UNPKG

@genkit-ai/ai

Version:

Genkit AI framework generative AI APIs.

459 lines 19 kB
"use strict"; var __defProp = Object.defineProperty; var __getOwnPropDesc = Object.getOwnPropertyDescriptor; var __getOwnPropNames = Object.getOwnPropertyNames; var __hasOwnProp = Object.prototype.hasOwnProperty; var __export = (target, all) => { for (var name in all) __defProp(target, name, { get: all[name], enumerable: true }); }; var __copyProps = (to, from, except, desc) => { if (from && typeof from === "object" || typeof from === "function") { for (let key of __getOwnPropNames(from)) if (!__hasOwnProp.call(to, key) && key !== except) __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable }); } return to; }; var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod); var model_exports = {}; __export(model_exports, { CandidateErrorSchema: () => CandidateErrorSchema, CandidateSchema: () => CandidateSchema, CustomPartSchema: () => CustomPartSchema, DataPartSchema: () => DataPartSchema, GenerateActionOptionsSchema: () => GenerateActionOptionsSchema, GenerateRequestSchema: () => GenerateRequestSchema, GenerateResponseChunkSchema: () => GenerateResponseChunkSchema, GenerateResponseSchema: () => GenerateResponseSchema, GenerationCommonConfigSchema: () => GenerationCommonConfigSchema, GenerationUsageSchema: () => GenerationUsageSchema, MediaPartSchema: () => MediaPartSchema, MessageSchema: () => MessageSchema, ModelInfoSchema: () => ModelInfoSchema, ModelRequestSchema: () => ModelRequestSchema, ModelResponseChunkSchema: () => ModelResponseChunkSchema, ModelResponseSchema: () => ModelResponseSchema, OutputConfigSchema: () => OutputConfigSchema, PartSchema: () => PartSchema, RoleSchema: () => RoleSchema, TextPartSchema: () => TextPartSchema, ToolDefinitionSchema: () => ToolDefinitionSchema, ToolRequestPartSchema: () => ToolRequestPartSchema, ToolResponsePartSchema: () => ToolResponsePartSchema, defineGenerateAction: () => import_action.defineGenerateAction, defineModel: () => defineModel, getBasicUsageStats: () => getBasicUsageStats, modelRef: () => modelRef, resolveModel: () => resolveModel, simulateConstrainedGeneration: () => import_middleware.simulateConstrainedGeneration }); module.exports = __toCommonJS(model_exports); var import_core = require("@genkit-ai/core"); var import_logging = require("@genkit-ai/core/logging"); var import_schema = require("@genkit-ai/core/schema"); var import_node_perf_hooks = require("node:perf_hooks"); var import_document = require("./document.js"); var import_middleware = require("./model/middleware.js"); var import_action = require("./generate/action.js"); const EmptyPartSchema = import_core.z.object({ text: import_core.z.never().optional(), media: import_core.z.never().optional(), toolRequest: import_core.z.never().optional(), toolResponse: import_core.z.never().optional(), data: import_core.z.unknown().optional(), metadata: import_core.z.record(import_core.z.unknown()).optional(), custom: import_core.z.record(import_core.z.unknown()).optional() }); const TextPartSchema = EmptyPartSchema.extend({ /** The text of the message. */ text: import_core.z.string() }); const MediaPartSchema = EmptyPartSchema.extend({ media: import_core.z.object({ /** The media content type. Inferred from data uri if not provided. */ contentType: import_core.z.string().optional(), /** A `data:` or `https:` uri containing the media content. */ url: import_core.z.string() }) }); const ToolRequestPartSchema = EmptyPartSchema.extend({ /** A request for a tool to be executed, usually provided by a model. */ toolRequest: import_core.z.object({ /** The call id or reference for a specific request. */ ref: import_core.z.string().optional(), /** The name of the tool to call. */ name: import_core.z.string(), /** The input parameters for the tool, usually a JSON object. */ input: import_core.z.unknown().optional() }) }); const ToolResponsePartSchema = EmptyPartSchema.extend({ /** A provided response to a tool call. */ toolResponse: import_core.z.object({ /** The call id or reference for a specific request. */ ref: import_core.z.string().optional(), /** The name of the tool. */ name: import_core.z.string(), /** The output data returned from the tool, usually a JSON object. */ output: import_core.z.unknown().optional() }) }); const DataPartSchema = EmptyPartSchema.extend({ data: import_core.z.unknown() }); const CustomPartSchema = EmptyPartSchema.extend({ custom: import_core.z.record(import_core.z.any()) }); const