Spice AI Provider

Vercel AI Provider for Spice runtime

Installation

1npm install @spiceai/spice-ai-provider

Usage

Check Spice.ai to learn how to install and get started with local runtime:

• Install and start spice runtime
• Configure OpenAI or compatible LLM models

Connect to local spice runtime

You can import default spice provider instance spice from @spiceai/spice-ai-provider

1import { spice } from "@spiceai/spice-ai-provider";

If you need a cusom setup, use createSpice and create provider with your settings:

1import { createSpice } from "@spiceai/spice-ai-provider";
2
3const spice = createSpice({
4  // ...
5});

Provider settings:

• baseURL - optional, defaults to http://localhost:8090/v1

Example

Navigate to examples/sample-app

1. Install Spice runtime. Learn more.

1curl https://install.spiceai.org | /bin/bash

2. Copy .env.example → .env, and set your OpenAI API key and GitHub PAT for data connector

3. Start Spice runtime

1spice run
2Checking for latest Spice runtime release...
3Spice.ai runtime starting...
42024-09-10T14:17:37.598982Z  INFO runtime::flight: Spice Runtime Flight listening on 127.0.0.1:50051
52024-09-10T14:17:37.599070Z  INFO runtime::metrics_server: Spice Runtime Metrics listening on 127.0.0.1:9090
62024-09-10T14:17:37.599263Z  INFO runtime::http: Spice Runtime HTTP listening on 127.0.0.1:8090
72024-09-10T14:17:37.600010Z  INFO runtime::opentelemetry: Spice Runtime OpenTelemetry listening on 127.0.0.1:50052
82024-09-10T14:17:38.175204Z  INFO runtime: Embedding [openai_embeddings] ready to embed
92024-09-10T14:17:38.175396Z  INFO runtime: Tool [document_similarity] ready to use
102024-09-10T14:17:38.175416Z  INFO runtime: Tool [table_schema] ready to use
112024-09-10T14:17:38.175421Z  INFO runtime: Tool [sql] ready to use
122024-09-10T14:17:38.175428Z  INFO runtime: Tool [list_datasets] ready to use
132024-09-10T14:17:38.179149Z  INFO runtime: Initialized results cache; max size: 128.00 MiB, item ttl: 1s
142024-09-10T14:17:38.179779Z  INFO runtime: Loading model [gpt-4o] from openai:gpt-4o...
152024-09-10T14:17:38.686558Z  INFO runtime: Model [gpt-4o] deployed, ready for inferencing
162024-09-10T14:17:39.174429Z  INFO runtime: Dataset vercel_ai_docs registered (github:github.com/vercel/ai/files/main), acceleration (arrow), results cache enabled.
172024-09-10T14:17:39.175632Z  INFO runtime::accelerated_table::refresh_task: Loading data for dataset vercel_ai_docs
182024-09-10T14:17:51.810106Z  INFO runtime::accelerated_table::refresh_task: Loaded 143 rows (2.29 MiB) for dataset vercel_ai_docs in 12s 634ms.

4. Start the web app

1npm run dev

Navigate to http://localhost:3000, and now you can chat with Vercel AI SDK docs: