Skip to main content

LLM Providers

Configure which AI model powers the Lens chat assistant. LLM Providers is a per-user, bring-your-own-key (BYOK) setting: you add your own API key for a provider, and that configuration is used across every site you have access to.

Per-user, not per-account

Your LLM Providers configuration belongs to you, not to a site or organization. The same key is applied to every site you can access, and only you can see or modify it. There is no per-account dropdown — other users configure their own keys independently.

Accessing LLM Providers​

In the sidebar, open My Account → LLM Providers (/account/llm).

Not the same as LENS Settings

LLM Providers (My Account) controls which AI model the Lens chat uses, per user.

LENS Settings (at /settings/lens) is a separate, per-site setting that controls the automated insights analysis frequency (daily / weekly / manual) and email notifications. Don't confuse the two.

Available Providers​

LLM Providers is BYOK only — every provider requires your own API key. The self-hosted DeepSeek Local model used by shared infrastructure is not available here for the chat.

ProviderAPI Key RequiredNotes
Anthropic (Claude)YesRecommended
OpenAI (GPT)YesGeneral-purpose
DeepSeek CloudYesOptional custom base URL
Google (Gemini)YesGemini models

Anthropic is the recommended default provider.

Adding a Provider​

Step 1: Open the form​

If you have no providers configured yet, the Configure LLM Provider form appears automatically. Otherwise, click Add Provider.

Step 2: Select a provider​

Choose a provider from the dropdown. Each provider shows a short description and a link to obtain an API key. The provider can't be changed after a configuration is created — to switch providers, add a new one or remove the existing one.

Step 3: Enter your API key​

All providers here require an API key. Paste it into the API Key field (use the eye toggle to reveal it). Your key is encrypted before storage and only the last 4 characters are ever shown afterwards (****XXXX).

For DeepSeek Cloud, you can optionally set a custom Base URL (defaults to https://api.deepseek.com/v1).

Step 4: Configure model overrides (optional)​

You can override the default model used at three complexity levels. Leave a field blank to use the provider's default.

LevelAnthropic defaultOpenAI defaultGemini defaultDeepSeek Cloud default
Simpleclaude-3-5-haiku-20241022gpt-4o-minigemini-2.5-flashdeepseek-chat
Complexclaude-sonnet-4-20250514gpt-4ogemini-2.5-flashdeepseek-chat
Criticalclaude-opus-4-20250514gpt-4ogemini-2.5-prodeepseek-chat

Use Reset to defaults to clear all overrides.

Step 5: Test the connection​

Click Test Connection to verify the key and provider are reachable before saving. A successful test shows a "Connection Successful" message with the response latency in milliseconds.

Step 6: Save​

Click Save Settings to store the configuration.

Managing Your Configurations​

Each configured provider appears as a card showing:

  • Provider name and whether it's the default
  • API key status (****XXXX when configured)
  • Enabled / disabled status
  • Any model overrides
  • Last updated timestamp

From a card you can:

  • Set as Default — make this provider the one Lens chat uses (only one default at a time)
  • Edit — update the API key, model overrides, base URL, or enabled status
  • Remove — delete the configuration (if it was the default, another configured provider becomes the default)

Enable / disable​

Use the Enable Configuration toggle when editing a provider. When a configuration is disabled, that provider can't be used by Lens chat, but the saved key is kept so you can re-enable it later.

Security​

API Key Storage​

  • API keys are encrypted at rest before storage
  • Only the last 4 characters are displayed (****XXXX)
  • Keys are never returned in API responses or logged

Data Privacy​

When you use a cloud provider, aggregated analytics data (not raw events, no personally identifiable information) is sent to that provider to generate chat responses. Because this is BYOK, the data is processed under your own provider account and API key.

Troubleshooting​

"Connection Failed" on test​

  1. Verify your API key is correct and active
  2. Check the provider's status page for outages
  3. Ensure your network allows outbound API calls
  4. Try a different provider

"API key is required" when saving​

Every provider here requires a key. Get one from:

Last updated on