix timestamp of response creation. */
-  readonly created?: number;
+  readonly created?: number
   /** Model used to generate the response. */
-  readonly model?: string;
+  readonly model?: string
   /** List of completion choices. */
-  readonly choices?: Array<Choice>;
+  readonly choices?: Array<Choice>
   /** Token usage statistics. */
-  readonly usage?: Usage;
+  readonly usage?: Usage
   /** Fingerprint of the system configuration (OpenAI-specific). */
-  readonly systemFingerprint?: string;
+  readonly systemFingerprint?: string
   /** Service tier used (OpenAI-specific). */
-  readonly serviceTier?: string;
+  readonly serviceTier?: string
 }
 
 /** A tool the model can invoke (currently, all tools are functions). */
 export interface ChatCompletionTool {
   /** Tool type (always "function" in OpenAI spec). */
-  readonly toolType: ToolType;
+  readonly toolType: ToolType
   /** Function definition with name, description, and JSON schema parameters. */
-  readonly function: FunctionDefinition;
+  readonly function: FunctionDefinition
 }
 
 /** A single completion choice. */
 export interface Choice {
   /** Index of this choice in the choices array. */
-  readonly index?: number;
+  readonly index?: number
   /** The assistant's message response. */
-  readonly message?: AssistantMessage;
+  readonly message?: AssistantMessage
   /** Why the model stopped generating (stop, length, tool_calls, content_filter, etc.). */
-  readonly finishReason?: FinishReason;
+  readonly finishReason?: FinishReason
 }
 
 /**
@@ -515,7 +570,7 @@ export interface ChunkMiddleware {
    * - `Ok(None)` — drop this chunk silently.
    * - `Err(e)` — propagate as a stream error.
    */
-  process(chunk?: ChatCompletionChunk | undefined | null): ChatCompletionChunk | null;
+  process(chunk?: ChatCompletionChunk | undefined | null): ChatCompletionChunk | null
 }
 
 /** Observable state of a circuit breaker. */
@@ -530,111 +585,111 @@ export declare enum CircuitState {
 
 /** A single content part in a user message — text, image, document, or audio. */
 export type ContentPart =
-  | { type: "text"; text: string }
-  | { type: "image_url"; imageUrl: ImageUrl }
-  | { type: "document"; document: DocumentContent }
-  | { type: "input_audio"; inputAudio: AudioContent };
+  | { type: 'text'; text: string }
+  | { type: 'image_url'; imageUrl: ImageUrl }
+  | { type: 'document'; document: DocumentContent }
+  | { type: 'input_audio'; inputAudio: AudioContent }
 
 /** Request to create a batch job. */
 export interface CreateBatchRequest {
   /** ID of the uploaded input file (JSONL format). */
-  readonly inputFileId?: string;
+  readonly inputFileId?: string
   /** API endpoint (e.g., `"/v1/chat/completions"`). */
-  readonly endpoint?: string;
+  readonly endpoint?: string
   /** Completion window (e.g., `"24h"`). */
-  readonly completionWindow?: string;
+  readonly completionWindow?: string
   /** Optional metadata to attach to the batch. */
-  readonly metadata?: JsonValue;
+  readonly metadata?: JsonValue
 }
 
 /** Request to upload a file. */
 export interface CreateFileRequest {
   /** Base64-encoded file data. */
-  readonly file?: string;
+  readonly file?: string
   /** Purpose for the file. */
-  readonly purpose?: FilePurpose;
+  readonly purpose?: FilePurpose
   /** Optional filename to associate with the upload. */
-  readonly filename?: string;
+  readonly filename?: string
 }
 
 /** Request to create images from a text prompt. */
 export interface CreateImageRequest {
   /** Text description of the image to generate. */
-  readonly prompt?: string;
+  readonly prompt?: string
   /** Model ID (e.g., `"dall-e-3"`). Optional; API may use default if unset. */
-  readonly model?: string;
+  readonly model?: string
   /** Number of images to generate. Defaults to 1. */
-  readonly n?: number;
+  readonly n?: number
   /** Image size (e.g., `"1024x1024"`, `"1792x1024"`). */
-  readonly size?: string;
+  readonly size?: string
   /** Image quality: `"standard"` or `"hd"`. */
-  readonly quality?: string;
+  readonly quality?: string
   /** Style: `"natural"` or `"vivid"` (DALL-E 3 only). */
-  readonly style?: string;
+  readonly style?: string
   /** Response format: `"url"` or `"b64_json"`. */
-  readonly responseFormat?: string;
+  readonly responseFormat?: string
   /** User identifier for request tracking. */
-  readonly user?: string;
+  readonly user?: string
 }
 
 /** Request to create a structured response. */
 export interface CreateResponseRequest {
   /** Model ID. */
-  readonly model?: string;
+  readonly model?: string
   /** Input data to process (e.g., a document to extract from). */
-  readonly input?: JsonValue;
+  readonly input?: JsonValue
   /** Instructions for processing the input. */
-  readonly instructions?: string;
+  readonly instructions?: string
   /** Available tools the model can use. */
-  readonly tools?: Array<ResponseTool>;
+  readonly tools?: Array<ResponseTool>
   /** Sampling temperature in `[0.0, 2.0]`. Defaults to 1.0. */
-  readonly temperature?: number;
+  readonly temperature?: number
   /** Maximum output tokens. */
-  readonly maxOutputTokens?: number;
+  readonly maxOutputTokens?: number
   /** Optional metadata. */
-  readonly metadata?: JsonValue;
+  readonly metadata?: JsonValue
 }
 
 /** Request to generate speech audio from text. */
 export interface CreateSpeechRequest {
   /** Model ID (e.g., `"tts-1"`, `"tts-1-hd"`). */
-  readonly model?: string;
+  readonly model?: string
   /** Text to synthesize into speech. */
-  readonly input?: string;
+  readonly input?: string
   /** Voice name (e.g., `"alloy"`, `"echo"`, `"fable"`, `"onyx"`, `"nova"`, `"shimmer"`). */
-  readonly voice?: string;
+  readonly voice?: string
   /** Audio format (e.g., `"mp3"`, `"opus"`, `"aac"`, `"flac"`, `"wav"`, `"pcm"`). */
-  readonly responseFormat?: string;
+  readonly responseFormat?: string
   /** Playback speed in `[0.25, 4.0]`. Defaults to 1.0. */
-  readonly speed?: number;
+  readonly speed?: number
 }
 
