diff --git a/build_with_arch/agent.html b/build_with_arch/agent.html index cfaf6dd2..7619585e 100755 --- a/build_with_arch/agent.html +++ b/build_with_arch/agent.html @@ -418,7 +418,7 @@ the user’s intent.

diff --git a/build_with_arch/multi_turn.html b/build_with_arch/multi_turn.html index 57dce2c1..826302a7 100755 --- a/build_with_arch/multi_turn.html +++ b/build_with_arch/multi_turn.html @@ -354,7 +354,7 @@ that you can test and modify locally for multi-turn RAG scenarios.

diff --git a/build_with_arch/rag.html b/build_with_arch/rag.html index b6c04ac1..857e589d 100755 --- a/build_with_arch/rag.html +++ b/build_with_arch/rag.html @@ -302,7 +302,7 @@ minutes. For additional details on how to build multi-turn RAG applications plea diff --git a/concepts/llm_providers/client_libraries.html b/concepts/llm_providers/client_libraries.html index 79f8d2a4..a065a018 100755 --- a/concepts/llm_providers/client_libraries.html +++ b/concepts/llm_providers/client_libraries.html @@ -584,7 +584,7 @@ Implement fallback logic for better reliability:

diff --git a/concepts/llm_providers/llm_providers.html b/concepts/llm_providers/llm_providers.html index cb4f0300..8ea02649 100755 --- a/concepts/llm_providers/llm_providers.html +++ b/concepts/llm_providers/llm_providers.html @@ -238,6 +238,7 @@ Use your preferred client library without changing existing code (see Supported API Endpoints
  • First-Class Providers
  • Providers Requiring Base URL
    • +
  • Base URL Configuration
  • Advanced Configuration
  • Model Selection Guidelines
  • See Also
    • @@ -301,7 +302,7 @@ Use your preferred client library without changing existing code (see
    -

    © 2025, Katanemo Labs, Inc Last updated: Oct 28, 2025. 

    +

    © 2025, Katanemo Labs, Inc Last updated: Oct 30, 2025. 

    diff --git a/concepts/llm_providers/model_aliases.html b/concepts/llm_providers/model_aliases.html index c97391b8..1292fa96 100755 --- a/concepts/llm_providers/model_aliases.html +++ b/concepts/llm_providers/model_aliases.html @@ -435,7 +435,7 @@ diff --git a/concepts/llm_providers/supported_providers.html b/concepts/llm_providers/supported_providers.html index 5fead73a..8560dc5d 100755 --- a/concepts/llm_providers/supported_providers.html +++ b/concepts/llm_providers/supported_providers.html @@ -194,7 +194,7 @@

  • access_key: API key for authentication (supports environment variables)

  • default: Mark a model as the default (optional, boolean)

  • name: Custom name for the provider instance (optional)

    • -

  • base_url: Custom endpoint URL (required for some providers)

    • +

  • base_url: Custom endpoint URL (required for some providers, optional for others - see Base URL Configuration)

    • @@ -700,6 +700,7 @@ Any provider that implements the OpenAI API interface can be configured using cu

    Providers Requiring Base URL

    +

    The following providers require a base_url parameter to be configured. For detailed information on base URL configuration including path prefix behavior and examples, see Base URL Configuration.

    Azure OpenAI

    Provider Prefix: azure_openai/

    @@ -804,6 +805,63 @@ Any provider that implements the OpenAI API interface can be configured using cu
    +
    +

    Base URL Configuration

    +

    The base_url parameter allows you to specify custom endpoints for model providers. It supports both hostname and path components, enabling flexible routing to different API endpoints.

    +

    Format: <scheme>://<hostname>[:<port>][/<path>]

    +

    Components:

    + +

    How Path Prefixes Work:

    +

    When you include a path in base_url, it replaces the provider’s default path prefix while preserving the endpoint suffix:

    + +

    Configuration Examples:

    +
    llm_providers:
+  # Simple hostname only - uses provider's default path
+  - model: zhipu/glm-4.6
+    access_key: $ZHIPU_API_KEY
+    base_url: https://api.z.ai
+    # Results in: https://api.z.ai/api/paas/v4/chat/completions
+
+  # With custom path prefix - replaces provider's default path
+  - model: zhipu/glm-4.6
+    access_key: $ZHIPU_API_KEY
+    base_url: https://api.z.ai/api/coding/paas/v4
+    # Results in: https://api.z.ai/api/coding/paas/v4/chat/completions
+
+  # Azure with custom path
+  - model: azure_openai/gpt-4
+    access_key: $AZURE_API_KEY
+    base_url: https://mycompany.openai.azure.com/custom/deployment/path
+    # Results in: https://mycompany.openai.azure.com/custom/deployment/path/chat/completions
+
+  # Behind a proxy or API gateway
+  - model: openai/gpt-4o
+    access_key: $OPENAI_API_KEY
+    base_url: https://proxy.company.com/ai-gateway/openai
+    # Results in: https://proxy.company.com/ai-gateway/openai/chat/completions
+
+  # Local endpoint with custom port
+  - model: ollama/llama3.1
+    base_url: http://localhost:8080
+    # Results in: http://localhost:8080/v1/chat/completions
+
+  # Custom provider with path prefix
+  - model: vllm/custom-model
+    access_key: $VLLM_API_KEY
+    base_url: https://vllm.example.com/models/v2
+    provider_interface: openai
+    # Results in: https://vllm.example.com/models/v2/chat/completions
+
    +
    +

    Advanced Configuration

    @@ -920,6 +978,7 @@ Any provider that implements the OpenAI API interface can be configured using cu
  • OpenAI-Compatible Providers
    • +
  • Base URL Configuration
  • Advanced Configuration