novelai-sdk-unofficial / OpenAIGeneration

Class: OpenAIGeneration

High-level OpenAI-compatible generation interface

Provides user-friendly methods for OpenAI-compatible text generation with automatic parameter validation and camelCase to snake_case conversion.

Example

typescript

const client = new NovelAI({ apiKey: 'your-api-key' });

// Text completion
const response = await client.openai.completion({
  prompt: 'Once upon a time',
  maxTokens: 100,
});

// Chat completion
const chat = await client.openai.chatCompletion({
  messages: [{ role: 'user', content: 'Hello!' }],
  maxTokens: 100,
});

// Streaming
for await (const chunk of client.openai.chatCompletionStream({
  messages: [{ role: 'user', content: 'Tell me a story' }],
})) {
  const content = chunk.choices[0]?.delta?.content;
  if (content) process.stdout.write(content);
}

Constructors

Constructor

new OpenAIGeneration(client): OpenAIGeneration

Defined in: src/client.ts:490

Parameters

client

NovelAI

Returns

OpenAIGeneration

Methods

chatCompletion()

chatCompletion(params): Promise<OAIChatCompletionResponse>

Defined in: src/client.ts:576

Chat completion with validation (OpenAI-compatible)

Parameters

params

User-friendly chat completion parameters (camelCase)

bestOf?

number = ...

Number of completions to generate server-side and return the best

echo?

boolean = ...

Echo back the prompt in addition to the completion

enableThinking?

boolean = ...

Enable thinking/reasoning mode

frequencyPenalty?

number = ...

Frequency penalty (-2.0 to 2.0)

logitBias?

Record<string, number> = ...

Logit bias for specific tokens

logprobs?

number = ...

Number of log probabilities to return

maxTokens?

number = ...

Maximum tokens to generate (1-2048)

messages

object[] = ...

Array of chat messages

minP?

number = ...

Min-P sampling threshold (0-1)

model?

"llama-3-erato-v1" | "kayra-v1" | "clio-v1" = ...

Model to use for generation

n?

number = ...

Number of completions to generate

presencePenalty?

number = ...

Presence penalty (-2.0 to 2.0)

seed?

number = ...

Random seed for reproducibility

stop?

string | string[] = ...

Stop sequences - generation stops when any is encountered

stream?

boolean = ...

Enable streaming response

suffix?

string = ...

Suffix to append after the completion

temperature?

number = ...

Sampling temperature (0-2). Higher = more random

topK?

number = ...

Top-K sampling. 0 = disabled

topP?

number = ...

Nucleus sampling threshold (0-1)

unifiedCubic?

number = ...

Unified cubic sampling parameter

unifiedIncreaseLinearWithEntropy?

number = ...

Unified increase linear with entropy parameter

unifiedLinear?

number = ...

Unified linear sampling parameter

unifiedQuadratic?

number = ...

Unified quadratic sampling parameter

user?

string = ...

User identifier for tracking

Returns

Promise<OAIChatCompletionResponse>

Chat completion response with generated message

Throws

ZodError if parameters fail validation

Example

typescript

const response = await client.openai.chatCompletion({
  messages: [
    { role: 'system', content: 'You are a helpful assistant.' },
    { role: 'user', content: 'Hello!' },
  ],
  maxTokens: 100,
});
console.log(response.choices[0].message.content);

chatCompletionStream()

chatCompletionStream(params, signal?): AsyncGenerator<OAIChatCompletionChunk, void, unknown>

Defined in: src/client.ts:609

Chat completion with streaming and validation

Parameters

params

User-friendly chat completion parameters (camelCase)

bestOf?

number = ...

Number of completions to generate server-side and return the best

echo?

boolean = ...

Echo back the prompt in addition to the completion

enableThinking?

boolean = ...

Enable thinking/reasoning mode

frequencyPenalty?

number = ...

Frequency penalty (-2.0 to 2.0)

logitBias?

Record<string, number> = ...

Logit bias for specific tokens

logprobs?

number = ...

Number of log probabilities to return

maxTokens?

number = ...

Maximum tokens to generate (1-2048)

messages

object[] = ...

Array of chat messages

minP?

number = ...

Min-P sampling threshold (0-1)

model?

"llama-3-erato-v1" | "kayra-v1" | "clio-v1" = ...

Model to use for generation

n?

number = ...

Number of completions to generate

presencePenalty?

number = ...

Presence penalty (-2.0 to 2.0)

seed?

number = ...

Random seed for reproducibility

stop?

string | string[] = ...

Stop sequences - generation stops when any is encountered

stream?

boolean = ...