 /** Request to transcribe audio into text. */
 export interface CreateTranscriptionRequest {
   /** Model ID (e.g., `"whisper-1"`). */
-  readonly model?: string;
+  readonly model?: string
   /** Base64-encoded audio file data. */
-  readonly file?: string;
+  readonly file?: string
   /** Language ISO-639-1 code (e.g., `"en"`, `"fr"`, `"de"`). Optional; model auto-detects. */
-  readonly language?: string;
+  readonly language?: string
   /** Optional text to guide the model (improves accuracy for domain-specific terms). */
-  readonly prompt?: string;
+  readonly prompt?: string
   /** Output format (e.g., `"json"`, `"text"`, `"vtt"`, `"srt"`, `"verbose_json"`). */
-  readonly responseFormat?: string;
+  readonly responseFormat?: string
   /** Sampling temperature in `[0.0, 1.0]`. Higher increases variability. Defaults to 0. */
-  readonly temperature?: number;
+  readonly temperature?: number
 }
 
 /** Configuration for registering a custom LLM provider at runtime. */
 export interface CustomProviderConfig {
   /** Unique name for this provider (e.g., "my-provider"). */
-  readonly name: string;
+  readonly name: string
   /** Base URL for the provider's API (e.g., `<https://api.my-provider.com/v1>`). */
-  readonly baseUrl: string;
+  readonly baseUrl: string
   /** Authentication header format. */
-  readonly authHeader: AuthHeaderFormat;
+  readonly authHeader: AuthHeaderFormat
   /** Model name prefixes that route to this provider (e.g., `["my-"]`). */
-  readonly modelPrefixes: Array<string>;
+  readonly modelPrefixes: Array<string>
 }
 
 /**
@@ -645,15 +700,15 @@ export interface CustomProviderConfig {
  */
 export interface DecodedDataUrl {
   /** MIME type extracted from the URL prefix (verbatim, not normalised). */
-  readonly mime?: string;
+  readonly mime?: string
   /** Decoded base64 payload. */
-  readonly data?: Uint8Array;
+  readonly data?: Uint8Array
 }
 
 /**
  * Default client implementation backed by `reqwest`.
  *
- * Sends requests to 143 LLM providers with automatic provider detection
+ * Sends requests to 163 LLM providers with automatic provider detection
  * and per-request routing. The provider is resolved at construction time
  * from `model_hint` (or defaults to OpenAI), but individual requests can
  * override the provider via model name prefix (e.g. `"anthropic/claude-3-5-sonnet"`
@@ -668,29 +723,27 @@ export interface DecodedDataUrl {
  * headers are cached at construction to avoid redundant encoding on every request.
  */
 export declare class DefaultClient {
-  chat(req?: ChatCompletionRequest | undefined | null): Promise<ChatCompletionResponse>;
-  chatStream(
-    req?: ChatCompletionRequest | undefined | null,
-  ): Promise<AsyncGenerator<ChatCompletionChunk, void, undefined>>;
-  embed(req?: EmbeddingRequest | undefined | null): Promise<EmbeddingResponse>;
-  listModels(): Promise<ModelsListResponse>;
-  imageGenerate(req?: CreateImageRequest | undefined | null): Promise<ImagesResponse>;
-  speech(req?: CreateSpeechRequest | undefined | null): Promise<Uint8Array>;
-  transcribe(req?: CreateTranscriptionRequest | undefined | null): Promise<TranscriptionResponse>;
-  moderate(req?: ModerationRequest | undefined | null): Promise<ModerationResponse>;
-  rerank(req?: RerankRequest | undefined | null): Promise<RerankResponse>;
-  search(req?: SearchRequest | undefined | null): Promise<SearchResponse>;
-  ocr(req?: OcrRequest | undefined | null): Promise<OcrResponse>;
-  createFile(req?: CreateFileRequest | undefined | null): Promise<FileObject>;
-  retrieveFile(fileId: string): Promise<FileObject>;
-  deleteFile(fileId: string): Promise<DeleteResponse>;
-  listFiles(query?: FileListQuery | undefined | null): Promise<FileListResponse>;
-  fileContent(fileId: string): Promise<Uint8Array>;
-  createBatch(req?: CreateBatchRequest | undefined | null): Promise<BatchObject>;
-  retrieveBatch(batchId: string): Promise<BatchObject>;
-  listBatches(query?: BatchListQuery | undefined | null): Promise<BatchListResponse>;
-  cancelBatch(batchId: string): Promise<BatchObject>;
-  fetchBatchForPolling(batchId: string): Promise<BatchObject>;
+  chat(req?: ChatCompletionRequest | undefined | null): Promise<ChatCompletionResponse>
+  chatStream(req?: ChatCompletionRequest | undefined | null): Promise<AsyncGenerator<ChatCompletionChunk, void, undefined>>
+  embed(req?: EmbeddingRequest | undefined | null): Promise<EmbeddingResponse>
+  listModels(): Promise<ModelsListResponse>
+  imageGenerate(req?: CreateImageRequest | undefined | null): Promise<ImagesResponse>
+  speech(req?: CreateSpeechRequest | undefined | null): Promise<Uint8Array>
+  transcribe(req?: CreateTranscriptionRequest | undefined | null): Promise<TranscriptionResponse>
+  moderate(req?: ModerationRequest | undefined | null): Promise<ModerationResponse>
+  rerank(req?: RerankRequest | undefined | null): Promise<RerankResponse>
+  search(req?: SearchRequest | undefined | null): Promise<SearchResponse>
+  ocr(req?: OcrRequest | undefined | null): Promise<OcrResponse>
+  createFile(req?: CreateFileRequest | undefined | null): Promise<FileObject>
+  retrieveFile(fileId: string): Promise<FileObject>
+  deleteFile(fileId: string): Promise<DeleteResponse>
+  listFiles(query?: FileListQuery | undefined | null): Promise<FileListResponse>
+  fileContent(fileId: string): Promise<Uint8Array>
+  createBatch(req?: CreateBatchRequest | undefined | null): Promise<BatchObject>
+  retrieveBatch(batchId: string): Promise<BatchObject>
+  listBatches(query?: BatchListQuery | undefined | null): Promise<BatchListResponse>
+  cancelBatch(batchId: string): Promise<BatchObject>
+  fetchBatchForPolling(batchId: string): Promise<BatchObject>
   /**
    * Poll a batch until it reaches a terminal status (Completed, Failed, Expired, Cancelled).
    *
@@ -700,36 +753,36 @@ export declare class DefaultClient {
    * Returns `BatchWaitError::Timeout` if the configured timeout is exceeded.
    * Returns `BatchWaitError::Client` for underlying client errors.
    */
-  waitForBatch(batchId: string, config?: WaitForBatchConfig | undefined | null): Promise<BatchObject>;
-  createResponse(req?: CreateResponseRequest | undefined | null): Promise<ResponseObject>;
-  retrieveResponse(responseId: string): Promise<ResponseObject>;
-  cancelResponse(responseId: string): Promise<ResponseObject>;
+  waitForBatch(batchId: string, config?: WaitForBatchConfig | undefined | null): Promise<BatchObject>
+  createResponse(req?: CreateResponseRequest | undefined | null): Promise<ResponseObject>
+  retrieveResponse(responseId: string): Promise<ResponseObject>
+  cancelResponse(responseId: string): Promise<ResponseObject>
 }
 
