@eidentic/model

Model and embedder wrappers for Eidentic — AIModel wraps any Vercel AI SDK 7 provider, AIEmbedder wraps embedding models, and the package ships a live price table (sourced from LiteLLM).

Install

pnpm add @eidentic/model ai @ai-sdk/anthropic
# or any other @ai-sdk/* provider

@eidentic/model follows AI SDK 7 and is ESM-only. Use ESM import, not CommonJS require().

Usage

import { AIModel, AIEmbedder } from "@eidentic/model";
import { anthropic } from "@ai-sdk/anthropic";
import { openai } from "@ai-sdk/openai";

// Wrap a provider model
const model = new AIModel(anthropic("claude-sonnet-4-5"));

// Wrap an embedding model (async factory — the dimension is probed automatically)
const embedder = await AIEmbedder.create(openai.embedding("text-embedding-3-small"));

// Check current prices
import { defaultPrices, pricesUpdatedAt } from "@eidentic/model";
console.log(defaultPrices["claude-sonnet-4-5"]); // { inputUsd: ..., outputUsd: ... }

// Ollama for local inference
// pnpm add ai-sdk-ollama
import { ollama } from "ai-sdk-ollama";
const local = new AIModel(ollama("llama3.2"));

Remote image URLs are denied at the model boundary. Resolve untrusted URLs through a guarded, size-bounded fetcher and pass base64 data to the Agent's multimodal policy. The deprecated unsafeAllowRemoteImageUrls: true option restores provider-side URL fetching only for controlled legacy deployments with an equivalent egress boundary.

Links

• GitHub
• Issue tracker
• Root README

Apache-2.0