EverMemoryArchive

Appearance

ema / ema / OpenAIClient

Class: OpenAIClient ​

Defined in: llm/openai_client.ts:21

LLM client using OpenAI's protocol.

This client uses the official OpenAI SDK and supports:

  • Reasoning content (via reasoning_split=True)
  • Tool calling
  • Retry logic

Extends ​

  • LLMClient

Constructors ​

Constructor ​

ts
new OpenAIClient(
   apiKey, 
   apiBase, 
   model, 
   retryConfig?, 
   proxyUrl?): OpenAIClient;

Defined in: llm/openai_client.ts:24

Parameters ​

apiKey ​

string

API key for authentication

apiBase ​

string

Base URL for the API

model ​

string

Model name to use

retryConfig? ​

RetryConfig

Optional retry configuration

proxyUrl? ​

string

Optional proxy URL

Returns ​

OpenAIClient

Overrides ​

ts
LLMClient.constructor

Properties ​

apiBase ​

ts
protected readonly apiBase: string;

Defined in: llm/base.ts:30

Base URL for the API

Inherited from ​

ts
LLMClient.apiBase

apiKey ​

ts
protected readonly apiKey: string;

Defined in: llm/base.ts:26

API key for authentication

Inherited from ​

ts
LLMClient.apiKey

model ​

ts
protected readonly model: string;

Defined in: llm/base.ts:34

Model name to use

Inherited from ​

ts
LLMClient.model

retryCallback ​

ts
retryCallback: (exception, attempt) => void | undefined = undefined;

Defined in: llm/base.ts:19

Inherited from ​

ts
LLMClient.retryCallback

retryConfig ​

ts
protected readonly retryConfig: RetryConfig;

Defined in: llm/base.ts:38

Optional retry configuration

Inherited from ​

ts
LLMClient.retryConfig

Methods ​

_convertMessages() ​

ts
_convertMessages(messages): [string | undefined, Record<string, unknown>[]];

Defined in: llm/openai_client.ts:170

Converts internal messages to OpenAI format. Note: OpenAI includes system message in the messages array

Parameters ​

messages ​

Message[]

List of internal Message objects

Returns ​

[string | undefined, Record<string, unknown>[]]

Tuple of (system_message, api_messages)

Overrides ​

ts
LLMClient._convertMessages

_makeApiRequest() ​

ts
_makeApiRequest(apiMessages, tools?): Promise<any>;

Defined in: llm/openai_client.ts:71

Executes API request (core method that can be retried).

Parameters ​

apiMessages ​

Record<string, unknown>[]

List of messages in OpenAI format

tools? ​

any[]

Optional list of tools

Returns ​

Promise<any>

OpenAI ChatCompletion response (full response including usage)

_parseResponse() ​

ts
_parseResponse(response): LLMResponse;

Defined in: llm/openai_client.ts:124

Parses OpenAI response into LLMResponse.

Parameters ​

response ​

any

OpenAI ChatCompletion response (full response object)

Returns ​

LLMResponse

LLMResponse object

_prepareRequest() ​

ts
_prepareRequest(messages, tools?): object;

Defined in: llm/openai_client.ts:230

Prepares the request for OpenAI API.

Parameters ​

messages ​

Message[]

List of conversation messages

tools? ​

any[]

Optional list of available tools

Returns ​

object

Dictionary containing request parameters

apiMessages ​
ts
apiMessages: Record<string, unknown>[];
tools? ​
ts
optional tools: any[];

Overrides ​

ts
LLMClient._prepareRequest

generate() ​

ts
generate(messages, tools?): Promise<LLMResponse>;

Defined in: llm/openai_client.ts:249

Generates response from OpenAI LLM.

Parameters ​

messages ​

Message[]

List of conversation messages

tools? ​

any[]

Optional list of available tools

Returns ​

Promise<LLMResponse>

LLMResponse containing the generated content

Overrides ​

ts
LLMClient.generate