@xpell/xai-providers

Official AI provider implementations for the Xpell AI Runtime.

@xpell/xai-providers contains provider adapters that implement the XAIProvider interface used by @xpell/node and the XAI runtime.

The goal is to provide a consistent API across different AI backends while allowing Xpell applications, modules, and services to switch providers without changing business logic.

⸻

Installation

pnpm add @xpell/xai-providers

⸻

Available Providers

AzureProvider

Direct integration with Azure OpenAI deployments.

import { AzureProvider } from "@xpell/xai-providers/azure";

Supports:

• GPT-4o
• GPT-4o-mini
• GPT-5
• Azure OpenAI compatible deployments

⸻

AimeProvider

Integration with XAI Router.

Instead of talking directly to Azure/OpenAI, the provider sends requests to an XAI Router instance using an Xpell AI API Key.

import { AimeProvider } from "@xpell/xai-providers/aime";

Supports:

• XAI Router
• Usage tracking
• Token accounting
• Routing rules
• Centralized billing
• Multi-provider routing

⸻

MockProvider

Testing provider.

import { MockProvider } from "@xpell/xai-providers/mock";

Useful for:

• Unit tests
• Local development
• Runtime validation

⸻

Usage

Registering Providers

Providers are registered through the XAI runtime.

import { _xai } from "@xpell/node"; import { AzureProvider } from "@xpell/xai-providers/azure"; _xai.registerProvider( "azure", new AzureProvider({ endpoint: process.env.AZURE_OPENAI_ENDPOINT!, apiKey: process.env.AZURE_OPENAI_API_KEY!, deployment: process.env.AZURE_OPENAI_DEPLOYMENT!, }) );

Set default provider:

await _x.execute({ _module: "xai", _op: "set_default", _params: { _provider: "azure" } });

⸻

Using AimeProvider

import { _xai } from "@xpell/node"; import { AimeProvider } from "@xpell/xai-providers/aime"; _xai.registerProvider( "aime", new AimeProvider({ endpoint: process.env.AIME_ENDPOINT!, apiKey: process.env.AIME_API_KEY!, }) );

Set as default:

await _x.execute({ _module: "xai", _op: "set_default", _params: { _provider: "aime" } });

All XAI calls will now be routed through XAI Router.

⸻

XAIProvider Interface

Every provider implements:

interface XAIProvider { generate( input: XAIInput ): Promise; }

⸻

XAIInput

type XAIInput = { prompt: string; system?: string; context?: any; response_format?: { type: "text" | "json_object"; }; };

⸻

XAIResult

type XAIUsage = { input_tokens?: number; output_tokens?: number; total_tokens?: number; }; type XAIResult = { text: string; raw?: unknown; usage?: XAIUsage; };

⸻

Token Usage

Providers should return token usage whenever available.

Example:

{ text: "Hello world", usage: { input_tokens: 14, output_tokens: 10, total_tokens: 24 } }

The XAI runtime, XVibe, XStudio, and XAI Router use this information for:

• telemetry
• analytics
• usage tracking
• billing
• routing decisions

⸻

Architecture

Application ↓ XAI ↓ XAIProvider ↓ Provider API

Examples:

XVibe ↓ XAI ↓ AzureProvider ↓ Azure OpenAI XVibe ↓ XAI ↓ AimeProvider ↓ XAI Router ↓ Azure/OpenAI/Anthropic

⸻

Project Goals

• Unified provider interface
• Runtime provider switching
• Token usage normalization
• Structured output support
• Provider-agnostic application code
• Seamless integration with XAI Router

⸻

License

Part of the Xpell AI ecosystem.

https://xpell.ai