Document Ask Fern LLM provider options

fern/products/ask-fern/ask-fern.yml

@@ -12,6 +12,8 @@ navigation:
       - page: Setup
         path: ./pages/configuration/setup.mdx
         slug: setup
+      - page: LLM providers
+        path: ./pages/configuration/llm-providers.mdx
       - page: Guidance
         path: ./pages/features/guidance.mdx
       - page: Additional content sources

fern/products/ask-fern/pages/configuration/llm-providers.mdx

@@ -0,0 +1,57 @@
+---
+title: LLM providers
+description: Configure which language model provider Ask Fern uses to generate responses.
+---
+
+Ask Fern supports multiple language model (LLM) providers for generating AI responses. The system automatically selects an available provider based on your configuration and falls back to alternatives if the primary provider is unavailable.
+
+## Supported providers
+
+Ask Fern supports the following LLM providers:
+
+### Anthropic
+
+Direct integration with Anthropic's Claude models via the Anthropic API. This is the default provider when configured.
+
+**Supported models:**
+- Claude 3.7
+- Claude 4 Sonnet
+- Claude 4.5 Sonnet
+- Claude 4.5 Haiku
+
+**Configuration:** Requires `ANTHROPIC_API_KEY` environment variable.
+
+### AWS Bedrock
+
+Access Claude models through AWS Bedrock, which provides enterprise features like VPC endpoints, CloudWatch logging, and AWS IAM integration.
+
+**Supported models:**
+- Claude 3.7 (via Bedrock)
+- Claude 4 Sonnet (via Bedrock)
+- Claude 4.5 Sonnet (via Bedrock)
+- Claude 4.5 Haiku (via Bedrock)
+
+**Configuration:** Requires AWS credentials with Bedrock access permissions.
+
+### Cohere
+
+Integration with Cohere's Command models for AI-powered responses.
+
+**Supported models:**
+- Command A 03 2025
+
+**Configuration:** Requires `COHERE_API_KEY` environment variable.
+
+## Provider selection
+
+Ask Fern uses a fallback system to ensure high availability:
+
+1. The system attempts to use providers in order of preference: Bedrock, Anthropic, Cohere
+2. If a provider is unavailable (missing credentials or API errors), the system automatically falls back to the next available provider
+3. Within each provider, Ask Fern attempts multiple models in order of capability, falling back to smaller models if needed
+
+This fallback system ensures Ask Fern remains available even if individual providers experience issues.
+
+## Configuration
+
+LLM providers are configured at the infrastructure level through environment variables. Contact your Fern account team to configure or change your LLM provider preferences.