neutron AI
Providers

OpenAI-Compatible Providers

Use OpenAI-compatible chat adapters when a provider accepts messages with system, user, and assistant roles.

Supported adapter IDs:

  • mistral
  • xai
  • deepseek
  • perplexity
  • openrouter

Shared Pattern

const { request } = await connector.createProviderContextRequest(
  "mistral",
  {
    task: "Use account memory to draft a concise technical answer.",
    tokenBudget: 1400
  },
  {
    model: "provider-model-id",
    input: "What should we tell the customer?"
  }
);

// Send `request` to the provider's OpenAI-compatible chat endpoint.

The SDK prepends a system message containing Neutron memory context.

Provider Examples

await connector.createProviderContextRequest("xai", { task }, { model: "grok-model-id", input });
await connector.createProviderContextRequest("deepseek", { task }, { model: "deepseek-model-id", input });
await connector.createProviderContextRequest("perplexity", { task }, { model: "sonar-model-id", input });
await connector.createProviderContextRequest("openrouter", { task }, { model: "openrouter/model-id", input });

Remote MCP support varies for OpenAI-compatible providers. When native remote MCP is not available, call Neutron from your backend first and inject the returned context into the provider request with the SDK adapter.

Google Gemini and Vertex AI

Groq