weave-typescript 0.8.0 → 0.10.0

package/dist/weaveapi/llmx/v1/capabilities.pb.d.ts

@@ -3,18 +3,32 @@ export declare const protobufPackage = "weaveapi.llmx.v1";
 /** Core capability types that models can support */
 export declare enum CapabilityType {
     CAPABILITY_TYPE_UNSPECIFIED = 0,
-    CAPABILITY_TYPE_STRUCTURED_RESPONSE = 1,
-    CAPABILITY_TYPE_STREAMING = 2,
-    CAPABILITY_TYPE_FUNCTION_CALLING = 3,
-    CAPABILITY_TYPE_VISION = 4,
-    CAPABILITY_TYPE_TOOL_USE = 5,
-    CAPABILITY_TYPE_SYSTEM_PROMPT = 6,
-    CAPABILITY_TYPE_CACHING = 7,
-    CAPABILITY_TYPE_REASONING = 8,
-    CAPABILITY_TYPE_AUDIO = 9,
-    CAPABILITY_TYPE_VIDEO = 10,
-    CAPABILITY_TYPE_EMBEDDINGS = 11,
-    CAPABILITY_TYPE_FINE_TUNING = 12,
+    /** CAPABILITY_TYPE_TEXT - Basic text input/output capability */
+    CAPABILITY_TYPE_TEXT = 1,
+    /** CAPABILITY_TYPE_STRUCTURED_RESPONSE - Structured output (JSON, XML, etc.) */
+    CAPABILITY_TYPE_STRUCTURED_RESPONSE = 2,
+    /** CAPABILITY_TYPE_STREAMING - Server-sent or websocket streaming of partial outputs */
+    CAPABILITY_TYPE_STREAMING = 3,
+    /** CAPABILITY_TYPE_FUNCTION_CALLING - Tool/function calling with schemaed arguments */
+    CAPABILITY_TYPE_FUNCTION_CALLING = 4,
+    /** CAPABILITY_TYPE_VISION - Image understanding and/or generation */
+    CAPABILITY_TYPE_VISION = 5,
+    /** CAPABILITY_TYPE_TOOL_USE - General tool use beyond plain function calls */
+    CAPABILITY_TYPE_TOOL_USE = 6,
+    /** CAPABILITY_TYPE_SYSTEM_PROMPT - System messages/prompt steering support */
+    CAPABILITY_TYPE_SYSTEM_PROMPT = 7,
+    /** CAPABILITY_TYPE_CACHING - Prompt/context caching capabilities */
+    CAPABILITY_TYPE_CACHING = 8,
+    /** CAPABILITY_TYPE_REASONING - Advanced reasoning controls/strategies */
+    CAPABILITY_TYPE_REASONING = 9,
+    /** CAPABILITY_TYPE_AUDIO - Audio STT/TTS support */
+    CAPABILITY_TYPE_AUDIO = 10,
+    /** CAPABILITY_TYPE_VIDEO - Video analysis/generation support */
+    CAPABILITY_TYPE_VIDEO = 11,
+    /** CAPABILITY_TYPE_EMBEDDINGS - Vector embedding generation */
+    CAPABILITY_TYPE_EMBEDDINGS = 12,
+    /** CAPABILITY_TYPE_FINE_TUNING - Custom model training/fine-tuning */
+    CAPABILITY_TYPE_FINE_TUNING = 13,
     UNRECOGNIZED = -1
 }
 export declare function capabilityTypeFromJSON(object: any): CapabilityType;
@@ -160,10 +174,35 @@ export declare enum Hyperparameter {
 }
 export declare function hyperparameterFromJSON(object: any): Hyperparameter;
 export declare function hyperparameterToJSON(object: Hyperparameter): string;
-/** Base capability configuration */
+/** Modality direction - whether a modality supports input, output, or both */
+export declare enum ModalityDirection {
+    MODALITY_DIRECTION_UNSPECIFIED = 0,
+    /** MODALITY_DIRECTION_INPUT_ONLY - Can only process/analyze (e.g., image analysis) */
+    MODALITY_DIRECTION_INPUT_ONLY = 1,
+    /** MODALITY_DIRECTION_OUTPUT_ONLY - Can only generate (e.g., TTS without STT) */
+    MODALITY_DIRECTION_OUTPUT_ONLY = 2,
+    /** MODALITY_DIRECTION_INPUT_OUTPUT - Can both process and generate */
+    MODALITY_DIRECTION_INPUT_OUTPUT = 3,
+    UNRECOGNIZED = -1
+}
+export declare function modalityDirectionFromJSON(object: any): ModalityDirection;
+export declare function modalityDirectionToJSON(object: ModalityDirection): string;
+/**
+ * Capability represents a specific feature/ability of a model with its configuration.