PartSchema = import_core.z.union([ TextPartSchema, MediaPartSchema, ToolRequestPartSchema, ToolResponsePartSchema, DataPartSchema, CustomPartSchema ]); const RoleSchema = import_core.z.enum(["system", "user", "model", "tool"]); const MessageSchema = import_core.z.object({ role: RoleSchema, content: import_core.z.array(PartSchema), metadata: import_core.z.record(import_core.z.unknown()).optional() }); const ModelInfoSchema = import_core.z.object({ /** Acceptable names for this model (e.g. different versions). */ versions: import_core.z.array(import_core.z.string()).optional(), /** Friendly label for this model (e.g. "Google AI - Gemini Pro") */ label: import_core.z.string().optional(), /** Supported model capabilities. */ supports: import_core.z.object({ /** Model can process historical messages passed with a prompt. */ multiturn: import_core.z.boolean().optional(), /** Model can process media as part of the prompt (multimodal input). */ media: import_core.z.boolean().optional(), /** Model can perform tool calls. */ tools: import_core.z.boolean().optional(), /** Model can accept messages with role "system". */ systemRole: import_core.z.boolean().optional(), /** Model can output this type of data. */ output: import_core.z.array(import_core.z.string()).optional(), /** Model supports output in these content types. */ contentType: import_core.z.array(import_core.z.string()).optional(), /** Model can natively support document-based context grounding. */ context: import_core.z.boolean().optional(), /** Model can natively support constrained generation. */ constrained: import_core.z.enum(["none", "all", "no-tools"]).optional(), /** Model supports controlling tool choice, e.g. forced tool calling. */ toolChoice: import_core.z.boolean().optional() }).optional(), /** At which stage of development this model is. * - `featured` models are recommended for general use. * - `stable` models are well-tested and reliable. * - `unstable` models are experimental and may change. * - `legacy` models are no longer recommended for new projects. * - `deprecated` models are deprecated by the provider and may be removed in future versions. */ stage: import_core.z.enum(["featured", "stable", "unstable", "legacy", "deprecated"]).optional() }); const ToolDefinitionSchema = import_core.z.object({ name: import_core.z.string(), description: import_core.z.string(), inputSchema: import_core.z.record(import_core.z.any()).describe("Valid JSON Schema representing the input of the tool.").nullish(), outputSchema: import_core.z.record(import_core.z.any()).describe("Valid JSON Schema describing the output of the tool.").nullish(), metadata: import_core.z.record(import_core.z.any()).describe("additional metadata for this tool definition").optional() }); const GenerationCommonConfigSchema = import_core.z.object({ /** A specific version of a model family, e.g. `gemini-1.0-pro-001` for the `gemini-1.0-pro` family. */ version: import_core.z.string().optional(), temperature: import_core.z.number().optional(), maxOutputTokens: import_core.z.number().optional(), topK: import_core.z.number().optional(), topP: import_core.z.number().optional(), stopSequences: import_core.z.array(import_core.z.string()).optional() }); const OutputConfigSchema = import_core.z.object({ format: import_core.z.string().optional(), schema: import_core.z.record(import_core.z.any()).optional(), constrained: import_core.z.boolean().optional(), instructions: import_core.z.string().optional(), contentType: import_core.z.string().optional() }); const ModelRequestSchema = import_core.z.object({ messages: import_core.z.array(MessageSchema), config: import_core.z.any().optional(), tools: import_core.z.array(ToolDefinitionSchema).optional(), toolChoice: import_core.z.enum(["auto", "required", "none"]).optional(), output: OutputConfigSchema.optional(), docs: import_core.z.array(import_document.DocumentDataSchema).optional() }); const GenerateRequestSchema = ModelRequestSchema.extend({ /** @deprecated All responses now return a single candidate. This will always be `undefined`. */ candidates: import_core.z.number().optional() }); const GenerationUsageSchema = import_core.z.object({ inputTokens: import_core.z.number().optional(), outputTokens: import_core.z.number().optional(), totalTokens: import_core.z.number().optional(), inputCharacters: import_core.z.number().optional(), outputCharacters: import_core.z.number().optional(), inputImages: import_core.z.number().optional(), outputImages: import_core.z.number().optional(), inputVideos: import_core.z.number().optional(), outputVideos: import_core.z.number().optional(), inputAudioFiles: import_core.z.number().optional(), outputAudioFiles: import_core.z.number().optional(), custom: import_core.z.record(import_core.z.number()).optional() }); const FinishReasonSchema = import_core.z.enum([ "stop", "length", "blocked", "interrupted", "other", "unknown" ]); const CandidateSchema = import_core.z.object({ index: import_core.z.number(), message: MessageSchema, usage: GenerationUsageSchema.optional(), finishReason: FinishReasonSchema, finishMessage: import_core.z.string().optional(), custom: import_core.z.unknown() }); const CandidateErrorSchema = import_core.z.object({ index: import_core.z.number(), code: import_core.z.enum(["blocked", "other", "unknown"]), message: import_core.z.string().optional() }); const ModelResponseSchema = import_core.z.object({ message: MessageSchema.optional(), finishReason: FinishReasonSchema, finishMessage: import_core.z.string().optional(), latencyMs: import_core.z.number().optional(), usage: GenerationUsageSchema.optional(), /** @deprecated use `raw` instead */ custom: import_core.z.unknown(), raw: import_core.z.unknown(), request: GenerateRequestSchema.optional() }); const GenerateResponseSchema = ModelResponseSchema.extend({ /** @deprecated All responses now return a single candidate. Only the first candidate will be used if supplied. Return `message`, `finishReason`, and `finishMessage` instead. */ candidates: import_core.z.array(CandidateSchema).optional(), finishReason: FinishReasonSchema.optional() }); const ModelResponseChunkSchema = import_core.z.object({ role: RoleSchema.optional(), /** index of the message this chunk belongs to. */ index: import_core.z.number().optional(), /** The chunk of content to stream right now. */ content: import_core.z.array(PartSchema), /** Model-specific extra information attached to this chunk. */ custom: import_core.z.unknown().optional(), /** If true, the chunk includes all data from previous chunks. Otherwise, considered to be incremental. */ aggregated: import_core.z.boolean().optional() }); const GenerateResponseChunkSchema = ModelResponseChunkSchema; function defineModel(registry, options, runner) { const label = options.label || options.name; const middleware = [ ...options.use || [], (0, import_middleware.validateSupport)(options) ]; if (!options?.supports?.context) middleware.push((0, import_middleware.augmentWithContext)()); const constratedSimulator = (0, import_middleware.simulateConstrainedGeneration)(); middleware.push((req, next) => { if (!options?.supports?.constrained || options?.supports?.constrained === "none" || options?.supports?.constrained === "no-tools" && (req.tools?.length ?? 0) > 0) { return constratedSimulator(req, next); } return next(req); }); const act = (0, import_core.defineAction)( registry, { actionType: "model", name: options.name, description: label, inputSchema: GenerateRequestSchema, outputSchema: GenerateResponseSchema, metadata: { model: { label, customOptions: options.configSchema ? (0, import_schema.toJsonSchema)({ schema: options.configSchema }) : void 0, versions: options.versions, supports: options.supports } }, use: middleware }, (input) => { const startTimeMs = import_node_perf_hooks.performance.now(); return runner(input, (0, import_core.getStreamingCallback)(registry)).then((response) => { const timedResponse = { ...response, latencyMs: import_node_perf_hooks.performance.now() - startTimeMs }; return timedResponse; }); } ); Object.assign(act, { __configSchema: options.configSchema || import_core.z.unknown() }); return act; } function modelRef(options) { const ref = { ...options }; ref.withConfig = (cfg) => { return modelRef({ ...options, config: cfg }); }; ref.withVersion = (version) => { return modelRef({ ...options, version }); }; return ref; } function getBasicUsageStats(input, response) { const inputCounts = getPartCounts(input.flatMap((md) => md.content)); const outputCounts = getPartCounts( Array.isArray(response) ? response.flatMap((c) => c.message.content) : response.content ); return { inputCharacters: inputCounts.characters, inputImages: inputCounts.images, inputVideos: inputCounts.videos, inputAudioFiles: inputCounts.audio, outputCharacters: outputCounts.characters, outputImages: outputCounts.images, outputVideos: outputCounts.videos, outputAudioFiles: outputCounts.audio }; } function getPartCounts(parts) { return parts.reduce( (counts, part) => { const isImage = part.media?.contentType?.startsWith("image") || part.media?.url?.startsWith("data:image"); const isVideo = part.media?.contentType?.startsWith("video") || part.media?.url?.startsWith("data:video"); const isAudio = part.media?.contentType?.startsWith("audio") || part.media?.url?.startsWith("data:audio"); return { characters: counts.characters + (part.text?.length || 0), images: counts.images + (isImage ? 