 /** Response from a delete operation. */
 export interface DeleteResponse {
   /** ID of the deleted resource. */
-  readonly id?: string;
+  readonly id?: string
   /** Object type. */
-  readonly object?: string;
+  readonly object?: string
   /** Confirmation that the resource was deleted. */
-  readonly deleted?: boolean;
+  readonly deleted?: boolean
 }
 
 /** Developer message (system-like message for Claude models). */
 export interface DeveloperMessage {
   /** Developer-specific instructions or context. */
-  readonly content?: string;
+  readonly content?: string
   /** Optional name for the developer message source. */
-  readonly name?: string;
+  readonly name?: string
 }
 
 /** PDF/document content part for vision-capable models. */
 export interface DocumentContent {
   /** Base64-encoded document data or URL. */
-  readonly data?: string;
+  readonly data?: string
   /** MIME type (e.g., "application/pdf", "text/csv"). */
-  readonly mediaType?: string;
+  readonly mediaType?: string
 }
 
 /** The format in which the embedding vectors are returned. */
@@ -754,25 +807,25 @@ export interface EmbeddingObject {
    * Always `"embedding"` from OpenAI-compatible APIs.  Stored as a plain
    * `String` so non-standard provider values do not break deserialization.
    */
-  readonly object: string;
+  readonly object: string
   /** The embedding vector. */
-  readonly embedding: Array<number>;
+  readonly embedding: Array<number>
   /** Index in the batch (corresponds to input order). */
-  readonly index: number;
+  readonly index: number
 }
 
 /** Embedding request. */
 export interface EmbeddingRequest {
   /** Model ID (e.g., `"text-embedding-3-small"`). */
-  readonly model?: string;
+  readonly model?: string
   /** Text or texts to embed. */
-  readonly input?: EmbeddingInput;
+  readonly input?: EmbeddingInput
   /** Output format: float (native) or base64. */
-  readonly encodingFormat?: EmbeddingFormat;
+  readonly encodingFormat?: EmbeddingFormat
   /** Requested embedding dimensions (if supported by the model). */
-  readonly dimensions?: number;
+  readonly dimensions?: number
   /** User identifier for request tracking. */
-  readonly user?: string;
+  readonly user?: string
 }
 
 /** Embedding response. */
@@ -781,13 +834,13 @@ export interface EmbeddingResponse {
    * Always `"list"` from OpenAI-compatible APIs.  Stored as a plain
    * `String` so non-standard provider values do not break deserialization.
    */
-  readonly object: string;
+  readonly object: string
   /** List of embeddings. */
-  readonly data: Array<EmbeddingObject>;
+  readonly data: Array<EmbeddingObject>
   /** Model used to generate embeddings. */
-  readonly model: string;
+  readonly model: string
   /** Token usage (input tokens only; embeddings have zero output tokens). */
-  readonly usage?: Usage;
+  readonly usage?: Usage
 }
 
 /** How budget limits are enforced. */
@@ -807,39 +860,39 @@ export declare enum Enforcement {
 /** Query parameters for listing files. */
 export interface FileListQuery {
   /** Filter by file purpose (e.g., `"batch"`, `"fine-tune"`). */
-  readonly purpose?: string;
+  readonly purpose?: string
   /** Maximum number of results to return. Defaults to 20. */
-  readonly limit?: number;
+  readonly limit?: number
   /** Pagination cursor: return results after this file ID. */
-  readonly after?: string;
+  readonly after?: string
 }
 
 /** Response from listing files. */
 export interface FileListResponse {
   /** Object type (always `"list"`). */
-  readonly object?: string;
+  readonly object?: string
   /** List of file objects. */
-  readonly data?: Array<FileObject>;
+  readonly data?: Array<FileObject>
   /** Whether more results are available. */
-  readonly hasMore?: boolean;
+  readonly hasMore?: boolean
 }
 
 /** An uploaded file object. */
 export interface FileObject {
   /** Unique file ID. */
-  readonly id?: string;
+  readonly id?: string
   /** Object type (always `"file"`). */
-  readonly object?: string;
+  readonly object?: string
   /** File size in bytes. */
-  readonly bytes?: number;
+  readonly bytes?: number
   /** Unix timestamp of file creation. */
-  readonly createdAt?: number;
+  readonly createdAt?: number
   /** Filename. */
-  readonly filename?: string;
+  readonly filename?: string
   /** File purpose. */
-  readonly purpose?: string;
+  readonly purpose?: string
   /** Processing status (e.g., `"uploaded"`, `"processed"`). */
-  readonly status?: string;
+  readonly status?: string
 }
 
 /** Purpose of an uploaded file. */
@@ -877,27 +930,27 @@ export declare enum FinishReason {
 /** Function call details. */
 export interface FunctionCall {
   /** Function name. */
-  readonly name: string;
+  readonly name: string
   /** Arguments as a JSON string (parse with serde_json::from_str). */
-  readonly arguments: string;
+  readonly arguments: string
 }
 
 /** Function definition exposed to the model. */
 export interface FunctionDefinition {
   /** Name of the function. Required and must be alphanumeric + underscores. */
-  readonly name: string;
+  readonly name: string
   /** Human-readable description explaining what the function does. */
-  readonly description?: string;
+  readonly description?: string
   /** JSON Schema defining the function's parameters. */
-  readonly parameters?: JsonValue;
+  readonly parameters?: JsonValue
   /** If true, enforce strict JSON schema validation for arguments. */
-  readonly strict?: boolean;
+  readonly strict?: boolean
 }
 
 /** Deprecated legacy function-role message body. */
 export interface FunctionMessage {
-  readonly content?: string;
-  readonly name?: string;
+  readonly content?: string
+  readonly name?: string
 }
 
 /**
@@ -914,7 +967,7 @@ export interface HealthChecker {
    * move it into the returned future without a clone, making the
    * `'static + Send` bound on the future trivially satisfiable.
    */
-  check(upstream: string): Promise<HealthStatus>;
+  check(upstream: string): Promise<HealthStatus>
 }
 
 /** The result of a single health probe. */
@@ -928,11 +981,11 @@ export declare enum HealthStatus {
 /** A single generated image, returned as either a URL or base64 data. */
 export interface Image {
   /** Image URL (if response_format was "url"). */
-  readonly url?: string;
+  readonly url?: string
   /** Base64-encoded image data (if response_format was "b64_json"). */
-  readonly b64Json?: string;
+  readonly b64Json?: string
   /** The final prompt used to generate the image (DALL-E 3). */
-  readonly revisedPrompt?: string;
+  readonly revisedPrompt?: string
 }
 
 /** Image detail level controlling token cost and processing. */
@@ -948,49 +1001,49 @@ export declare enum ImageDetail {
 /** Response containing generated images. */
 export interface ImagesResponse {
   /** Unix timestamp of image creation. */
-  readonly created?: number;
+  readonly created?: number
   /** List of generated images. */
-  readonly data?: Array<Image>;
+  readonly data?: Array<Image>
 }
 