+ * Each capability has a type and optional detailed configuration.
+ */
 export interface Capability {
+    /**
+     * The type of capability this represents.
+     * Example: CAPABILITY_TYPE_FUNCTION_CALLING for function/tool calling
+     */
     type: CapabilityType;
+    /**
+     * Whether this capability is currently enabled/available.
+     * Example: true if the model supports and has this feature active
+     */
     enabled: boolean;
+    text?: Text | undefined;
     structuredResponse?: StructuredResponse | undefined;
     streaming?: Streaming | undefined;
     functionCalling?: FunctionCalling | undefined;
@@ -176,117 +215,467 @@ export interface Capability {
     video?: Video | undefined;
     embeddings?: Embeddings | undefined;
     fineTuning?: FineTuning | undefined;
+    /**
+     * Unstructured additional information about this capability.
+     * Used for provider-specific details that don't fit the structured fields.
+     * Example: "Supports up to 10 parallel function calls with automatic retry"
+     * Example: "Beta feature - may have unexpected behavior"
+     * Example: "Optimized for conversational use cases"
+     */
+    additionalInfo: string;
+}
+/** Text capability configuration for basic text input/output */
+export interface Text {
+    /**
+     * Direction of text support (input, output, or both)
+     * Example: MODALITY_DIRECTION_INPUT_OUTPUT for chat models
+     */
+    direction: ModalityDirection;
+    /**
+     * Maximum input text length in characters (if limited)
+     * Example: 32000 for models with character limits
+     */
+    maxInputLength: number;
+    /**
+     * Maximum output text length in characters (if limited)
+     * Example: 4096 for models with output limits
+     */
+    maxOutputLength: number;
+    /**
+     * Supported languages for text processing
+     * Examples: ["en", "es", "fr", "de", "zh", "ja"]
+     */
+    supportedLanguages: string[];
+    /**
+     * Whether the model supports multi-turn conversations
+     * Example: true for chat models, false for completion-only models
+     */
+    supportsConversation: boolean;
+    /**
+     * Whether the model can maintain context across messages
+     * Example: true for stateful models
+     */
+    supportsContext: boolean;
 }
 /** Structured response capability configuration */
 export interface StructuredResponse {
+    /**
+     * Optional guidance to include in the system prompt to elicit structured output
+     * Example: "Always return valid JSON matching the provided schema"
+     */
     systemPromptHint: string;
+    /**
+     * Supported output formats for structured responses
+     * Examples: [DATA_FORMAT_JSON, DATA_FORMAT_YAML]
+     */
     supportedFormats: DataFormat[];
+    /**
+     * Maximum allowed nesting depth for JSON schema objects/arrays
+     * Example: 5 for moderately complex schemas
+     */
     maxSchemaDepth: number;
+    /**
+     * Whether structured output requires tool/function use (vs. direct generation)
+     * Example: true for providers that only return JSON via tool calls
+     */
     requiresToolUse: boolean;
+    /**
+     * Whether the model must be in a special "JSON mode" to honor schemas
+     * Example: true for models that enforce strict JSON output when enabled
+     */
     requiresJsonMode: boolean;
+    /**
+     * Maximum number of top-level properties in object schemas
+     * Example: 50
+     */
     maxProperties: number;
+    /**
+     * JSON Schema primitive/types supported by the model
+     * Examples: [JSON_SCHEMA_TYPE_OBJECT, JSON_SCHEMA_TYPE_ARRAY, JSON_SCHEMA_TYPE_STRING]
+     */
     supportedTypes: JsonSchemaType[];
-    /** Can stream structured responses (not just generate them) */
+    /**
+     * Can stream structured responses (not just generate them)
+     * Example: true if partial JSON chunks are streamed
+     */
     supportsStreaming: boolean;
 }
 /** Streaming capability configuration */
 export interface Streaming {
+    /**
+     * Delimiter used between streamed chunks, if any