Enable streaming response

suffix?

string = ...

Suffix to append after the completion

temperature?

number = ...

Sampling temperature (0-2). Higher = more random

topK?

number = ...

Top-K sampling. 0 = disabled

topP?

number = ...

Nucleus sampling threshold (0-1)

unifiedCubic?

number = ...

Unified cubic sampling parameter

unifiedIncreaseLinearWithEntropy?

number = ...

Unified increase linear with entropy parameter

unifiedLinear?

number = ...

Unified linear sampling parameter

unifiedQuadratic?

number = ...

Unified quadratic sampling parameter

user?

string = ...

User identifier for tracking

signal?

AbortSignal

Optional AbortSignal for cancellation

Returns

AsyncGenerator<OAIChatCompletionChunk, void, unknown>

Yields

Chat completion chunks as they are generated

Example

typescript

const controller = new AbortController();

for await (const chunk of client.openai.chatCompletionStream({
  messages: [{ role: 'user', content: 'Tell me a story' }],
  maxTokens: 200,
}, controller.signal)) {
  const content = chunk.choices[0]?.delta?.content;
  if (content) process.stdout.write(content);
}

completion()

completion(params): Promise<OAICompletionResponse>

Defined in: src/client.ts:511

Text completion with validation (OpenAI-compatible)

Parameters

params

User-friendly completion parameters (camelCase)

bestOf?

number = ...

Number of completions to generate server-side and return the best

echo?

boolean = ...

Echo back the prompt in addition to the completion

frequencyPenalty?

number = ...

Frequency penalty (-2.0 to 2.0)

logitBias?

Record<string, number> = ...

Logit bias for specific tokens

logprobs?

number = ...

Number of log probabilities to return

maxTokens?

number = ...

Maximum tokens to generate (1-2048)

minP?

number = ...

Min-P sampling threshold (0-1)

model?

"llama-3-erato-v1" | "kayra-v1" | "clio-v1" = ...

Model to use for generation

n?

number = ...

Number of completions to generate

presencePenalty?

number = ...

Presence penalty (-2.0 to 2.0)

prompt

string | number[] | number[][] | string[] = OAIPromptSchema

Input prompt - string, array of strings, or array of token arrays

seed?

number = ...

Random seed for reproducibility

stop?

string | string[] = ...

Stop sequences - generation stops when any is encountered

stream?

boolean = ...

Enable streaming response

suffix?

string = ...

Suffix to append after the completion

temperature?

number = ...

Sampling temperature (0-2). Higher = more random

topK?

number = ...

Top-K sampling. 0 = disabled

topP?

number = ...

Nucleus sampling threshold (0-1)

unifiedCubic?

number = ...

Unified cubic sampling parameter

unifiedIncreaseLinearWithEntropy?

number = ...

Unified increase linear with entropy parameter

unifiedLinear?

number = ...

Unified linear sampling parameter

unifiedQuadratic?

number = ...

Unified quadratic sampling parameter

user?

string = ...

User identifier for tracking

Returns

Promise<OAICompletionResponse>

Completion response with generated text

Throws

ZodError if parameters fail validation

Example

typescript

const response = await client.openai.completion({
  prompt: 'The quick brown fox',
  maxTokens: 50,
  temperature: 0.7,
});
console.log(response.choices[0].text);

completionStream()

completionStream(params, signal?): AsyncGenerator<OAICompletionChunk, void, unknown>

Defined in: src/client.ts:543

Text completion with streaming and validation

Parameters

params

User-friendly completion parameters (camelCase)

bestOf?

number = ...

Number of completions to generate server-side and return the best

echo?

boolean = ...

Echo back the prompt in addition to the completion

frequencyPenalty?

number = ...

Frequency penalty (-2.0 to 2.0)

logitBias?

Record<string, number> = ...

Logit bias for specific tokens

logprobs?

number = ...

Number of log probabilities to return

maxTokens?

number = ...

Maximum tokens to generate (1-2048)

minP?

number = ...

Min-P sampling threshold (0-1)

model?

"llama-3-erato-v1" | "kayra-v1" | "clio-v1" = ...

Model to use for generation

n?

number = ...

Number of completions to generate

presencePenalty?

number = ...

Presence penalty (-2.0 to 2.0)

prompt

string | number[] | number[][] | string[] = OAIPromptSchema

Input prompt - string, array of strings, or array of token arrays

seed?

number = ...

Random seed for reproducibility

stop?

string | string[] = ...

Stop sequences - generation stops when any is encountered

stream?

boolean = ...

Enable streaming response

suffix?

string = ...

Suffix to append after the completion

temperature?

number = ...