1 : 0), videos: counts.videos + (isVideo ? 1 : 0), audio: counts.audio + (isAudio ? 1 : 0) }; }, { characters: 0, images: 0, videos: 0, audio: 0 } ); } async function resolveModel(registry, model, options) { let out; let modelId; if (!model) { model = await registry.lookupValue("defaultModel", "defaultModel"); } if (!model) { throw new import_core.GenkitError({ status: "INVALID_ARGUMENT", message: "Must supply a `model` to `generate()` calls." }); } if (typeof model === "string") { modelId = model; out = { modelAction: await registry.lookupAction(`/model/${model}`) }; } else if (model.hasOwnProperty("__action")) { modelId = model.__action.name; out = { modelAction: model }; } else { const ref = model; modelId = ref.name; out = { modelAction: await registry.lookupAction( `/model/${ref.name}` ), config: { ...ref.config }, version: ref.version }; } if (!out.modelAction) { throw new import_core.GenkitError({ status: "NOT_FOUND", message: `Model '${modelId}' not found` }); } if (options?.warnDeprecated && out.modelAction.__action.metadata?.model?.stage === "deprecated") { import_logging.logger.warn( `Model '${out.modelAction.__action.name}' is deprecated and may be removed in a future release.` ); } return out; } const GenerateActionOptionsSchema = import_core.z.object({ /** A model name (e.g. `vertexai/gemini-1.0-pro`). */ model: import_core.z.string(), /** Retrieved documents to be used as context for this generation. */ docs: import_core.z.array(import_document.DocumentDataSchema).optional(), /** Conversation history for multi-turn prompting when supported by the underlying model. */ messages: import_core.z.array(MessageSchema), /** List of registered tool names for this generation if supported by the underlying model. */ tools: import_core.z.array(import_core.z.string()).optional(), /** Tool calling mode. `auto` lets the model decide whether to use tools, `required` forces the model to choose a tool, and `none` forces the model not to use any tools. Defaults to `auto`. */ toolChoice: import_core.z.enum(["auto", "required", "none"]).optional(), /** Configuration for the generation request. */ config: import_core.z.any().optional(), /** Configuration for the desired output of the request. Defaults to the model's default output if unspecified. */ output: import_core.z.object({ format: import_core.z.string().optional(), contentType: import_core.z.string().optional(), instructions: import_core.z.union([import_core.z.boolean(), import_core.z.string()]).optional(), jsonSchema: import_core.z.any().optional(), constrained: import_core.z.boolean().optional() }).optional(), /** Options for resuming an interrupted generation. */ resume: import_core.z.object({ respond: import_core.z.array(ToolResponsePartSchema).optional(), restart: import_core.z.array(ToolRequestPartSchema).optional(), metadata: import_core.z.record(import_core.z.any()).optional() }).optional(), /** When true, return tool calls for manual processing instead of automatically resolving them. */ returnToolRequests: import_core.z.boolean().optional(), /** Maximum number of tool call iterations that can be performed in a single generate call (default 5). */ maxTurns: import_core.z.number().optional() }); // Annotate the CommonJS export names for ESM import in node: 0 && (module.exports = { CandidateErrorSchema, CandidateSchema, CustomPartSchema, DataPartSchema, GenerateActionOptionsSchema, GenerateRequestSchema, GenerateResponseChunkSchema, GenerateResponseSchema, GenerationCommonConfigSchema, GenerationUsageSchema, MediaPartSchema, MessageSchema, ModelInfoSchema, ModelRequestSchema, ModelResponseChunkSchema, ModelResponseSchema, OutputConfigSchema, PartSchema, RoleSchema, TextPartSchema, ToolDefinitionSchema, ToolRequestPartSchema, ToolResponsePartSchema, defineGenerateAction, defineModel, getBasicUsageStats, modelRef, resolveModel, simulateConstrainedGeneration }); //# sourceMappingURL=model.js.map