Add support for GLM-4.7 and Minimax-2.1 models

[DrKyro]

Feature Request

It would be great to add support for GLM-4.7 (by Zhipu AI) and Minimax-2.1 (by MiniMax) models to this project.

Use Case

These models are popular in the AI development community and having support for them would allow more flexibility in choosing AI providers for autonomous coding tasks.

Proposed Implementation

1. Add model aliases in :

   • ->
   • ->

2. Create new providers:

   • for Zhipu AI's API
   • for MiniMax's API

3. Update type definitions:

   • Add to type
   • Add provider prefixes and utility functions

4. Update UI:

   • Add model selection options in settings
   • Add configuration tabs for API credentials

References

• GLM-4.7: https://www.zhipuai.cn/
• Minimax-2.1: https://www.minimaxi.com/

Thank you for considering this feature request!

[sschneiders]

Hey,

in release 0.6.0 it was possible to configure Claude Code to use a different api compatible provider and automaker worked fine using it. Since 0.7.0 it looks like this does not work anymore. See #359

[imetallica]

Can we also include support for the non-chinese version of the GLM API:

https://api.z.ai/api/coding/paas/v4
https://docs.z.ai/guides/overview/quick-start

[jmvl]

Custom Endpoint Support (Prerequisite for GLM)

Commit: 1b16ab8f - feat: support ANTHROPIC_BASE_URL and ANTHROPIC_AUTH_TOKEN for custom endpoints

Enables AutoMaker to use custom API endpoints (gateways, GLM, Z.AI, etc.) by allowing additional environment variables to pass through to the Claude Agent SDK.

Why This Change Is Needed

The Claude Agent SDK by default only allows a minimal set of environment variables. This change adds support for:

• ANTHROPIC_BASE_URL: Redirect API requests to custom endpoints (e.g., Z.AI gateway)
• ANTHROPIC_AUTH_TOKEN: Alternative authentication method used by some gateways
• API_TIMEOUT_MS: Configure request timeout for long-running operations

Without this change, the .env configuration for GLM/Z.AI would be ignored by the SDK.

Files Modified

apps/server/src/providers/claude-provider.ts (Lines 12-48)

Add to ALLOWED_ENV_VARS array (after ANTHROPIC_API_KEY):

const ALLOWED_ENV_VARS = [
  'ANTHROPIC_API_KEY',
  'ANTHROPIC_BASE_URL', // Custom API endpoints (gateways, GLM, Z.AI, etc.)
  'ANTHROPIC_AUTH_TOKEN', // Alternative auth for gateways
  'API_TIMEOUT_MS', // Request timeout configuration
  'PATH',
  'HOME',
  'SHELL',
  'TERM',
  'USER',
  'LANG',
  'LC_ALL',
];

Add validation function (after ALLOWED_ENV_VARS, around line 24-35):

/**
 * Validate URL format and warn if it looks invalid
 */
function validateBaseUrl(url: string): void {
  try {
    const parsed = new URL(url);
    if (!['http:', 'https:'].includes(parsed.protocol)) {
      logger.warn(`ANTHROPIC_BASE_URL has unexpected protocol: ${parsed.protocol}`);
    }
  } catch {
    logger.warn(`ANTHROPIC_BASE_URL appears to be invalid: ${url}`);
  }
}

Update buildEnv() function (around line 37-48):

/**
 * Build environment for the SDK with only explicitly allowed variables
 */
function buildEnv(): Record<string, string | undefined> {
  const env: Record<string, string | undefined> = {};
  for (const key of ALLOWED_ENV_VARS) {
    if (process.env[key]) {
      env[key] = process.env[key];

      // Validate ANTHROPIC_BASE_URL format
      if (key === 'ANTHROPIC_BASE_URL') {
        validateBaseUrl(env[key]!);
      }
    }
  }
  return env;
}

Update sdkOptions to use buildEnv() (around line 107):

const sdkOptions: Options = {
  model,
  systemPrompt,
  maxTurns,
  cwd,
  // Pass only explicitly allowed environment variables to SDK
  env: buildEnv(),
  // ... rest of options
};

