NeuroLink API Reference / GenerateOptions

Type Alias: GenerateOptions

GenerateOptions = object

Generate function options type - Primary method for content generation Supports multimodal content while maintaining backward compatibility

Properties

input

input: object

Defined in: types/generate.ts:37

text

text: string

images?

optional images?: (Buffer | string | ImageWithAltText)[]

Images to include in the request. Supports simple image data (Buffer, string) or objects with alt text for accessibility.

Examples

images: [imageBuffer, "https://example.com/image.jpg"];

images: [
  { data: imageBuffer, altText: "Product screenshot showing main dashboard" },
  { data: "https://example.com/chart.png", altText: "Sales chart for Q3 2024" },
];

csvFiles?

optional csvFiles?: (Buffer | string)[]

pdfFiles?

optional pdfFiles?: (Buffer | string)[]

videoFiles?

optional videoFiles?: (Buffer | string)[]

files?

optional files?: (Buffer | string | FileWithMetadata)[]

content?

optional content?: Content[]

segments?

optional segments?: DirectorSegment[]

Director Mode segments. When provided, Director Mode is activated automatically. Each segment contains its own prompt and image. Must contain 2-10 segments.

output?

optional output?: object

Defined in: types/generate.ts:91

Output configuration options

format?

optional format?: "text" | "structured" | "json"

Output format for text generation

mode?

optional mode?: "text" | "video" | "ppt"

Output mode - determines the type of content generated

"text": Standard text generation (default)
"video": Video generation using models like Veo 3.1
"ppt": PowerPoint presentation generation

video?

optional video?: VideoOutputOptions

Video generation configuration (used when mode is "video") Requires an input image and text prompt

ppt?

optional ppt?: PPTOutputOptions

PowerPoint generation configuration (used when mode is "ppt") Generates slides based on text prompt

director?

optional director?: DirectorModeOptions

Director Mode configuration (only used when input.segments is provided) Controls transition prompts, durations, and concurrency.

Examples

output: {
  format: "text";
}

output: {
  mode: "video",
  video: {
    resolution: "1080p",
    length: 8,
    aspectRatio: "16:9",
    audio: true
  }
}

csvOptions?

optional csvOptions?: object

Defined in: types/generate.ts:119

maxRows?

optional maxRows?: number

formatStyle?

optional formatStyle?: "raw" | "markdown" | "json"

includeHeaders?

optional includeHeaders?: boolean

videoOptions?

optional videoOptions?: object

Defined in: types/generate.ts:126

frames?

optional frames?: number

quality?

optional quality?: number

format?

optional format?: "jpeg" | "png"

transcribeAudio?

optional transcribeAudio?: boolean

tts?

optional tts?: TTSOptions

Defined in: types/generate.ts:165

Text-to-Speech (TTS) configuration

Enable audio generation from the text response. The generated audio will be returned in the result's audio field as a TTSResult object.

Examples

const result = await neurolink.generate({
  input: { text: "Tell me a story" },
  provider: "google-ai",
  tts: { enabled: true, voice: "en-US-Neural2-C" },
});
console.log(result.audio?.buffer); // Audio Buffer

const result = await neurolink.generate({
  input: { text: "Speak slowly and clearly" },
  provider: "google-ai",
  tts: {
    enabled: true,
    voice: "en-US-Neural2-D",
    speed: 0.8,
    pitch: 2.0,
    format: "mp3",
    quality: "standard",
  },
});

stt?

optional stt?: STTOptions & object

Defined in: types/generate.ts:184

Speech-to-Text (STT) configuration

Enable audio transcription. When enabled, the audio provided via stt.audio will be transcribed to text and used as the prompt.

Type Declaration

provider?

optional provider?: string

audio?

optional audio?: Buffer | ArrayBuffer

Example

const neurolink = new NeuroLink();
const result = await neurolink.generate({
  input: { text: "" },
  provider: "openai",
  stt: {
    enabled: true,
    provider: "whisper",
    language: "en-US",
    audio: audioBuffer,
  },
});
// STT transcribes the audio, result.transcription contains the transcription

thinkingConfig?

optional thinkingConfig?: object

Defined in: types/generate.ts:226

Thinking/reasoning configuration for extended thinking models

Enables extended thinking capabilities for supported models.