 /** An image URL reference with optional detail level for processing. */
 export interface ImageUrl {
   /** URL of the image (data URI or HTTP/HTTPS URL). */
-  readonly url?: string;
+  readonly url?: string
   /** Detail level: low (512x512), high (2x2 tiles), or auto (model-selected). */
-  readonly detail?: ImageDetail;
+  readonly detail?: ImageDetail
 }
 
 /** An intent prototype: `(intent_name, prototype_embedding, target_model_id)`. */
 export interface IntentPrototype {
   /** Human-readable name for the intent (used in logs/metrics). */
-  readonly name: string;
+  readonly name: string
   /** Pre-computed embedding vector for this intent. */
-  readonly embedding: Array<number>;
+  readonly embedding: Array<number>
   /** Model to route to when this intent is detected. */
-  readonly model: string;
+  readonly model: string
 }
 
 /** JSON Schema specification for constrained output. */
 export interface JsonSchemaFormat {
   /** Name of the schema (must be unique in the request). */
-  readonly name?: string;
+  readonly name?: string
   /** Description of what the schema represents. */
-  readonly description?: string;
+  readonly description?: string
   /** JSON Schema object defining the output structure. */
-  readonly schema?: JsonValue;
+  readonly schema?: JsonValue
   /** If true, enforce strict schema validation. */
-  readonly strict?: boolean;
+  readonly strict?: boolean
 }
 
 /** A chat message in a conversation. */
 export type Message =
-  | { role: "system"; 0: SystemMessage }
-  | { role: "user"; 0: UserMessage }
-  | { role: "assistant"; 0: AssistantMessage }
-  | { role: "tool"; 0: ToolMessage }
-  | { role: "developer"; 0: DeveloperMessage }
-  | { role: "function"; 0: FunctionMessage };
+  | { role: 'system'; 0: SystemMessage }
+  | { role: 'user'; 0: UserMessage }
+  | { role: 'assistant'; 0: AssistantMessage }
+  | { role: 'tool'; 0: ToolMessage }
+  | { role: 'developer'; 0: DeveloperMessage }
+  | { role: 'function'; 0: FunctionMessage }
 
 /**
  * Output modality requested from the model.
@@ -1007,19 +1060,80 @@ export declare enum Modality {
   Image = "image",
 }
 
+/**
+ * Public, FFI-friendly snapshot of a model's pricing and capability
+ * metadata, projected from [`ModelPricing`].
+ *
+ * Unlike [`ModelPricing`] (which is excluded from binding generation),
+ * `ModelInfo` is an owned plain-data DTO safe to hand across the FFI
+ * boundary — see [`model_info`].
+ */
+export interface ModelInfo {
+  /** Cost in USD per input (prompt) token. */
+  readonly inputCostPerToken?: number
+  /** Cost in USD per output (completion) token. */
+  readonly outputCostPerToken?: number
+  /** Cost in USD per cached input token (cache hit / read). */
+  readonly cacheReadInputTokenCost?: number
+  /** Cost in USD per token written to the prompt cache. */
+  readonly cacheCreationInputTokenCost?: number
+  /** Cost in USD per input audio token. */
+  readonly inputCostPerAudioToken?: number
+  /** Cost in USD per output audio token. */
+  readonly outputCostPerAudioToken?: number
+  /** Cost in USD per reasoning (extended-thinking) output token. */
+  readonly outputCostPerReasoningToken?: number
+  /** Total context window size in tokens (input + output). */
+  readonly maxTokens?: number
+  /** Maximum input (prompt) tokens accepted. */
+  readonly maxInputTokens?: number
+  /** Maximum output (completion) tokens the model can generate. */
+  readonly maxOutputTokens?: number
+  /** Best-effort operating mode, e.g. `"chat"`, `"embedding"`. */
+  readonly mode?: string
+  /** The model accepts image input. */
+  readonly supportsVision?: boolean
+  /** The model supports tool / function calling. */
+  readonly supportsFunctionCalling?: boolean
+  /** The model supports extended-thinking / reasoning tokens. */
+  readonly supportsReasoning?: boolean
+  /** The model supports JSON-mode or `response_format` structured output. */
+  readonly supportsStructuredOutput?: boolean
+  /** The model accepts audio input. */
+  readonly supportsAudioInput?: boolean
+  /** The model can generate audio output. */
+  readonly supportsAudioOutput?: boolean
+  /** The model supports prompt caching. */
+  readonly supportsPromptCaching?: boolean
+  /**
+   * Context-tiered pricing overrides, sorted by ascending
+   * `min_context_tokens`. Empty when the model has flat pricing.
+   */
+  readonly tiers?: Array<ModelTier>
+}
+
 /** A model available from the API. */
 export interface ModelObject {
   /** Model ID (e.g., `"gpt-4o"`, `"claude-3-5-sonnet"`). */
-  readonly id?: string;
+  readonly id?: string
   /**
    * Always `"model"` from OpenAI-compatible APIs.  Stored as a plain
    * `String` so non-standard provider values do not break deserialization.
+   * Defaults to empty when a provider omits the field.
    */
-  readonly object?: string;
-  /** Unix timestamp of model creation (or release date). */
-  readonly created?: number;
-  /** Organization or entity that owns the model. */
-  readonly ownedBy?: string;
+  readonly object?: string
+  /**
+   * Unix timestamp of model creation (or release date).
+   *
+   * Defaults to `0` when a provider omits it — DeepSeek and some other
+   * OpenAI-compatible providers do not return `created` from `/v1/models`.
+   */
+  readonly created?: number
+  /**
+   * Organization or entity that owns the model.
+   * Defaults to empty when a provider omits the field.
+   */
+  readonly ownedBy?: string
 }
 
 /** Response listing available models. */
@@ -1028,61 +1142,87 @@ export interface ModelsListResponse {
    * Always `"list"` from OpenAI-compatible APIs.  Stored as a plain
    * `String` so non-standard provider values do not break deserialization.
    */
-  readonly object?: string;
+  readonly object?: string
   /** List of available models. */
-  readonly data?: Array<ModelObject>;
+  readonly data?: Array<ModelObject>
+}
+
+/**
+ * Public, FFI-friendly snapshot of a single context-window pricing tier,
+ * projected from [`PricingTier`].
+ */
+export interface ModelTier {
+  /**
+   * The tier applies when the prompt/context token count is at least
+   * this value.
+   */
+  readonly minContextTokens?: number
+  /** Cost in USD per input (prompt) token within this tier. */
+  readonly inputCostPerToken?: number
+  /** Cost in USD per output (completion) token within this tier. */
+  readonly outputCostPerToken?: number
+  /** Cost in USD per cached input token within this tier. */
+  readonly cacheReadInputTokenCost?: number
+  /** Cost in USD per cache-write token within this tier. */
+  readonly cacheCreationInputTokenCost?: number
+  /** Cost in USD per input audio token within this tier. */
+  readonly inputCostPerAudioToken?: number
+  /** Cost in USD per output audio token within this tier. */
+  readonly outputCostPerAudioToken?: number
+  /** Cost in USD per reasoning output token within this tier. */
+  readonly outputCostPerReasoningToken?: number
 }
 
