@genkit-ai/vertexai/lib/modelgarden/v2/llama.mjs

Genkit AI framework plugin for Google Cloud Vertex AI APIs including Gemini APIs, Imagen, and more.

import { GenkitError, z } from "genkit";
import {
  modelRef
} from "genkit/model";
import OpenAI from "openai";
import { getGenkitClientHeader } from "../../common/index.mjs";
import {
  OpenAIConfigSchema,
  defineOpenaiCompatibleModel
} from "./openai_compatibility.mjs";
import { checkModelName } from "./utils.mjs";
const LlamaConfigSchema = OpenAIConfigSchema.extend({
  location: z.string().optional()
}).passthrough();
function commonRef(name, info, configSchema = LlamaConfigSchema) {
  return modelRef({
    name: `vertex-model-garden/${name}`,
    configSchema,
    info: info ?? {
      supports: {
        multiturn: true,
        tools: true,
        media: true,
        systemRole: true,
        output: ["text", "json"]
      }
    }
  });
}
const GENERIC_MODEL = commonRef("llama");
const KNOWN_MODELS = {
  "meta/llama-4-maverick-17b-128e-instruct-maas": commonRef(
    "meta/llama-4-maverick-17b-128e-instruct-maas"
  ),
  "meta/llama-4-scout-17b-16e-instruct-maas": commonRef(
    "meta/llama-4-scout-17b-16e-instruct-maas"
  ),
  "meta/llama-3.3-70b-instruct-maas": commonRef(
    "meta/llama-3.3-70b-instruct-maas"
  )
};
function isLlamaModelName(value) {
  return !!value?.startsWith("meta/llama-");
}
function model(version, options = {}) {
  const name = checkModelName(version);
  return modelRef({
    name: `vertex-model-garden/${name}`,
    config: options,
    configSchema: LlamaConfigSchema,
    info: {
      ...GENERIC_MODEL.info
    }
  });
}
function listActions(clientOptions) {
  return [];
}
function listKnownModels(clientOptions, pluginOptions) {
  return Object.keys(KNOWN_MODELS).map(
    (name) => defineModel(name, clientOptions, pluginOptions)
  );
}
function defineModel(name, clientOptions, pluginOptions) {
  const ref = model(name);
  const clientFactory = async (request) => {
    const options = await resolveOptions(clientOptions, request.config);
    return new OpenAI(options);
  };
  return defineOpenaiCompatibleModel(ref, clientFactory);
}
async function resolveOptions(clientOptions, requestConfig) {
  const baseUrlTemplate = clientOptions.baseUrlTemplate ?? "https://{location}-aiplatform.googleapis.com/v1/projects/{projectId}/locations/{location}/endpoints/openapi";
  const location = requestConfig?.location || clientOptions.location;
  const baseURL = baseUrlTemplate.replace(/{location}/g, location).replace(/{projectId}/g, clientOptions.projectId);
  const apiKey = await clientOptions.authClient.getAccessToken();
  if (!apiKey) {
    throw new GenkitError({
      status: "PERMISSION_DENIED",
      message: "Unable to get accessToken"
    });
  }
  const defaultHeaders = {
    "X-Goog-Api-Client": getGenkitClientHeader()
  };
  return { baseURL, apiKey, defaultHeaders };
}
export {
  GENERIC_MODEL,
  KNOWN_MODELS,
  LlamaConfigSchema,
  defineModel,
  isLlamaModelName,
  listActions,
  listKnownModels,
  model
};
//# sourceMappingURL=llama.mjs.map