fix: propagate alternative API provider settings to agent subprocesses

When users configured GLM/Ollama/Kimi via the Settings UI, agents still
used Claude because conflicting env vars leaked through subprocess env.

Root cause: get_effective_sdk_env() set ANTHROPIC_AUTH_TOKEN for GLM but
didn't clear ANTHROPIC_API_KEY, which leaked from os.environ. The CLI
prioritized the wrong credential.

Changes:
- registry.py: Clear conflicting auth vars (API_KEY vs AUTH_TOKEN) and
  Vertex AI vars when building env for alternative providers
- client.py: Replace manual os.getenv() loop with get_effective_sdk_env()
  so agent SDK reads provider settings from the database
- autonomous_agent_demo.py: Apply UI-configured provider settings to
  process env so CLI-launched agents also respect Settings UI config
- start.py: Pass --model from settings when launching agent subprocess
- server/schemas.py: Allow non-Claude model names when an alternative
  provider is configured (prevents 422 errors for glm-4.7, etc.)
- .env.example: Document env vars for GLM, Ollama, and Kimi providers

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

.env.example

@@ -32,5 +32,25 @@
 # ===================
 # Alternative API Providers (GLM, Ollama, Kimi, Custom)
 # ===================
-# Configure alternative providers via the Settings UI (gear icon > API Provider).
-# The Settings UI is the recommended way to switch providers and models.
+# Configure via Settings UI (recommended) or set env vars below.
+# When both are set, env vars take precedence.
+#
+# GLM (Zhipu AI):
+# ANTHROPIC_BASE_URL=https://api.z.ai/api/anthropic
+# ANTHROPIC_AUTH_TOKEN=your-glm-api-key
+# ANTHROPIC_DEFAULT_OPUS_MODEL=glm-4.7
+# ANTHROPIC_DEFAULT_SONNET_MODEL=glm-4.7
+# ANTHROPIC_DEFAULT_HAIKU_MODEL=glm-4.7
+#
+# Ollama (Local):
+# ANTHROPIC_BASE_URL=http://localhost:11434
+# ANTHROPIC_DEFAULT_OPUS_MODEL=qwen3-coder
+# ANTHROPIC_DEFAULT_SONNET_MODEL=qwen3-coder
+# ANTHROPIC_DEFAULT_HAIKU_MODEL=qwen3-coder
+#
+# Kimi (Moonshot):
+# ANTHROPIC_BASE_URL=https://api.kimi.com/coding/
+# ANTHROPIC_API_KEY=your-kimi-api-key
+# ANTHROPIC_DEFAULT_OPUS_MODEL=kimi-k2.5
+# ANTHROPIC_DEFAULT_SONNET_MODEL=kimi-k2.5
+# ANTHROPIC_DEFAULT_HAIKU_MODEL=kimi-k2.5

autonomous_agent_demo.py

@@ -44,8 +44,10 @@ from dotenv import load_dotenv
 # IMPORTANT: Must be called BEFORE importing other modules that read env vars at load time
 load_dotenv()
 
+import os
+
 from agent import run_autonomous_agent
-from registry import DEFAULT_MODEL, get_project_path
+from registry import DEFAULT_MODEL, get_effective_sdk_env, get_project_path
 
 
 def parse_args() -> argparse.Namespace:
@@ -195,6 +197,14 @@ def main() -> None:
     # Note: Authentication is handled by start.bat/start.sh before this script runs.
     # The Claude SDK auto-detects credentials from ~/.claude/.credentials.json
 
+    # Apply UI-configured provider settings to this process's environment.
+    # This ensures CLI-launched agents respect Settings UI provider config (GLM, Ollama, etc.).
+    # Uses setdefault so explicit env vars / .env file take precedence.
+    sdk_overrides = get_effective_sdk_env()
+    for key, value in sdk_overrides.items():
+        if value:  # Only set non-empty values (empty values are used to clear conflicts)
+            os.environ.setdefault(key, value)
+
     # Handle deprecated --parallel flag
     if args.parallel is not None:
         print("WARNING: --parallel is deprecated. Use --concurrency instead.", flush=True)

client.py

@@ -16,7 +16,6 @@ from claude_agent_sdk import ClaudeAgentOptions, ClaudeSDKClient
 from claude_agent_sdk.types import HookContext, HookInput, HookMatcher, SyncHookJSONOutput
 from dotenv import load_dotenv
 
-from env_constants import API_ENV_VARS
 from security import SENSITIVE_DIRECTORIES, bash_security_hook
 
 # Load environment variables from .env file if present
@@ -450,14 +449,11 @@ def create_client(
         }
 
     # Build environment overrides for API endpoint configuration
-    # These override system env vars for the Claude CLI subprocess,
-    # allowing AutoForge to use alternative APIs (e.g., GLM) without
-    # affecting the user's global Claude Code settings
-    sdk_env = {}
-    for var in API_ENV_VARS:
-        value = os.getenv(var)
-        if value:
-            sdk_env[var] = value
+    # Uses get_effective_sdk_env() which reads provider settings from the database,
+    # ensuring UI-configured alternative providers (GLM, Ollama, Kimi, Custom) propagate
+    # correctly to the Claude CLI subprocess
+    from registry import get_effective_sdk_env
+    sdk_env = get_effective_sdk_env()
 