+     * Example: "\n\n" for event-stream boundaries
+     */
     chunkDelimiter: string;
+    /**
+     * Preferred buffer size used by the provider when batching output
+     * Example: 1024 (bytes)
+     */
     bufferSize: number;
+    /**
+     * Whether Server-Sent Events (SSE) is supported
+     * Example: true if provider uses text/event-stream
+     */
     supportsSse: boolean;
+    /**
+     * Whether token usage info can be emitted incrementally during streaming
+     * Example: true if usage deltas are included in stream
+     */
     supportsUsage: boolean;
+    /**
+     * Typical size of individual streamed chunks in bytes
+     * Example: 256
+     */
     avgChunkSizeBytes: number;
+    /**
+     * Maximum delay between streamed chunks in milliseconds
+     * Example: 200
+     */
     maxChunkDelayMs: number;
 }
 /** Function calling capability configuration */
 export interface FunctionCalling {
+    /**
+     * Maximum number of functions that can be registered per request
+     * Example: 128
+     */
     maxFunctions: number;
+    /**
+     * Maximum number of tool/function calls that may be executed in parallel
+     * Example: 5
+     */
     maxParallelCalls: number;
+    /**
+     * Whether the model can plan and invoke multiple tools concurrently
+     * Example: true for models with parallel tool-use orchestration
+     */
     supportsParallel: boolean;
+    /**
+     * Whether requests must use a special "tool" role or channel for calls
+     * Example: true if provider requires tool role messages
+     */
     requiresToolRole: boolean;
+    /**
+     * Whether intermediate tool call tokens/results are streamable
+     * Example: true if tool call arguments/results are streamed as they are produced
+     */
     supportsStreaming: boolean;
+    /**
+     * JSON schema types supported for function parameters
+     * Examples: [JSON_SCHEMA_TYPE_OBJECT, JSON_SCHEMA_TYPE_ARRAY, JSON_SCHEMA_TYPE_STRING]
+     */
     supportedParameterTypes: JsonSchemaType[];
+    /**
+     * Maximum nesting depth allowed for function/tool call plans
+     * Example: 3 for limited recursion
+     */
     maxNestingDepth: number;
 }
 /** Vision capability configuration */
 export interface Vision {
+    /**
+     * Direction of vision support
+     * Example: MODALITY_DIRECTION_INPUT_ONLY for analysis-only models
+     * Example: MODALITY_DIRECTION_OUTPUT_ONLY for image generation models (DALL-E)
+     * Example: MODALITY_DIRECTION_INPUT_OUTPUT for models that can both analyze and generate
+     */
+    direction: ModalityDirection;
+    /**
+     * Supported image file formats
+     * Examples: [IMAGE_FORMAT_JPEG, IMAGE_FORMAT_PNG, IMAGE_FORMAT_WEBP]
+     */
     supportedFormats: ImageFormat[];
+    /**
+     * Maximum size per image in bytes
+     * Example: 20971520 (20MB) for GPT-4-vision
+     */
     maxImageSizeBytes: number;
+    /**
+     * Maximum images per API request (for input)
+     * Example: 10 for GPT-4-vision, 1 for some models
+     */
     maxImagesPerRequest: number;
-    supportsImageGeneration: boolean;
-    supportsVideoFrames: boolean;
+    /**
+     * Maximum image width in pixels
+     * Example: 4096 for high-resolution support
+     */
     maxResolutionWidth: number;
+    /**
+     * Maximum image height in pixels
+     * Example: 4096 for high-resolution support
+     */
     maxResolutionHeight: number;
+    /**
+     * Supports optical character recognition
+     * Example: true if model can extract text from images
+     */
     supportsOcr: boolean;
+    /**
+     * Supports object detection/localization
+     * Example: true if model can identify and locate objects
+     */
     supportsObjectDetection: boolean;
+    /**
+     * Can process video frames as images
+     * Example: true if model accepts video frame extraction
+     */
+    supportsVideoFrames: boolean;
 }
 /** Tool use capability configuration */
 export interface ToolUse {
+    /**
+     * Maximum number of tools that can be registered or considered
+     * Example: 64
+     */
     maxTools: number;