Testing

1. Set the environment variables in apps/server/.env:

   ANTHROPIC_BASE_URL=https://api.z.ai/api/anthropic
   ANTHROPIC_AUTH_TOKEN=your_token_here
   API_TIMEOUT_MS=3000000

2. Start the server and check logs for validation warnings (if any)

3. Create a feature and verify it uses the custom endpoint:

   • Check network requests in browser DevTools
   • Verify requests go to the custom BASE_URL
   • Confirm AUTH_TOKEN is used instead of API_KEY if set

---

1. GLM Model Support (Z.AI Endpoint)

Commit: 8c734f60 - feat: add GLM model support for Z.AI endpoint

Enable AutoMaker to use GLM models via Z.AI gateway instead of default Anthropic Claude models.

Files Modified

apps/server/.env (Create this file - NOT in git)

# GLM/Z.AI Configuration
ANTHROPIC_API_KEY=your_glm_api_key_here
ANTHROPIC_AUTH_TOKEN=your_glm_api_key_here
ANTHROPIC_BASE_URL=https://api.z.ai/api/anthropic
API_TIMEOUT_MS=3000000
ANTHROPIC_DEFAULT_HAIKU_MODEL=glm-4.5-air
ANTHROPIC_DEFAULT_SONNET_MODEL=glm-4.7
ANTHROPIC_DEFAULT_OPUS_MODEL=glm-4.7

Important: This file is in .gitignore and must be recreated manually after updates.

libs/types/src/model.ts (Line 4-8)

Before:

export const CLAUDE_MODEL_MAP: Record<string, string> = {
  haiku: 'claude-haiku-4-5-20251001',
  sonnet: 'claude-sonnet-4-5-20250929',
  opus: 'claude-opus-4-5-20251101',
} as const;

After:

export const CLAUDE_MODEL_MAP: Record<string, string> = {
  haiku: 'glm-4.5-air',
  sonnet: 'glm-4.7',
  opus: 'glm-4.7',
} as const;

libs/model-resolver/src/resolver.ts (Line 69-73)

Before:

// Full Claude model string - pass through unchanged
if (modelKey.includes('claude-')) {
  console.log(`[ModelResolver] Using full Claude model string: ${modelKey}`);
  return modelKey;
}

After:

// Full Claude or GLM model string - pass through unchanged
if (modelKey.includes('claude-') || modelKey.includes('glm-')) {
  console.log(`[ModelResolver] Using full model string: ${modelKey}`);
  return modelKey;
}

---

Re-application Checklist

After pulling updates from upstream, verify these changes are still present:

Custom Endpoint Support (Prerequisite)

• apps/server/src/providers/claude-provider.ts has ANTHROPIC_BASE_URL, ANTHROPIC_AUTH_TOKEN, and API_TIMEOUT_MS in ALLOWED_ENV_VARS
• validateBaseUrl() function exists and is called in buildEnv()
• sdkOptions uses env: buildEnv() to pass environment variables

GLM Support

• apps/server/.env exists with GLM configuration
• libs/types/src/model.ts has GLM model mappings
• libs/model-resolver/src/resolver.ts includes GLM in pass-through check

---

Build Commands

After re-applying changes, rebuild packages:

npm run build:packages
npm run build -w @automaker/ui

---

Testing

GLM Support

1. Start the app
2. Check logs for: [dotenv] injecting env (5) or similar
3. Create a feature and verify GLM models are used
4. Check agent output for GLM model references

---

Commit History Reference

8c734f60 feat: add GLM model support for Z.AI endpoint (Section 1)
1b16ab8f feat: support ANTHROPIC_BASE_URL and ANTHROPIC_AUTH_TOKEN for custom endpoints (Section 0 - Prerequisite)

---

Notes

• Section 0 (Custom Endpoint Support) is a prerequisite for Section 1 (GLM Support) - without it, the GLM .env configuration would be ignored
• The .env file is intentionally not tracked in git (security measure)
• Always backup the .env file content before major updates
• GLM model strings must include "glm-" prefix to pass through the model resolver

[GhostDragonAlpha]

If you can do it then why haven't we done it already What are you hiding?