     # Detect alternative API mode (Ollama, GLM, or Vertex AI)
     base_url = sdk_env.get("ANTHROPIC_BASE_URL", "")

registry.py

@@ -731,7 +731,22 @@ def get_effective_sdk_env() -> dict[str, str]:
                 sdk_env[var] = value
         return sdk_env
 
-    sdk_env = {}
+    sdk_env: dict[str, str] = {}
+
+    # Explicitly clear credentials that could leak from the server process env.
+    # For providers using ANTHROPIC_AUTH_TOKEN (GLM, Custom), clear ANTHROPIC_API_KEY.
+    # For providers using ANTHROPIC_API_KEY (Kimi), clear ANTHROPIC_AUTH_TOKEN.
+    # This prevents the Claude CLI from using the wrong credentials.
+    auth_env_var = provider.get("auth_env_var", "ANTHROPIC_AUTH_TOKEN")
+    if auth_env_var == "ANTHROPIC_AUTH_TOKEN":
+        sdk_env["ANTHROPIC_API_KEY"] = ""
+    elif auth_env_var == "ANTHROPIC_API_KEY":
+        sdk_env["ANTHROPIC_AUTH_TOKEN"] = ""
+
+    # Clear Vertex AI vars when using non-Vertex alternative providers
+    sdk_env["CLAUDE_CODE_USE_VERTEX"] = ""
+    sdk_env["CLOUD_ML_REGION"] = ""
+    sdk_env["ANTHROPIC_VERTEX_PROJECT_ID"] = ""
 
     # Base URL
     base_url = all_settings.get("api_base_url") or provider.get("base_url")
@@ -741,7 +756,6 @@ def get_effective_sdk_env() -> dict[str, str]:
     # Auth token
     auth_token = all_settings.get("api_auth_token")
     if auth_token:
-        auth_env_var = provider.get("auth_env_var", "ANTHROPIC_AUTH_TOKEN")
         sdk_env[auth_env_var] = auth_token
 
     # Model - set all three tier overrides to the same model

server/schemas.py

@@ -190,9 +190,12 @@ class AgentStartRequest(BaseModel):
     @field_validator('model')
     @classmethod
     def validate_model(cls, v: str | None) -> str | None:
-        """Validate model is in the allowed list."""
+        """Validate model is in the allowed list (Claude) or allow any model for alternative providers."""
         if v is not None and v not in VALID_MODELS:
-            raise ValueError(f"Invalid model. Must be one of: {VALID_MODELS}")
+            from registry import get_all_settings
+            settings = get_all_settings()
+            if settings.get("api_provider", "claude") == "claude":
+                raise ValueError(f"Invalid model. Must be one of: {VALID_MODELS}")
         return v
 
     @field_validator('max_concurrency')
@@ -571,9 +574,12 @@ class ScheduleCreate(BaseModel):
     @field_validator('model')
     @classmethod
     def validate_model(cls, v: str | None) -> str | None:
-        """Validate model is in the allowed list."""
+        """Validate model is in the allowed list (Claude) or allow any model for alternative providers."""
         if v is not None and v not in VALID_MODELS:
-            raise ValueError(f"Invalid model. Must be one of: {VALID_MODELS}")
+            from registry import get_all_settings
+            settings = get_all_settings()
+            if settings.get("api_provider", "claude") == "claude":
+                raise ValueError(f"Invalid model. Must be one of: {VALID_MODELS}")
         return v
 
 
@@ -593,9 +599,12 @@ class ScheduleUpdate(BaseModel):
     @field_validator('model')
     @classmethod
     def validate_model(cls, v: str | None) -> str | None:
-        """Validate model is in the allowed list."""
+        """Validate model is in the allowed list (Claude) or allow any model for alternative providers."""
         if v is not None and v not in VALID_MODELS:
-            raise ValueError(f"Invalid model. Must be one of: {VALID_MODELS}")
+            from registry import get_all_settings
+            settings = get_all_settings()
+            if settings.get("api_provider", "claude") == "claude":
+                raise ValueError(f"Invalid model. Must be one of: {VALID_MODELS}")
         return v
 
 

start.py

@@ -390,8 +390,11 @@ def run_agent(project_name: str, project_dir: Path) -> None:
     print(f"Location: {project_dir}")
     print("-" * 50)
 
-    # Build the command - pass absolute path
-    cmd = [sys.executable, "autonomous_agent_demo.py", "--project-dir", str(project_dir.resolve())]
+    # Build the command - pass absolute path and model from settings
+    from registry import DEFAULT_MODEL, get_all_settings
+    settings = get_all_settings()
+    model = settings.get("api_model") or settings.get("model", DEFAULT_MODEL)
+    cmd = [sys.executable, "autonomous_agent_demo.py", "--project-dir", str(project_dir.resolve()), "--model", model]
 
     # Run the agent with stderr capture to detect auth errors
     # stdout goes directly to terminal for real-time output