 /** Boolean flags for each moderation category. */
 export interface ModerationCategories {
   /** Sexual content. */
-  readonly sexual?: boolean;
+  readonly sexual?: boolean
   /** Hate speech. */
-  readonly hate?: boolean;
+  readonly hate?: boolean
   /** Harassment. */
-  readonly harassment?: boolean;
+  readonly harassment?: boolean
   /** Self-harm content. */
-  readonly selfHarm?: boolean;
+  readonly selfHarm?: boolean
   /** Sexual content involving minors. */
-  readonly sexualMinors?: boolean;
+  readonly sexualMinors?: boolean
   /** Hate speech that threatens violence. */
-  readonly hateThreatening?: boolean;
+  readonly hateThreatening?: boolean
   /** Graphic violence. */
-  readonly violenceGraphic?: boolean;
+  readonly violenceGraphic?: boolean
   /** Intent to self-harm. */
-  readonly selfHarmIntent?: boolean;
+  readonly selfHarmIntent?: boolean
   /** Instructions for self-harm. */
-  readonly selfHarmInstructions?: boolean;
+  readonly selfHarmInstructions?: boolean
   /** Harassment that threatens violence. */
-  readonly harassmentThreatening?: boolean;
+  readonly harassmentThreatening?: boolean
   /** Non-graphic violence. */
-  readonly violence?: boolean;
+  readonly violence?: boolean
 }
 
 /** Confidence scores for each moderation category. */
 export interface ModerationCategoryScores {
   /** Sexual content score. */
-  readonly sexual?: number;
+  readonly sexual?: number
   /** Hate speech score. */
-  readonly hate?: number;
+  readonly hate?: number
   /** Harassment score. */
-  readonly harassment?: number;
+  readonly harassment?: number
   /** Self-harm content score. */
-  readonly selfHarm?: number;
+  readonly selfHarm?: number
   /** Sexual content involving minors score. */
-  readonly sexualMinors?: number;
+  readonly sexualMinors?: number
   /** Hate speech that threatens violence score. */
-  readonly hateThreatening?: number;
+  readonly hateThreatening?: number
   /** Graphic violence score. */
-  readonly violenceGraphic?: number;
+  readonly violenceGraphic?: number
   /** Intent to self-harm score. */
-  readonly selfHarmIntent?: number;
+  readonly selfHarmIntent?: number
   /** Instructions for self-harm score. */
-  readonly selfHarmInstructions?: number;
+  readonly selfHarmInstructions?: number
   /** Harassment that threatens violence score. */
-  readonly harassmentThreatening?: number;
+  readonly harassmentThreatening?: number
   /** Non-graphic violence score. */
-  readonly violence?: number;
+  readonly violence?: number
 }
 
 /** Input to the moderation endpoint — a single string or multiple strings. */
@@ -1096,82 +1236,84 @@ export declare enum ModerationInput {
 /** Request to classify content for policy violations. */
 export interface ModerationRequest {
   /** Text or texts to check. */
-  readonly input?: ModerationInput;
+  readonly input?: ModerationInput
   /** Model ID (e.g., `"text-moderation-latest"`). Optional; API uses default if unset. */
-  readonly model?: string;
+  readonly model?: string
 }
 
 /** Response from the moderation endpoint. */
 export interface ModerationResponse {
   /** Unique identifier for this moderation request. */
-  readonly id: string;
+  readonly id: string
   /** Model used for classification. */
-  readonly model: string;
+  readonly model: string
   /** Results for each input string. */
-  readonly results: Array<ModerationResult>;
+  readonly results: Array<ModerationResult>
 }
 
 /** A single moderation classification result. */
 export interface ModerationResult {
   /** True if any category was flagged. */
-  readonly flagged: boolean;
+  readonly flagged: boolean
   /** Boolean flags for each moderation category. */
-  readonly categories: ModerationCategories;
+  readonly categories: ModerationCategories
   /** Confidence scores for each category. */
-  readonly categoryScores: ModerationCategoryScores;
+  readonly categoryScores: ModerationCategoryScores
 }
 
 /** Document input for OCR — either a URL or inline base64 data. */
-export type OcrDocument = { type: "document_url"; url: string } | { type: "base64"; data: string; mediaType: string };
+export type OcrDocument =
+  | { type: 'document_url'; url: string }
+  | { type: 'base64'; data: string; mediaType: string }
 
 /** An image extracted from an OCR page. */
 export interface OcrImage {
   /** Unique image identifier within the document. */
-  readonly id: string;
+  readonly id: string
   /** Base64-encoded image data (if `include_image_base64` was true). */
-  readonly imageBase64?: string;
+  readonly imageBase64?: string
 }
 
 /** A single page of OCR output. */
 export interface OcrPage {
   /** Page index (0-based). */
-  readonly index: number;
+  readonly index: number
   /** Extracted page content as Markdown. */
-  readonly markdown: string;
+  readonly markdown: string
   /** Embedded images extracted from the page (if `include_image_base64` was true). */
-  readonly images?: Array<OcrImage>;
+  readonly images?: Array<OcrImage>
   /** Page dimensions in pixels, if available. */
-  readonly dimensions?: PageDimensions;
+  readonly dimensions?: PageDimensions
 }
 
 /** An OCR request. */
 export interface OcrRequest {
   /** The model/provider to use (e.g. `"mistral/mistral-ocr-latest"`). */
-  readonly model?: string;
+  readonly model?: string
   /** The document to process (URL or base64). */
-  readonly document?: OcrDocument;
+  readonly document?: OcrDocument
   /** Specific pages to process (1-indexed). `None` means all pages. */
-  readonly pages?: Array<number>;
+  readonly pages?: Array<number>
   /** Whether to include base64-encoded images of each processed page. */
-  readonly includeImageBase64?: boolean;
+  readonly includeImageBase64?: boolean
 }
 
 /** An OCR response. */
 export interface OcrResponse {
   /** Extracted pages in order. */
-  readonly pages: Array<OcrPage>;
+  readonly pages: Array<OcrPage>
   /** Model/provider used for OCR. */
-  readonly model: string;
+  readonly model: string
   /** Token usage, if reported by the provider. */
-  readonly usage?: Usage;
+  readonly usage?: Usage
 }
 
 /** Page dimensions in pixels. */
 export interface PageDimensions {
   /** Width in pixels. */
-  readonly width: number;
+  readonly width: number
   /** Height in pixels. */
-  readonly height: number;
+  readonly height: number
 }
 