+    /**
+     * Whether the model can chain tool invocations one after another
+     * Example: true for stepwise tool planning
+     */
     supportsSequential: boolean;
+    /**
+     * Whether the model can invoke multiple tools concurrently
+     * Example: true for parallelized tool execution
+     */
     supportsParallel: boolean;
+    /**
+     * Maximum number of tool-use rounds allowed in a single request
+     * Example: 10
+     */
     maxToolRounds: number;
+    /**
+     * Types of tools supported by the provider/model
+     * Examples: [TOOL_TYPE_FUNCTION, TOOL_TYPE_RETRIEVAL]
+     */
     supportedToolTypes: ToolType[];
 }
 /** System prompt capability configuration */
 export interface SystemPrompt {
+    /**
+     * Maximum allowed length of system prompt content (characters or tokens)
+     * Example: 8000 (characters)
+     */
     maxLength: number;
+    /**
+     * Whether multiple system prompts (or segments) can be supplied
+     * Example: true if the provider supports multiple system messages
+     */
     supportsMultiple: boolean;
+    /**
+     * Whether system prompts can be cached for re-use
+     * Example: true if provider supports prompt caching hints
+     */
     supportsCaching: boolean;
+    /**
+     * Preferred/required format for system prompts (plain, markdown, etc.)
+     * Example: DATA_FORMAT_MARKDOWN
+     */
     format: DataFormat;
 }
 /** Caching capability configuration */
 export interface Caching {
+    /**
+     * Strategy used to compute cache keys for requests/responses
+     * Examples: CACHE_STRATEGY_HASH, CACHE_STRATEGY_SEMANTIC
+     */
     cacheKeyStrategy: CacheStrategy;
+    /**
+     * Maximum total cache capacity (bytes)
+     * Example: 1073741824 for 1GB
+     */
     maxCacheSizeBytes: number;
+    /**
+     * Default time-to-live for cache entries (seconds)
+     * Example: 3600 for 1 hour
+     */
     cacheTtlSeconds: number;
+    /**
+     * Whether the model/provider supports caching of conversation context
+     * Example: true for context window reuse
+     */
     supportsContextCaching: boolean;
+    /**
+     * Whether the model/provider supports caching of prompts/prefixes
+     * Example: true for prefix caching/token reuse
+     */
     supportsPromptCaching: boolean;
+    /**
+     * Minimum number of tokens required for an entry to be cacheable
+     * Example: 128
+     */
     minCacheableTokens: number;
 }
 /** Reasoning capability configuration */
 export interface Reasoning {
+    /**
+     * Whether the model can produce hidden chain-of-thought/assistant reasoning
+     * Example: true for models supporting private CoT traces
+     */
     supportsChainOfThought: boolean;
+    /**
+     * Whether the model can expose discrete step indices or markers
+     * Example: true for models that number reasoning steps
+     */
     supportsStepTracking: boolean;
+    /**
+     * Maximum number of internal reasoning steps supported
+     * Example: 64
+     */
     maxReasoningSteps: number;
+    /**
+     * Whether the model can revise earlier steps based on later insights
+     * Example: true for self-correction/self-reflection
+     */
     supportsSelfCorrection: boolean;
+    /**
+     * Supported high-level reasoning strategies
+     * Examples: [REASONING_STRATEGY_STEP_BY_STEP, REASONING_STRATEGY_SELF_CONSISTENCY]
+     */
     reasoningStrategies: ReasoningStrategy[];
 }
 /** Audio capability configuration */
 export interface Audio {
+    /**
+     * Direction of audio support
+     * Example: MODALITY_DIRECTION_INPUT_ONLY for speech-to-text only
+     * Example: MODALITY_DIRECTION_OUTPUT_ONLY for text-to-speech only
+     * Example: MODALITY_DIRECTION_INPUT_OUTPUT for models supporting both STT and TTS
+     */
+    direction: ModalityDirection;
+    /**
+     * Supported audio file formats
+     * Examples: [AUDIO_FORMAT_MP3, AUDIO_FORMAT_WAV, AUDIO_FORMAT_M4A]
+     */
     supportedFormats: AudioFormat[];
+    /**
+     * Maximum audio duration in seconds
+     * Example: 600 for 10-minute limit
