Easy AI Assist

API keys and models

Version: 1.0.15

API keys and models

Approved Knowledge Base answers work with no key. Add a key when the chat should write replies that are not already approved. The key stays in Configuration. Your provider bills the account that owns the key.

  1. Open Components > Easy AI Assist > Configuration.
  2. Choose the provider. The address and the model fill in. Paste the key, click Connect to check that the provider accepts it, then click Save. Connect does not save the key.
OpenAI
  1. Choose OpenAI. The address becomes https://api.openai.com/v1 and the model becomes gpt-4o-mini.
  2. Open platform.openai.com/api-keys and sign in.
  3. Create a secret key and copy it. OpenAI shows the full key only once.
  4. Paste it into Secret key and click Connect. Save when it says connected.
  5. Leave the model on gpt-4o-mini. You can type another chat model from the OpenAI list if your account can use it.
Claude
  1. Choose Anthropic Claude. The model becomes claude-sonnet-5.
  2. Open Claude settings, API keys and sign in.
  3. Create a key and copy it. The key starts with sk-ant- and is shown only once.
  4. Paste it into Secret key and click Connect. Save when it says connected.
  5. Leave the model on claude-sonnet-5. For a smaller model, type claude-haiku-4-5. Any model name that starts with claude is sent to Anthropic.
Other providers

Choose the name. The address and the model below fill in. Create the key on that provider's site, paste it, and click Connect. Connected means the key works. Then save. Ollama and LM Studio have no website key. Connect checks that the program on this computer answers.

  • Google Gemini: model gemini-2.5-flash. Create the key in Google AI Studio, API keys.
  • Groq: model llama-3.3-70b-versatile. Create the key at console.groq.com/keys.
  • OpenRouter: model openai/gpt-4o-mini. Create the key at openrouter.ai/keys. Keys start with sk-or-.
  • Mistral: model mistral-small-latest. In Mistral Studio, open API keys and create a key. Steps: Mistral, create an API key.
  • DeepSeek: model deepseek-chat. Create the key at platform.deepseek.com/api_keys.
  • xAI Grok: model grok-3-mini. Create the key in the xAI console.
  • Qwen (Alibaba): model qwen-plus. In Alibaba Model Studio (international), open API keys. The filled address is the international endpoint.
  • Moonshot Kimi: model moonshot-v1-8k. Create the key at platform.moonshot.ai, API keys.
  • Together AI: model meta-llama/Llama-3.3-70B-Instruct-Turbo. Create the key at api.together.ai, API keys.
  • Fireworks: model accounts/fireworks/models/llama-v3p1-8b-instruct. Create the key at Fireworks, API keys.
  • Perplexity: model sonar. In console.perplexity.ai, open API keys and generate a key.
  • Cerebras: model llama-3.3-70b. Create the key at cloud.cerebras.ai.
  • SambaNova: model Meta-Llama-3.3-70B-Instruct. Create the key at cloud.sambanova.ai/apis.
  • NVIDIA: model meta/llama-3.1-8b-instruct. Sign in at build.nvidia.com and use Get API Key.
  • GitHub Models: model openai/gpt-4o-mini. Sign in at GitHub Models and create a token for that catalog.
  • Azure OpenAI: model gpt-4o-mini, or the deployment name in your Azure resource. Replace YOUR-RESOURCE in the address with that resource name. Copy the key from the Azure portal for the same resource.
  • Ollama on this computer: model llama3.2. No website key. The preset writes ollama into Secret key. Ollama must be running on the same server as the site, with that model pulled.
  • LM Studio on this computer: the preset writes lmstudio into Secret key. Change the model to the name loaded in LM Studio. The local server must be running.