 /**
@@ -1184,9 +1326,9 @@ export interface PageDimensions {
  */
 export interface PromptTokensDetails {
   /** Cached tokens present in the prompt. Defaults to 0 when absent. */
-  readonly cachedTokens?: number;
+  readonly cachedTokens?: number
   /** Audio input tokens present in the prompt. Defaults to 0 when absent. */
-  readonly audioTokens?: number;
+  readonly audioTokens?: number
 }
 
 /**
@@ -1216,19 +1358,19 @@ export interface PromptTokensDetails {
  */
 export interface ProviderCapabilities {
   /** The provider accepts image input in chat messages. */
-  readonly vision?: boolean;
+  readonly vision?: boolean
   /** The provider supports extended-thinking / reasoning tokens. */
-  readonly reasoning?: boolean;
+  readonly reasoning?: boolean
   /** The provider supports JSON-mode or `response_format` structured output. */
-  readonly structuredOutput?: boolean;
+  readonly structuredOutput?: boolean
   /** The provider supports tool / function calling. */
-  readonly functionCalling?: boolean;
+  readonly functionCalling?: boolean
   /** The provider accepts audio as input. */
-  readonly audioIn?: boolean;
+  readonly audioIn?: boolean
   /** The provider can generate audio / TTS output. */
-  readonly audioOut?: boolean;
+  readonly audioOut?: boolean
   /** The provider accepts video as input. */
-  readonly videoIn?: boolean;
+  readonly videoIn?: boolean
 }
 
 /**
@@ -1239,17 +1381,17 @@ export interface ProviderCapabilities {
  */
 export interface ProviderConfig {
   /** Provider identifier (matches the entry key in providers.json). */
-  readonly name: string;
+  readonly name: string
   /** Human-readable provider name shown in UIs. */
-  readonly displayName?: string;
+  readonly displayName?: string
   /** Base URL used as the default for this provider's HTTP client. */
-  readonly baseUrl?: string;
+  readonly baseUrl?: string
   /** Authentication scheme metadata (auth type + env var holding the key). */
-  readonly auth?: AuthConfig;
+  readonly auth?: AuthConfig
   /** Supported endpoint kinds (e.g. `chat`, `embeddings`). */
-  readonly endpoints?: Array<string>;
+  readonly endpoints?: Array<string>
   /** Model-name prefixes claimed by this provider (e.g. `["gpt-", "o1-"]`). */
-  readonly modelPrefixes?: Array<string>;
+  readonly modelPrefixes?: Array<string>
   /**
    * Parameter key renaming for this provider.
    *
@@ -1257,17 +1399,17 @@ export interface ProviderConfig {
    * to the name this provider expects (e.g. `"max_tokens"`).  Applied
    * automatically by `ConfigDrivenProvider::transform_request`.
    */
-  readonly paramMappings?: Record<string, string>;
+  readonly paramMappings?: Record<string, string>
 }
 
 /** Configuration for per-model rate limits. */
 export interface RateLimitConfig {
   /** Maximum requests per window.  `None` means unlimited. */
-  readonly rpm?: number;
+  readonly rpm?: number
   /** Maximum tokens per window.  `None` means unlimited. */
-  readonly tpm?: number;
+  readonly tpm?: number
   /** Fixed window duration (defaults to 60 s). */
-  readonly window?: number;
+  readonly window?: number
 }
 
 /** Controls how much reasoning effort the model should use. */
@@ -1277,6 +1419,26 @@ export declare enum ReasoningEffort {
   High = "high",
 }
 
+/** Result of a [`refresh_catalog`] call. */
+export declare enum RefreshOutcome {
+  /**
+   * `config.enabled` was `false`; no network, filesystem, or overlay
+   * activity occurred.
+   */
+  Disabled = "Disabled",
+  /**
+   * The on-disk cache was fresh (age < `ttl_seconds`); the overlay was
+   * installed from the cached file without a network request.
+   */
+  FromCache = "FromCache",
+  /**
+   * The catalog was fetched over the network, the cache file was
+   * (best-effort) refreshed, and the overlay was installed from the
+   * fetched catalog.
+   */
+  Fetched = "Fetched",
+}
+
 /** A document to be reranked — either a plain string or an object with a text field. */
 export declare enum RerankDocument {
   /** Plain text document content. */
@@ -1288,41 +1450,41 @@ export declare enum RerankDocument {
 /** Request to rerank documents by relevance to a query. */
 export interface RerankRequest {
   /** Model ID (e.g., `"cohere/rerank-english-v3.0"`). */
-  readonly model?: string;
+  readonly model?: string
   /** The search query. */
-  readonly query?: string;
+  readonly query?: string
   /** Documents to rerank. */
-  readonly documents?: Array<RerankDocument>;
+  readonly documents?: Array<RerankDocument>
   /** Return only the top N results. Optional. */
-  readonly topN?: number;
+  readonly topN?: number
   /** Include the document content in results. Defaults to false. */
-  readonly returnDocuments?: boolean;
+  readonly returnDocuments?: boolean
 }
 
 /** Response from the rerank endpoint. */
 export interface RerankResponse {
   /** Unique identifier for this rerank request. */
-  readonly id?: string;
+  readonly id?: string
   /** Reranked documents in order of relevance. */
-  readonly results: Array<RerankResult>;
+  readonly results: Array<RerankResult>
   /** Optional metadata about the reranking operation. */
-  readonly meta?: JsonValue;
+  readonly meta?: JsonValue
 }
 
 /** A single reranked document with its relevance score. */
 export interface RerankResult {
   /** Original document index in the input list. */
-  readonly index: number;
+  readonly index: number
   /** Relevance score in `[0, 1]`. Higher indicates more relevant. */
-  readonly relevanceScore: number;
+  readonly relevanceScore: number
   /** Original document content (if `return_documents` was true). */
-  readonly document?: RerankResultDocument;
+  readonly document?: RerankResultDocument
 }
 
 /** The text content of a reranked document, returned when `return_documents` is true. */
 export interface RerankResultDocument {
   /** Document text. */
-  readonly text: string;
+  readonly text: string
 }
 
 /**
@@ -1344,88 +1506,88 @@ export interface RerankResultDocument {
  *   returned JSON if the schema is load-bearing.
  */
 export type ResponseFormat =
-  | { type: "text" }
-  | { type: "json_object" }
-  | { type: "json_schema"; jsonSchema: JsonSchemaFormat };
+  | { type: 'text' }
+  | { type: 'json_object' }
+  | { type: 'json_schema'; jsonSchema: JsonSchemaFormat }
 