Gemini 3 Models (gemini-3.1-pro-preview, gemini-3-flash-preview): Use thinkingLevel to control reasoning depth:

minimal - Near-zero thinking (Flash only)
low - Fast reasoning for simple tasks
medium - Balanced reasoning/latency
high - Maximum reasoning depth (default for Pro)

Anthropic Claude (claude-3-7-sonnet, etc.): Use budgetTokens to set token budget for thinking.

enabled?

optional enabled?: boolean

type?

optional type?: "enabled" | "disabled"

budgetTokens?

optional budgetTokens?: number

Token budget for thinking (Anthropic models)

thinkingLevel?

optional thinkingLevel?: "minimal" | "low" | "medium" | "high"

Thinking level for Gemini 3 models: minimal, low, medium, high

Examples

const result = await neurolink.generate({
  input: { text: "Solve this complex problem..." },
  provider: "google-ai",
  model: "gemini-3.1-pro-preview",
  thinkingConfig: {
    thinkingLevel: "high",
  },
});

const result = await neurolink.generate({
  input: { text: "Solve this complex math problem..." },
  provider: "anthropic",
  model: "claude-3-7-sonnet-20250219",
  thinkingConfig: {
    enabled: true,
    budgetTokens: 10000,
  },
});

provider?

optional provider?: AIProviderName | string

Defined in: types/generate.ts:236

model?

optional model?: string

Defined in: types/generate.ts:237

region?

optional region?: string

Defined in: types/generate.ts:238

temperature?

optional temperature?: number

Defined in: types/generate.ts:239

maxTokens?

optional maxTokens?: number

Defined in: types/generate.ts:240

systemPrompt?

optional systemPrompt?: string

Defined in: types/generate.ts:241

schema?

optional schema?: ValidationSchema

Defined in: types/generate.ts:274

Zod schema for structured output validation

Important

Google Gemini Limitation Google Vertex AI and Google AI Studio cannot combine function calling with structured output. You MUST use disableTools: true when using schemas with Google providers.

Error without disableTools: "Function calling with a response mime type: 'application/json' is unsupported"

This is a documented Google API limitation, not a NeuroLink bug. All frameworks (LangChain, Vercel AI SDK, Agno, Instructor) use this approach.

Example

// ✅ Correct for Google providers
const result = await neurolink.generate({
  schema: MySchema,
  provider: "vertex",
  disableTools: true, // Required for Google
});

// ✅ No restriction for other providers
const result = await neurolink.generate({
  schema: MySchema,
  provider: "openai", // Works without disableTools
});

See

https://ai.google.dev/gemini-api/docs/function-calling

tools?

optional tools?: Record<string, Tool>

Defined in: types/generate.ts:275

enabledToolNames?

optional enabledToolNames?: string[]

Defined in: types/generate.ts:289

Filter available tools by name. Only tools with names in this array will be made available. Used by dynamic arguments to dynamically select which tools to enable.

Example

await neurolink.generate({
  input: { text: "Search for information" },
  enabledToolNames: ["websearchGrounding", "readFile"],
});

timeout?

optional timeout?: number | string

Defined in: types/generate.ts:290

abortSignal?

optional abortSignal?: AbortSignal

Defined in: types/generate.ts:292

AbortSignal for external cancellation of the AI call

disableTools?

optional disableTools?: boolean

Defined in: types/generate.ts:310

Disable tool execution (including built-in tools)

Required

For Google Gemini providers when using schemas Google Vertex AI and Google AI Studio require this flag when using structured output (schemas) due to Google API limitations.

Example

// Required for Google providers with schemas
await neurolink.generate({
  schema: MySchema,
  provider: "vertex",
  disableTools: true,
});

toolFilter?

optional toolFilter?: string[]

Defined in: types/generate.ts:313

Include only these tools by name (whitelist). If set, only matching tools are available.

excludeTools?

optional excludeTools?: string[]

Defined in: types/generate.ts:316

Exclude these tools by name (blacklist). Applied after toolFilter.

skipToolPromptInjection?

optional skipToolPromptInjection?: boolean

Defined in: types/generate.ts:324

Skip injecting tool schemas into the system prompt. When true, tools are ONLY passed natively via the provider's tools parameter, avoiding duplicate tool definitions (~30K tokens savings per call). Default: false (backward compatible — tool schemas are injected into system prompt).

disableToolCache?

optional disableToolCache?: boolean

Defined in: types/generate.ts:327

Disable tool result caching for this request (overrides global mcp.cache.enabled)

maxSteps?

optional maxSteps?: number

Defined in: types/generate.ts:330

Maximum number of tool execution steps (default: 200)

toolChoice?

optional toolChoice?: ToolChoice<Record<string, Tool>>

Defined in: types/generate.ts:345

Tool choice configuration for the generation. Controls whether and which tools the model must call.

