Document and expose model.openai.base-url in application.yml and README

[devops-thiago]

Background

ModelConfig.OpenAISettings defines a baseUrl field, and OpenAIModelProvider (src/main/java/br/com/arquivolivre/myjavagenie/service/OpenAIModelProvider.java, ~line 57) already applies it (builder.baseUrl(...)). However, src/main/resources/application.yml does not define model.openai.base-url, so the only way to use a custom OpenAI-compatible endpoint with the OpenAI provider is the undocumented MODEL_OPENAI_BASE_URL env var via Spring relaxed binding.

Impact

• Users of OpenAI-compatible proxies/gateways (LM Studio, vLLM, OpenRouter, Ollama Cloud, etc.) cannot discover how to configure a custom base URL for the OpenAI provider.
• The config surface is incomplete relative to the code's capabilities.
• The self-hosted provider documents base-url in YAML, but the OpenAI provider does not.

How to reproduce

1. Set model.provider: openai and point the app at an OpenAI-compatible endpoint.
2. Search application.yml and the README for base-url under the openai section — nothing documents it.

Where the fix should land

Add the missing property to src/main/resources/application.yml:

model:
  openai:
    api-key: ${OPENAI_API_KEY:}
    base-url: ${OPENAI_BASE_URL:}
    model-name: gpt-4
    timeout-seconds: 300

Document the variable in the README configuration section.

Files touched

• src/main/resources/application.yml
• README.md
• Optionally src/test/java/.../OpenAIModelProviderTest.java asserting the custom base URL is applied to the builder.

Acceptance criteria

• model.openai.base-url (and the OPENAI_BASE_URL env var) is documented in YAML and README.
• With model.provider: openai and a custom base-url, requests are sent to that endpoint (verifiable via a WireMock test).
• Default behavior (no base-url set) is unchanged — falls back to the OpenAI default endpoint.