+     */
     maxDurationSeconds: number;
+    /**
+     * Maximum audio file size in bytes
+     * Example: 26214400 (25MB) limit
+     */
     maxFileSizeBytes: number;
-    supportsTranscription: boolean;
-    supportsGeneration: boolean;
+    /**
+     * Supported languages for audio processing
+     * Examples: ["en", "es", "fr", "de", "zh", "ja"]
+     */
     supportedLanguages: string[];
+    /**
+     * Supports real-time streaming (for live audio)
+     * Example: true for real-time voice models
+     */
+    supportsStreaming: boolean;
+    /**
+     * Supports voice cloning or voice selection
+     * Example: true if TTS can use different voices
+     */
+    supportsVoiceSelection: boolean;
 }
 /** Video capability configuration */
 export interface Video {
+    /**
+     * Direction of video support
+     * Example: MODALITY_DIRECTION_INPUT_ONLY for video analysis/understanding
+     * Example: MODALITY_DIRECTION_OUTPUT_ONLY for video generation
+     * Example: MODALITY_DIRECTION_INPUT_OUTPUT for models that can both analyze and generate
+     */
+    direction: ModalityDirection;
+    /**
+     * Supported video file formats
+     * Examples: [VIDEO_FORMAT_MP4, VIDEO_FORMAT_MOV, VIDEO_FORMAT_AVI]
+     */
     supportedFormats: VideoFormat[];
+    /**
+     * Maximum video duration in seconds
+     * Example: 120 for 2-minute limit
+     */
     maxDurationSeconds: number;
+    /**
+     * Maximum video file size in bytes
+     * Example: 1073741824 (1GB) limit
+     */
     maxFileSizeBytes: number;
+    /**
+     * Maximum frames per second supported
+     * Example: 30 for standard frame rate
+     */
     maxFps: number;
+    /**
+     * Supports extracting and analyzing individual frames
+     * Example: true if model can process video as a sequence of images
+     */
     supportsFrameExtraction: boolean;
+    /**
+     * Maximum number of frames that can be analyzed
+     * Example: 100 for frame-by-frame analysis limit
+     */
+    maxFrames: number;
 }
 /** Embeddings capability configuration */
 export interface Embeddings {
+    /**
+     * Dimensionality of the vector space produced by the model
+     * Example: 1536
+     */
     embeddingDimensions: number;
+    /**
+     * Maximum number of tokens accepted per input item
+     * Example: 8192
+     */
     maxInputTokens: number;
+    /**
+     * Whether the API supports batching multiple inputs in one call
+     * Example: true
+     */
     supportsBatch: boolean;
+    /**
+     * Maximum number of items allowed per batch
+     * Example: 128
+     */
     maxBatchSize: number;
+    /**
+     * Supported similarity/distance metrics for index/search
+     * Examples: [DISTANCE_METRIC_COSINE, DISTANCE_METRIC_DOT_PRODUCT]
+     */
     distanceMetrics: DistanceMetric[];
 }
 /** Fine-tuning capability configuration */
 export interface FineTuning {
+    /**
+     * Minimum number of training examples required to start fine-tuning
+     * Example: 50
+     */
     minExamples: number;
+    /**
+     * Maximum number of training examples supported in a single job
+     * Example: 500000
+     */
     maxExamples: number;
+    /**
+     * Supported dataset/file formats for fine-tuning
+     * Examples: [DATA_FORMAT_JSONL, DATA_FORMAT_CSV]
+     */
     supportedFormats: DataFormat[];
+    /**
+     * Maximum size of individual training files (MB)
+     * Example: 512
+     */
     maxFileSizeMb: number;
+    /**
+     * Whether a separate validation set can be provided
+     * Example: true
+     */
     supportsValidationSet: boolean;
+    /**
+     * Hyperparameters that can be configured for training
+     * Examples: [HYPERPARAMETER_LEARNING_RATE, HYPERPARAMETER_EPOCHS]
+     */
     hyperparameters: Hyperparameter[];
 }
 export declare const Capability: MessageFns<Capability>;
+export declare const Text: MessageFns<Text>;
 export declare const StructuredResponse: MessageFns<StructuredResponse>;
 export declare const Streaming: MessageFns<Streaming>;
 export declare const FunctionCalling: MessageFns<FunctionCalling>;