"auto" (default): the model can choose whether and which tools to call
"none": no tool calls allowed
"required": the model must call at least one tool
{ type: "tool", toolName: string }: the model must call the specified tool

Note: When used without prepareStep, this applies to every step in the maxSteps loop. Using "required" or { type: "tool" } without prepareStep will cause infinite tool calls until maxSteps is exhausted.

prepareStep?

optional prepareStep?: (options) => PromiseLike<{ model?: LanguageModel; toolChoice?: ToolChoice<Record<string, Tool>>; experimental_activeTools?: string[]; } | undefined>

Defined in: types/generate.ts:370

Optional callback that runs before each step in a multi-step generation. Allows dynamically changing toolChoice and available tools per step.

This is the recommended way to enforce specific tool calls on certain steps while allowing the model freedom on others.

Maps to Vercel AI SDK's experimental_prepareStep.

Parameters

options

steps

StepResult<Record<string, Tool>>[]

stepNumber

number

maxSteps

number

model

LanguageModel

Returns

PromiseLike<{ model?: LanguageModel; toolChoice?: ToolChoice<Record<string, Tool>>; experimental_activeTools?: string[]; } | undefined>

Example

prepareStep: ({ stepNumber, steps }) => {
  if (stepNumber === 0) {
    return {
      toolChoice: { type: "tool", toolName: "myTool" },
    };
  }
  return { toolChoice: "auto" };
};

See

https://ai-sdk.dev/docs/reference/ai-sdk-core/generate-text#parameters

enableEvaluation?

optional enableEvaluation?: boolean

Defined in: types/generate.ts:385

enableAnalytics?

optional enableAnalytics?: boolean

Defined in: types/generate.ts:386

context?

optional context?: StandardRecord

Defined in: types/generate.ts:387

evaluationDomain?

optional evaluationDomain?: string

Defined in: types/generate.ts:390

toolUsageContext?

optional toolUsageContext?: string

Defined in: types/generate.ts:391

conversationHistory?

optional conversationHistory?: object[]

Defined in: types/generate.ts:398

role

role: string

content

content: string

Deprecated

Use conversationMessages instead. This field uses a simple {role, content} shape that is not consumed by buildMessagesArray() — messages passed here will NOT reach the AI model as proper conversation turns. conversationMessages uses the full ChatMessage type and is correctly wired through the entire generate pipeline.

conversationMessages?

optional conversationMessages?: ChatMessage[]

Defined in: types/generate.ts:406

Previous conversation as a ChatMessage array. Messages are injected as proper multi-turn conversation history before the current prompt, so the AI model sees them as real prior exchanges (not text dumped into the prompt). Used by task continuation mode and available to external callers.

factoryConfig?

optional factoryConfig?: object

Defined in: types/generate.ts:409

domainType?

optional domainType?: string

domainConfig?

optional domainConfig?: StandardRecord

enhancementType?

optional enhancementType?: "domain-configuration" | "streaming-optimization" | "mcp-integration" | "legacy-migration" | "context-conversion"

preserveLegacyFields?

optional preserveLegacyFields?: boolean

validateDomainData?

optional validateDomainData?: boolean

streaming?

optional streaming?: object

Defined in: types/generate.ts:423

enabled?

optional enabled?: boolean

chunkSize?

optional chunkSize?: number

bufferSize?

optional bufferSize?: number

enableProgress?

optional enableProgress?: boolean

fallbackToGenerate?

optional fallbackToGenerate?: boolean

workflow?

optional workflow?: string

Defined in: types/generate.ts:432

workflowConfig?

optional workflowConfig?: WorkflowConfig

Defined in: types/generate.ts:433

rag?

optional rag?: RAGConfig

Defined in: types/generate.ts:467

RAG (Retrieval-Augmented Generation) configuration.

When provided, NeuroLink automatically loads the specified files, chunks them, generates embeddings, and creates a search tool that the AI model can invoke on demand to find relevant context before answering.

