SDK Reference

The free Decision Memos SDK. Query multiple AI models in parallel with typed responses — one call, four models, zero boilerplate.

Note

The free SDK provides raw multi-model querying. For structured verdicts with consensus scoring, advisor personas, and Decision Memos, use the hosted API.

Installation

npm install decisionmemos

Requires Node.js 18+ and TypeScript 5+.

createMultiModelQuery(options?)

Factory function that creates a MultiModelQuery instance configured with your API keys.

import { createMultiModelQuery } from 'decisionmemos';

// Reads API keys from process.env by default
const query = createMultiModelQuery();

// Or pass keys explicitly
const query = createMultiModelQuery({
  keys: {
    openai: 'sk-...',
    anthropic: 'sk-ant-...',
    xai: 'xai-...',
    google: 'AIza...',
  },
});

Options

Property	Type	Default	Description
keys	object	process.env	Provider API keys
clients	AIModelClient[]	auto-detected	Custom model client instances

query.ask(question, systemPrompt?)

Query all configured models in parallel with the same question. Returns typed responses from every model.

const result = await query.ask(
  "Should we adopt Kubernetes for our staging environment?"
);

// result.responses — ModelResponse[] (one per model)
// result.successCount — how many responded successfully
// result.errorCount — how many failed
// result.totalLatency — wall-clock time (ms)

Parameters

Parameter	Type	Description
question	string	The question to send to all models
systemPrompt	string?	Optional system-level instructions sent to every model

Return value: MultiModelResult

interface MultiModelResult {
  question: string;
  responses: ModelResponse[];
  successCount: number;
  errorCount: number;
  totalLatency: number; // ms
  timestamp: Date;
}

interface ModelResponse {
  modelName: string;   // e.g. "GPT-5.2"
  provider: string;    // e.g. "openai"
  response: string;    // The model's full response
  timestamp: Date;
  tokensUsed?: number;
  latency: number;     // ms
  error?: string;      // Set if this model failed
}

query.testConnections()

Test connectivity to all configured providers.

const results = await query.testConnections();
// [
//   { model: "GPT-5.2", provider: "openai", ok: true },
//   { model: "Claude Opus 4.6", provider: "anthropic", ok: true },
//   ...
// ]

query.getStatus()

Get the number and names of configured models.

const status = query.getStatus();
// { count: 4, models: [{ name: "GPT-5.2", provider: "openai" }, ...] }

Using individual clients

You can also import and use model clients directly:

import { OpenAIClient, AnthropicClient } from 'decisionmemos';

const openai = new OpenAIClient('sk-...', 'gpt-5.2');
const anthropic = new AnthropicClient('sk-ant-...', 'claude-sonnet-4-6');

const [gptResponse, claudeResponse] = await Promise.all([
  openai.query("Should we use GraphQL?"),
  anthropic.query("Should we use GraphQL?"),
]);

Tip

Want structured verdicts with consensus scoring instead of raw responses? See the POST /v1/deliberate API reference. BYOK gets you 20% off per key (all 4 = 80% off).