Anthropic
Integrating Anthropic models with Lakehousecat provides advanced AI capabilities for Chat interactions. This page describes how to configure Anthropic as a Provider Model.
Prerequisites
- Access to Lakehousecat as an Administrator
- An Anthropic API Key
Configuration
Navigate to Workspace → Models → Provider Models and click the + icon.
Step 1: Select Provider Type
From the Provider Type dropdown, select Anthropic.
Step 2: Select Model Type
Select Chat as the Model Type.
Step 3: Enter Configuration Details
| Field | Description |
|---|---|
| Configuration Name | A unique name for this configuration (e.g. anthropic-claude-chat). |
| API Key | Your Anthropic API key. Required for authentication. |
| Model ID | The Anthropic model identifier (e.g. claude-3-7-sonnet-20250219). Select from the dropdown or enter manually. |
Model ID accepts any Anthropic model identifier as free text — the table below lists models explicitly verified by the Lakehousecat team.
Step 4: Save
Click Save. After saving, you can optionally set this configuration as the default UI model or default Backend model.
Tested Configurations
Stand: 2026-07-20. This reflects what Lakehousecat has actually run and observed, not a benchmark or performance ranking — see There is no single "right" provider for why we don't publish comparative scores.
| Model ID | Chat | Chart Generation | Notes |
|---|---|---|---|
claude-haiku-4-5-20251001 | ✅ | ✅ (3/3 charts, 0 failures) | Fastest and cheapest of the tested models; ~$0.15/chart, ~$0.068/chart with prompt caching enabled |
claude-sonnet-4-6 | ✅ | ✅ (1/1 charts) | ~$0.34/chart — same chart output as Haiku for this structured task, at higher cost |
claude-opus-4-6 | ✅ | not yet tested for chart generation | Chat-verified only |
claude-opus-4-8 | ✅ | ✅ (1/1 charts) | ~$0.73/chart — most expensive of the tested models; no functional advantage observed over Haiku for chart generation |
Earlier model generations (Claude 3.x) are not part of this test round and are not listed here; they are expected to remain compatible but have not been re-verified recently.
Prompt Caching
For Anthropic models, Lakehousecat applies prompt caching automatically — there is nothing to configure. It marks the static part of each request (the system prompt, and in multi-step runs the conversation so far) as cacheable, so repeated steps re-read it at a fraction of the normal input price. This applies to the multi-step runs behind chart generation, voice-input optimization and semantic extraction, and noticeably lowers their input cost.
Other providers (OpenAI, Azure, Google) cache on their side without any marker from Lakehousecat; this setting does not affect them.
Emergency stop. If caching ever costs more than it saves — for example after a change on Anthropic's side — an administrator can switch it off in the instance's Custom Resource, without an image update:
spec:
features:
anthropicPromptCacheEnabled: false
Leave the field unset in normal operation; the default is true.
Best Practices
- Use a descriptive configuration name that includes the model name for easy identification.
- Keep your API key updated when rotating credentials.
- For chart generation, Claude Haiku is the recommended starting point — it is the fastest and cheapest of the tested models and produced the same chart output as Sonnet and Opus in testing. Move to Sonnet or Opus only if you have a specific reason (e.g. heavier reasoning needs elsewhere in your workflow). See Choosing a Model for Chart Generation.
Troubleshooting
- Validation Errors: Verify the API Key and Model ID for accuracy.
- Configuration Issues: Check the Anthropic documentation for valid model identifiers.