Feat/local llm

[rezawr]

Feature: Support Custom baseUrl and Local LLM Endpoints

Summary

This change extends the zerox SDK by adding a baseUrl field to the OpenAICredentials interface and model-factory logic. Users can now:

• Specify custom OpenAI-compatible endpoints (proxy, staging, internal)
• Connect to self-hosted or local LLM servers (for example, http://localhost:1234/v1)

When baseUrl is omitted, the SDK continues to use the default https://api.openai.com/v1.

Changes

1. Types

   • File: node-zerox/src/types.ts

     export interface OpenAICredentials {
       apiKey:  string;
       baseUrl: string;    // newly added field
     }

2. Credential Validation

   • File: node-zerox/src/models/index.ts
     Updated the isOpenAICredentials() type guard to require baseUrl in addition to apiKey.

3. OpenAI Model Implementation

   • File: node-zerox/src/models/openai.ts
     • Constructor reads credentials.baseUrl (falling back if necessary).
     • All HTTP requests (chat completions, embeddings, etc.) now use this.baseUrl.

4. Factory Logic

   • File: node-zerox/src/models/index.ts
     • createModel() passes through the new baseUrl property to the model implementation.

5. Examples and Tests

   • Added unit tests in node-zerox/src/models/openai.test.ts:
     • Verify default base URL when baseUrl is not provided.
     • Verify custom base URL is respected.
   • Updated example in examples/ocr.ts:

     import { zerox } from "zerox";
     
     const result = await zerox({
       filePath: "invoice.png",
       model:    "qwen-2-vl-7b-ocr",
       credentials: {
         apiKey:  "",
         baseUrl: "http://localhost:1234/v1"
       }
     });