 /** Response from a structured response request. */
 export interface ResponseObject {
   /** Unique response ID. */
-  readonly id?: string;
+  readonly id?: string
   /** Object type (e.g., `"response"`). */
-  readonly object?: string;
+  readonly object?: string
   /** Unix timestamp of response creation. */
-  readonly createdAt?: number;
+  readonly createdAt?: number
   /** Model used to generate the response. */
-  readonly model?: string;
+  readonly model?: string
   /** Status (e.g., `"succeeded"`, `"failed"`). */
-  readonly status?: string;
+  readonly status?: string
   /** Output items from the response. */
-  readonly output?: Array<ResponseOutputItem>;
+  readonly output?: Array<ResponseOutputItem>
   /** Token usage. */
-  readonly usage?: ResponseUsage;
+  readonly usage?: ResponseUsage
   /** Error details (if status is "failed"). */
-  readonly error?: JsonValue;
+  readonly error?: JsonValue
 }
 
 /** A single output item from the response. */
 export interface ResponseOutputItem {
   /** Output type (e.g., `"text"`, `"object"`, `"error"`). */
-  readonly itemType?: string;
+  readonly itemType?: string
   /** Output content (flattened into the object). */
-  readonly content?: JsonValue;
+  readonly content?: JsonValue
 }
 
 /** A tool available for the response request. */
 export interface ResponseTool {
   /** Tool type (e.g., "extractor", "search"). */
-  readonly toolType?: string;
+  readonly toolType?: string
   /** Tool configuration (flattened into the object). */
-  readonly config?: JsonValue;
+  readonly config?: JsonValue
 }
 
 /** Token usage for a response. */
 export interface ResponseUsage {
   /** Input tokens used. */
-  readonly inputTokens?: number;
+  readonly inputTokens?: number
   /** Output tokens used. */
-  readonly outputTokens?: number;
+  readonly outputTokens?: number
   /** Total tokens used. */
-  readonly totalTokens?: number;
+  readonly totalTokens?: number
 }
 
 /** A search request. */
 export interface SearchRequest {
   /** The model/provider to use (e.g. `"brave/web-search"`, `"tavily/search"`). */
-  readonly model?: string;
+  readonly model?: string
   /** The search query string. */
-  readonly query?: string;
+  readonly query?: string
   /** Maximum number of results to return. */
-  readonly maxResults?: number;
+  readonly maxResults?: number
   /** Domain filter — restrict results to specific domains. */
-  readonly searchDomainFilter?: Array<string>;
+  readonly searchDomainFilter?: Array<string>
   /** Country code for localized results (ISO 3166-1 alpha-2, e.g., `"US"`, `"FR"`). */
-  readonly country?: string;
+  readonly country?: string
 }
 
 /** A search response. */
 export interface SearchResponse {
   /** List of search results. */
-  readonly results: Array<SearchResult>;
+  readonly results: Array<SearchResult>
   /** Model/provider that performed the search. */
-  readonly model: string;
+  readonly model: string
 }
 
 /** An individual search result. */
 export interface SearchResult {
   /** Result title. */
-  readonly title: string;
+  readonly title: string
   /** Result URL. */
-  readonly url: string;
+  readonly url: string
   /** Text snippet or excerpt from the page. */
-  readonly snippet: string;
+  readonly snippet: string
   /** Publication or last-updated date, if available. */
-  readonly date?: string;
+  readonly date?: string
 }
 
 /**
@@ -1434,20 +1596,21 @@ export interface SearchResult {
  * The error value is shared so every follower receives the same upstream
  * failure without cloning the underlying error.
  */
-export declare class SingleflightResult {}
+export declare class SingleflightResult {
+}
 
 /** Name of the specific function to invoke. */
 export interface SpecificFunction {
   /** Function name. */
-  readonly name?: string;
+  readonly name?: string
 }
 
 /** Directive to call a specific tool. */
 export interface SpecificToolChoice {
   /** Tool type (always "function"). */
-  readonly choiceType?: ToolType;
+  readonly choiceType?: ToolType
   /** The specific function to invoke. */
-  readonly function?: SpecificFunction;
+  readonly function?: SpecificFunction
 }
 
 /** Stop sequence(s) that cause the model to stop generating. */
@@ -1461,25 +1624,25 @@ export declare enum StopSequence {
 /** A streaming choice with incremental delta. */
 export interface StreamChoice {
   /** Index of this choice in the choices array. */
-  readonly index?: number;
+  readonly index?: number
   /** Incremental update to the message (content, tool calls, etc.). */
-  readonly delta?: StreamDelta;
+  readonly delta?: StreamDelta
   /** Why the stream ended (present only in final chunk). */
-  readonly finishReason?: FinishReason;
+  readonly finishReason?: FinishReason
 }
 
 /** Incremental delta in a stream chunk. */
 export interface StreamDelta {
   /** Role (typically present only in the first chunk). */
-  readonly role?: string;
+  readonly role?: string
   /** Partial content chunk (e.g., a few words of the response). */
-  readonly content?: string;
+  readonly content?: string
   /** Partial tool calls being streamed. */
-  readonly toolCalls?: Array<StreamToolCall>;
+  readonly toolCalls?: Array<StreamToolCall>
   /** Deprecated legacy function_call delta; retained for API compatibility. */
-  readonly functionCall?: StreamFunctionCall;
+  readonly functionCall?: StreamFunctionCall
   /** Partial refusal message. */
-  readonly refusal?: string;
+  readonly refusal?: string
 }
 
 /**
@@ -1500,27 +1663,27 @@ export declare enum StreamFormat {
 /** Partial function call details in a stream. */
 export interface StreamFunctionCall {
   /** Function name (typically in the first chunk). */
-  readonly name?: string;
+  readonly name?: string
   /** Partial JSON arguments chunk. */
-  readonly arguments?: string;
+  readonly arguments?: string
 }
 
 /** Options for streaming responses. */
 export interface StreamOptions {
   /** If true, include token usage in the final stream chunk. */
-  readonly includeUsage?: boolean;
+  readonly includeUsage?: boolean
 }
 
 /** A streaming tool call being built incrementally. */
 export interface StreamToolCall {
   /** Index of this tool call in the tool_calls array. */
-  readonly index?: number;
+  readonly index?: number
   /** Tool call ID (typically in the first chunk for this call). */
-  readonly id?: string;
+  readonly id?: string
   /** Tool type (typically "function"). */
-  readonly callType?: ToolType;
+  readonly callType?: ToolType
   /** Partial function name and arguments. */
-  readonly function?: StreamFunctionCall;
+  readonly function?: StreamFunctionCall
 }
 
 /** System message guiding model behavior for the entire conversation. */
@@ -1531,19 +1694,19 @@ export interface SystemMessage {
    * Accepts either a plain text string or an array of content parts,
    * mirroring [`UserContent`] so that `Message::system_with_parts` works.
    */
-  readonly content?: UserContent;
+  readonly content?: UserContent
   /** Optional name for the system message source. */
-  readonly name?: string;
+  readonly name?: string
 }
 
 /** A tool call the model wants to execute. */
 export interface ToolCall {
   /** Unique ID for this call, used to reference in tool result messages. */
-  readonly id: string;
+  readonly id: string
   /** Tool type (always "function"). */
-  readonly callType: ToolType;
+  readonly callType: ToolType
   /** Function name and arguments. */
-  readonly function: FunctionCall;
+  readonly function: FunctionCall
 }
 
