Normalize Local AI Gateway base URL in dev proxy setups

[Copilot]

The Local AI Gateway status endpoint could return a frontend-origin base URL (http://localhost:5173/v1) when accessed through Vite, which caused downstream clients to hit invalid paths like /v1/v1/models and parse HTML error pages as JSON. This change normalizes the returned gateway base URL to the backend origin.

• Gateway base URL normalization

  • Added origin normalization in src/routes/gateway.ts via gatewayBaseUrl(...).
  • When request origin is Vite dev server (:5173), base URL is rewritten to backend WEB_PORT (default 3000).
  • Applied consistently to both GET /api/gateway and POST /api/gateway responses.

• Safer request-origin fallback

  • If ctx.origin is absent, origin is derived from ctx.protocol + ctx.host before normalization.

• Regression coverage

  • Extended src/routes/__tests__/gateway.test.ts with a case asserting Host: localhost:5173 returns gateway.baseUrl === "http://localhost:3000/v1".

function gatewayBaseUrl(origin: string | null | undefined): string {
  const rawOrigin = typeof origin === 'string' ? origin : '';
  try {
    const url = new URL(rawOrigin);
    if (url.port === '5173') url.port = process.env.WEB_PORT ?? '3000';
    return `${url.origin}/v1`;
  } catch {
    return rawOrigin ? `${rawOrigin.replace(/\/$/, '')}/v1` : '/v1';
  }
}

Original prompt

message: Gateway /v1/models was unreachable: Unexpected token '<', "<!doctype "... is not valid JSON
requestUrl: http://localhost:5173/v1/v1/models
endpoint: http://localhost:5173/
checkedAt: 2026-05-30T14:30:25.391Z

[Chronological Review: The conversation began with the user requesting the implementation of a feature related to the Local AI Gateway. The user then sought clarification on how to use this feature. Following that, the user suggested adding a toggle switch and configuration capability for the gatewayToken in the settings, indicating a desire for improved usability.]

[Intent Mapping:

• "实现这个feature" - The user explicitly requested the implementation of a feature related to the Local AI Gateway.
• "具体怎么使用呢？" - The user asked for specific instructions on how to use the feature.
• "gatewayToken 配置太极客了吧，或许可以在设置里面增加一个开关和配置能力？" - The user suggested enhancing the configuration of gatewayToken by adding a toggle switch in the settings.]

[Technical Inventory:

• Technologies mentioned include Koa for backend HTTP services, Vercel AI SDK for LLM integration, and various AI models (Gemini, DeepSeek, OpenAI, Anthropic, Claude Code, Ollama).
• Key files referenced include:
  • src/llm/client.ts: Handles AI SDK model calls, streaming, fallback, and usage recording.
  • src/config.ts: Manages model aliases, provider key access, and configuration.
  • src/routes/model.ts: Provides model lists, provider status, and routing configuration APIs.]

[Code Archaeology:

• No specific code changes were discussed in detail during this segment, but the user is focused on enhancing the configuration capabilities related to gatewayToken.]

[Progress Assessment:

• The user has outlined the existing capabilities of the Local AI Gateway and identified areas for improvement, particularly regarding user configuration options.
• No specific implementation tasks have been completed yet, as the conversation is still in the planning and suggestion phase.]

[Context Validation: All critical information regarding the user's requests and suggestions for the Local AI Gateway feature is captured, ensuring continuity in the conversation.]

[Recent Commands Analysis:

• Last Agent Commands: The user requested the implementation of a feature, asked for usage instructions, and suggested a configuration enhancement.
• Tool Results Summary: No specific tool results were generated as the conversation primarily consisted of user requests and suggestions.
• Pre-Summary State: The agent was actively discussing the implementation and usability of the Local AI Gateway feature when the token budget was exceeded.
• Operation Context: The user's commands were aimed at improving the functionality and user experience of the Local AI Gateway, particularly regarding the gatewayToken configuration.]

1. Conversation Overview: - Primary Objectives: - "实现这个feature" - User requested implementation of a feature for the Local AI Gateway. - "具体怎么使用呢？" - User sought clarification on how to use the feature. - "gatewayToken 配置太极客了吧，或许可以在设置里面增加一个开关和配置能力？" - User suggested adding a toggle switch for gatewayToken configuration. - Session Context: The conversation has focused on the Local AI Gateway, discussing its current capabilities and potential enhancements. - User Intent Evolution: The user transitioned from requesting implementation to seeking usage instructions and suggesting usability improvements.

2. Technical Foundation:

• Koa: Backend HTTP service framework used for API management.
• Vercel AI SDK: Framework for integrating various AI models.
• Local AI Gateway: Aims to provide a unified model entry point for external clients.

3. Codebase Status:

• File Name: src/llm/client.ts
  • Purpose: Handles AI SDK model calls and usage recording.
  • Current State: No recent changes discussed.
• File Name: src/config.ts
  • Purpose: Manages model aliases and provider key access.
  • Current State: No recent changes discussed.
• File Name: src/routes/model.ts
  • Purpose: Provides model lists and routing configuration.
  • Current State: No recent changes discussed.

4. Problem Resolution:

• Issues Encountered: User expressed concerns about the complexity of gatewayToken configuration.
• Solutions Implemented: User suggested adding a toggle switch for easier configuration.
• Debugging Context: No ongoing troubleshooting efforts were mentioned.
• Lessons Learned: User feedback indicates a need for improved usability in configuration settings.

5. Progress Tracking:

• Completed Tasks: None reported yet.
• Partially Complete Work: User has outlined feature requests and suggestions for improvement.
• Validated Outcomes: No features confirmed working through testing yet.

6. Active Work State:

• Current Focus: Discussing the implementation and usability of the Local AI Gateway feature.
• Recent Context: User has been actively requesting feature implementation and usability enhancements.
• Working Code: No speci...

Created from VS Code.