Sampling temperature (0-2). Higher = more random

topK?

number = ...

Top-K sampling. 0 = disabled

topP?

number = ...

Nucleus sampling threshold (0-1)

unifiedCubic?

number = ...

Unified cubic sampling parameter

unifiedIncreaseLinearWithEntropy?

number = ...

Unified increase linear with entropy parameter

unifiedLinear?

number = ...

Unified linear sampling parameter

unifiedQuadratic?

number = ...

Unified quadratic sampling parameter

user?

string = ...

User identifier for tracking

signal?

AbortSignal

Optional AbortSignal for cancellation

Returns

AsyncGenerator<OAICompletionChunk, void, unknown>

Yields

Completion chunks as they are generated

Example

typescript

const controller = new AbortController();

for await (const chunk of client.openai.completionStream({
  prompt: 'Once upon a time',
  maxTokens: 100,
}, controller.signal)) {
  process.stdout.write(chunk.choices[0]?.text ?? '');
}

listModels()

listModels(): Promise<OAIModel[]>

Defined in: src/client.ts:636

List available models

Returns

Promise<OAIModel[]>

Array of available model objects

Example

typescript

const models = await client.openai.listModels();
for (const model of models) {
  console.log(`${model.id} (owned by ${model.owned_by})`);
}

tokenCount()

tokenCount(params): Promise<number>

Defined in: src/client.ts:656

Count tokens in text

Parameters

params

Token count parameters

model?

"llama-3-erato-v1" | "kayra-v1" | "clio-v1" = ...

Model to use for tokenization

prompt

string | number[] | number[][] | string[] = OAIPromptSchema

Prompt to count tokens for - string, array of strings, or array of token arrays

Returns

Promise<number>

Number of tokens in the text

Example

typescript

const count = await client.openai.tokenCount({
  prompt: 'Hello, world!',
  model: 'llama-3-erato-v1',
});
console.log(`Token count: ${count}`);

Class: OpenAIGeneration ​

Example ​

Constructors ​

Constructor ​

Parameters ​

client ​

Returns ​

Methods ​

chatCompletion() ​

Parameters ​

params ​

bestOf? ​

echo? ​

enableThinking? ​

frequencyPenalty? ​

logitBias? ​

logprobs? ​

maxTokens? ​

messages ​

minP? ​

model? ​

n? ​

presencePenalty? ​

seed? ​

stop? ​

stream? ​

suffix? ​

temperature? ​

topK? ​

topP? ​

unifiedCubic? ​

unifiedIncreaseLinearWithEntropy? ​

unifiedLinear? ​

unifiedQuadratic? ​

user? ​

Returns ​

Throws ​

Example ​

chatCompletionStream() ​

Parameters ​

params ​

bestOf? ​

echo? ​

enableThinking? ​

frequencyPenalty? ​

logitBias? ​

logprobs? ​

maxTokens? ​

messages ​

minP? ​

model? ​

n? ​

presencePenalty? ​

seed? ​

stop? ​

stream? ​

suffix? ​

temperature? ​

topK? ​

topP? ​

unifiedCubic? ​

unifiedIncreaseLinearWithEntropy? ​

unifiedLinear? ​

unifiedQuadratic? ​

user? ​

signal? ​

Returns ​

Yields ​

Example ​

completion() ​

Parameters ​

params ​

bestOf? ​

echo? ​

frequencyPenalty? ​

logitBias? ​

logprobs? ​

maxTokens? ​

minP? ​

model? ​

Class: OpenAIGeneration

Example

Constructors

Constructor

Parameters

client

Returns

Methods

chatCompletion()

Parameters

params

bestOf?

echo?

enableThinking?

frequencyPenalty?

logitBias?

logprobs?

maxTokens?

messages

minP?

model?

n?

presencePenalty?

seed?

stop?

stream?

suffix?

temperature?

topK?

topP?

unifiedCubic?

unifiedIncreaseLinearWithEntropy?

unifiedLinear?

unifiedQuadratic?

user?

Returns

Throws

Example

chatCompletionStream()

Parameters

params

bestOf?

echo?

enableThinking?

frequencyPenalty?

logitBias?

logprobs?

maxTokens?

messages

minP?

model?

n?

presencePenalty?

seed?

stop?

stream?

suffix?

temperature?

topK?

topP?

unifiedCubic?

unifiedIncreaseLinearWithEntropy?

unifiedLinear?

unifiedQuadratic?

user?

signal?

Returns

Yields

Example

completion()

Parameters

params

bestOf?

echo?

frequencyPenalty?

logitBias?

logprobs?

maxTokens?

minP?

model?