 /** Tool usage mode or a specific tool to call. */
@@ -1567,11 +1730,11 @@ export declare enum ToolChoiceMode {
 /** Tool execution result returned to the model. */
 export interface ToolMessage {
   /** Result of the tool execution. */
-  readonly content?: string;
+  readonly content?: string
   /** ID of the tool call this result responds to. */
-  readonly toolCallId?: string;
+  readonly toolCallId?: string
   /** Optional tool/function name. */
-  readonly name?: string;
+  readonly name?: string
 }
 
 /**
@@ -1588,41 +1751,41 @@ export declare enum ToolType {
 /** Response from a transcription request. */
 export interface TranscriptionResponse {
   /** The transcribed text. */
-  readonly text?: string;
+  readonly text?: string
   /** Detected language (ISO-639-1 code). */
-  readonly language?: string;
+  readonly language?: string
   /** Total audio duration in seconds. */
-  readonly duration?: number;
+  readonly duration?: number
   /** Detailed segment-level transcription (if response_format is "verbose_json"). */
-  readonly segments?: Array<TranscriptionSegment>;
+  readonly segments?: Array<TranscriptionSegment>
 }
 
 /** A segment of transcribed audio with timing information. */
 export interface TranscriptionSegment {
   /** Segment index (0-based). */
-  readonly id?: number;
+  readonly id?: number
   /** Start time in seconds. */
-  readonly start?: number;
+  readonly start?: number
   /** End time in seconds. */
-  readonly end?: number;
+  readonly end?: number
   /** Transcribed text for this segment. */
-  readonly text?: string;
+  readonly text?: string
 }
 
 /** Token-usage accounting returned by the provider on each completion / embedding call. */
 export interface Usage {
   /** Prompt tokens used. Defaults to 0 when absent (some providers omit this). */
-  readonly promptTokens?: number;
+  readonly promptTokens?: number
   /** Completion tokens used. Defaults to 0 when absent (e.g. embedding responses). */
-  readonly completionTokens?: number;
+  readonly completionTokens?: number
   /** Total tokens used. Defaults to 0 when absent (some providers omit this). */
-  readonly totalTokens?: number;
+  readonly totalTokens?: number
   /**
    * Breakdown of tokens used in the prompt, including cached tokens served
    * at the provider's discounted cache-read rate. Absent when the provider
    * does not return prompt-token details.
    */
-  readonly promptTokensDetails?: PromptTokensDetails;
+  readonly promptTokensDetails?: PromptTokensDetails
 }
 
 /** User message content as either plain text or a list of multimodal parts. */
@@ -1636,9 +1799,9 @@ export declare enum UserContent {
 /** User message in the conversation. */
 export interface UserMessage {
   /** Message content as plain text or array of content parts (text, images, documents, audio). */
-  readonly content?: UserContent;
+  readonly content?: UserContent
   /** Optional name for the user. */
-  readonly name?: string;
+  readonly name?: string
 }
 
 /**
@@ -1649,15 +1812,54 @@ export interface UserMessage {
  */
 export interface WaitForBatchConfig {
   /** Initial interval between polls, in seconds. */
-  readonly initialIntervalSecs?: number;
+  readonly initialIntervalSecs?: number
   /** Maximum interval between polls (backoff plateau), in seconds. */
-  readonly maxIntervalSecs?: number;
+  readonly maxIntervalSecs?: number
   /** Exponential backoff multiplier (e.g., 1.5 increases delay by 50% each poll). */
-  readonly backoffMultiplier?: number;
+  readonly backoffMultiplier?: number
   /** Optional timeout in seconds — polling fails if this duration is exceeded. */
-  readonly timeoutSecs?: number;
+  readonly timeoutSecs?: number
 }
 
+/**
+ * Look up FFI-friendly pricing and capability metadata for a model.
+ *
+ * Returns `None` if the model is not present in the active pricing
+ * registry. Uses the same exact-match-then-prefix-fallback resolution as
+ * [`model_pricing`]; unlike `model_pricing`, the result is an owned
+ * [`ModelInfo`] value safe to hand across the FFI boundary.
+ *
+ * When a runtime catalog refresh has succeeded, this reflects the refreshed
+ * (overlay) catalog; otherwise it reflects the embedded catalog. See
+ * [`model_pricing`] for the embedded-only alternative.
+ */
+export declare function modelInfo(model: string): ModelInfo | null;
+
+/**
+ * Refresh the runtime catalog overlay per `config`.
+ *
+ * - `config.enabled == false`: returns `Ok(`[`RefreshOutcome::Disabled`]`)`
+ *   immediately. No network, filesystem, or overlay activity.
+ * - A fresh on-disk cache (age < `config.ttl_seconds`) exists at the
+ *   resolved cache path (`config.cache_path`, or a default under
+ *   `std::env::temp_dir()`): read + flatten it and install the overlay,
+ *   returning `Ok(`[`RefreshOutcome::FromCache`]`)`. No network request is
+ *   made.
+ * - Otherwise: validate `config.source_url` uses `https`
+ *   ([`CatalogRefreshError::InsecureUrl`] otherwise), fetch it, flatten it,
+ *   install the overlay, best-effort write the raw JSON to the cache path
+ *   (a cache write failure does not fail the refresh), and return
+ *   `Ok(`[`RefreshOutcome::Fetched`]`)`.
+ *
+ * On any error return, the overlay is left untouched: the previously
+ * active registry (a prior successful overlay, or the embedded catalog if
+ * none was ever installed) remains in effect. This is what makes the
+ * feature air-gap-safe — an unreachable or invalid `source_url` never
+ * degrades `completion_cost` / `model_info` below embedded-catalog
+ * availability.
+ */
+export declare function refreshCatalog(config?: CatalogRefreshConfig | undefined | null): Promise<RefreshOutcome>;
+
 /**
  * Register a custom provider in the global runtime registry.
  *
@@ -1678,12 +1880,12 @@ export declare function registerCustomProvider(config: CustomProviderConfig): vo
 export declare function unregisterCustomProvider(name: string): boolean;
 
 export declare class ChatStreamIterator {
-  next(value?: undefined): Promise<IteratorResult<ChatCompletionChunk, void>>;
-  [Symbol.asyncIterator](): AsyncGenerator<ChatCompletionChunk, void, undefined>;
+  next(value?: undefined): Promise<IteratorResult<ChatCompletionChunk, void>>
+  [Symbol.asyncIterator](): AsyncGenerator<ChatCompletionChunk, void, undefined>
 }
 
 export declare class LiterLlmErrorInfo {
-  statusCode(): number;
-  isTransient(): boolean;
-  errorType(): string;
+  statusCode(): number
+  isTransient(): boolean
+  errorType(): string
 }