Examples

const result = await neurolink.generate({
  input: { text: "What is RAG?" },
  provider: "vertex",
  rag: {
    files: ["./docs/guide.md"],
  },
});

const result = await neurolink.generate({
  input: { text: "Explain chunking strategies" },
  provider: "vertex",
  rag: {
    files: ["./docs/guide.md", "./docs/api.md"],
    strategy: "markdown",
    chunkSize: 512,
    topK: 5,
  },
});

maxBudgetUsd?

optional maxBudgetUsd?: number

Defined in: types/generate.ts:482

Maximum budget in USD for this session. When the accumulated cost of all generate() calls on this NeuroLink instance exceeds this value, subsequent calls will throw a budget-exceeded error before making the API request.

Example

const result = await neurolink.generate({
  input: { text: "Summarize this" },
  maxBudgetUsd: 1.0,
});

requestId?

optional requestId?: string

Defined in: types/generate.ts:489

Optional request identifier for observability and log correlation. When provided, this ID is forwarded to spans, logs, and telemetry so callers can correlate generation traces back to their own request lifecycle.

middleware?

optional middleware?: MiddlewareFactoryOptions

Defined in: types/generate.ts:504

Per-call middleware configuration.

onFinish?

optional onFinish?: OnFinishCallback

Defined in: types/generate.ts:507

Callback invoked when generation completes successfully.

onError?

optional onError?: OnErrorCallback

Defined in: types/generate.ts:510

Callback invoked when generation encounters an error.

requestContext?

optional requestContext?: Record<string, unknown>

Defined in: types/generate.ts:513

Pre-validated user context for the request

auth?

optional auth?: object

Defined in: types/generate.ts:516

Raw auth token — validated by configured auth provider

token

token: string

credentials?

optional credentials?: NeurolinkCredentials

Defined in: types/generate.ts:523

Per-provider credential overrides for this request. Overrides instance-level credentials set in new NeuroLink({ credentials }). Unset providers fall through to instance credentials, then environment variables.

providerFallback?

optional providerFallback?: (error) => Promise<{ provider?: string; model?: string; } | null>

Defined in: types/generate.ts:529

Curator P2-3: per-call fallback callback. Overrides any instance-level providerFallback set on new NeuroLink({...}).

Parameters

error

unknown

Returns

Promise<{ provider?: string; model?: string; } | null>

modelChain?

optional modelChain?: string[]

Defined in: types/generate.ts:537

Curator P2-3: per-call ordered model chain. Overrides any instance-level modelChain. Tried in order on model-access-denied.

memory?

optional memory?: object

Defined in: types/generate.ts:547

Per-call memory control.

Override the global memory SDK behavior for this specific call. All flags default to true when the global memory SDK is enabled. If the global memory SDK is disabled, these flags have no effect.

enabled?

optional enabled?: boolean

Master toggle for this call. When false, both read and write are skipped. Defaults to true.

read?

optional read?: boolean

Whether to read condensed memory and prepend to prompt. Defaults to true.

write?

optional write?: boolean

Whether to write (add/condense) the conversation into memory after completion. Defaults to true.

additionalUsers?

optional additionalUsers?: AdditionalMemoryUser[]

Additional users whose memory should be retrieved/stored alongside the primary user. Each entry can override the condensation prompt and maxWords for that user. Primary user is still determined by context.userId.

FilesExpand file tree

GenerateOptions.md

Latest commit

History

GenerateOptions.md

File metadata and controls

Type Alias: GenerateOptions

Properties

input

text

images?

Examples

csvFiles?

pdfFiles?

videoFiles?

files?

content?

segments?

output?

format?

mode?

video?

ppt?

director?

Examples

csvOptions?

maxRows?

formatStyle?

includeHeaders?

videoOptions?

frames?

quality?

format?

transcribeAudio?

tts?

Examples

stt?

Type Declaration

provider?

audio?

Example

thinkingConfig?

enabled?

type?

budgetTokens?

thinkingLevel?

Examples

provider?

model?

region?

temperature?

maxTokens?

systemPrompt?

schema?

Important

Example

See

tools?

enabledToolNames?

Example

timeout?

abortSignal?

disableTools?

Required

Example

toolFilter?

excludeTools?

skipToolPromptInjection?

disableToolCache?

maxSteps?

toolChoice?

prepareStep?

Parameters

options

steps

stepNumber

maxSteps

model

Returns

Example