mirror of
https://github.com/geoffsee/open-gsio.git
synced 2025-09-08 22:56:46 +00:00
adds eslint
This commit is contained in:

committed by
Geoff Seemueller

parent
9698fc6f3b
commit
02c3253343
@@ -1,95 +1,97 @@
|
||||
import { OpenAI } from "openai";
|
||||
import { Utils } from "../lib/utils";
|
||||
import { ChatCompletionCreateParamsStreaming } from "openai/resources/chat/completions/completions";
|
||||
import { BaseChatProvider, CommonProviderParams } from "./chat-stream-provider";
|
||||
import { OpenAI } from 'openai';
|
||||
import { ChatCompletionCreateParamsStreaming } from 'openai/resources/chat/completions/completions';
|
||||
|
||||
import { Utils } from '../lib/utils';
|
||||
|
||||
import { BaseChatProvider, CommonProviderParams } from './chat-stream-provider';
|
||||
|
||||
export class MlxOmniChatProvider extends BaseChatProvider {
|
||||
getOpenAIClient(param: CommonProviderParams): OpenAI {
|
||||
return new OpenAI({
|
||||
baseURL: "http://localhost:10240",
|
||||
apiKey: param.env.MLX_API_KEY,
|
||||
});
|
||||
getOpenAIClient(param: CommonProviderParams): OpenAI {
|
||||
return new OpenAI({
|
||||
baseURL: 'http://localhost:10240',
|
||||
apiKey: param.env.MLX_API_KEY,
|
||||
});
|
||||
}
|
||||
|
||||
getStreamParams(
|
||||
param: CommonProviderParams,
|
||||
safeMessages: any[],
|
||||
): ChatCompletionCreateParamsStreaming {
|
||||
const baseTuningParams = {
|
||||
temperature: 0.86,
|
||||
top_p: 0.98,
|
||||
presence_penalty: 0.1,
|
||||
frequency_penalty: 0.3,
|
||||
max_tokens: param.maxTokens as number,
|
||||
};
|
||||
|
||||
const getTuningParams = () => {
|
||||
return baseTuningParams;
|
||||
};
|
||||
|
||||
let completionRequest: ChatCompletionCreateParamsStreaming = {
|
||||
model: param.model,
|
||||
stream: true,
|
||||
messages: safeMessages,
|
||||
};
|
||||
|
||||
const client = this.getOpenAIClient(param);
|
||||
const isLocal = client.baseURL.includes('localhost');
|
||||
|
||||
if (isLocal) {
|
||||
completionRequest['messages'] = Utils.normalizeWithBlanks(safeMessages);
|
||||
completionRequest['stream_options'] = {
|
||||
include_usage: true,
|
||||
};
|
||||
} else {
|
||||
completionRequest = { ...completionRequest, ...getTuningParams() };
|
||||
}
|
||||
|
||||
getStreamParams(param: CommonProviderParams, safeMessages: any[]): ChatCompletionCreateParamsStreaming {
|
||||
const baseTuningParams = {
|
||||
temperature: 0.86,
|
||||
top_p: 0.98,
|
||||
presence_penalty: 0.1,
|
||||
frequency_penalty: 0.3,
|
||||
max_tokens: param.maxTokens as number,
|
||||
};
|
||||
return completionRequest;
|
||||
}
|
||||
|
||||
const getTuningParams = () => {
|
||||
return baseTuningParams;
|
||||
};
|
||||
async processChunk(chunk: any, dataCallback: (data: any) => void): Promise<boolean> {
|
||||
const isLocal = chunk.usage !== undefined;
|
||||
|
||||
let completionRequest: ChatCompletionCreateParamsStreaming = {
|
||||
model: param.model,
|
||||
stream: true,
|
||||
messages: safeMessages
|
||||
};
|
||||
|
||||
const client = this.getOpenAIClient(param);
|
||||
const isLocal = client.baseURL.includes("localhost");
|
||||
|
||||
if(isLocal) {
|
||||
completionRequest["messages"] = Utils.normalizeWithBlanks(safeMessages);
|
||||
completionRequest["stream_options"] = {
|
||||
include_usage: true
|
||||
};
|
||||
} else {
|
||||
completionRequest = {...completionRequest, ...getTuningParams()};
|
||||
}
|
||||
|
||||
return completionRequest;
|
||||
if (isLocal && chunk.usage) {
|
||||
dataCallback({
|
||||
type: 'chat',
|
||||
data: {
|
||||
choices: [
|
||||
{
|
||||
delta: { content: '' },
|
||||
logprobs: null,
|
||||
finish_reason: 'stop',
|
||||
},
|
||||
],
|
||||
},
|
||||
});
|
||||
return true; // Break the stream
|
||||
}
|
||||
|
||||
async processChunk(chunk: any, dataCallback: (data: any) => void): Promise<boolean> {
|
||||
const isLocal = chunk.usage !== undefined;
|
||||
|
||||
if (isLocal && chunk.usage) {
|
||||
dataCallback({
|
||||
type: "chat",
|
||||
data: {
|
||||
choices: [
|
||||
{
|
||||
delta: { content: "" },
|
||||
logprobs: null,
|
||||
finish_reason: "stop",
|
||||
},
|
||||
],
|
||||
},
|
||||
});
|
||||
return true; // Break the stream
|
||||
}
|
||||
|
||||
dataCallback({ type: "chat", data: chunk });
|
||||
return false; // Continue the stream
|
||||
}
|
||||
dataCallback({ type: 'chat', data: chunk });
|
||||
return false; // Continue the stream
|
||||
}
|
||||
}
|
||||
|
||||
export class MlxOmniChatSdk {
|
||||
private static provider = new MlxOmniChatProvider();
|
||||
private static provider = new MlxOmniChatProvider();
|
||||
|
||||
static async handleMlxOmniStream(
|
||||
ctx: any,
|
||||
dataCallback: (data: any) => any,
|
||||
) {
|
||||
if (!ctx.messages?.length) {
|
||||
return new Response("No messages provided", { status: 400 });
|
||||
}
|
||||
|
||||
return this.provider.handleStream(
|
||||
{
|
||||
systemPrompt: ctx.systemPrompt,
|
||||
preprocessedContext: ctx.preprocessedContext,
|
||||
maxTokens: ctx.maxTokens,
|
||||
messages: Utils.normalizeWithBlanks(ctx.messages),
|
||||
model: ctx.model,
|
||||
env: ctx.env
|
||||
},
|
||||
dataCallback,
|
||||
);
|
||||
static async handleMlxOmniStream(ctx: any, dataCallback: (data: any) => any) {
|
||||
if (!ctx.messages?.length) {
|
||||
return new Response('No messages provided', { status: 400 });
|
||||
}
|
||||
|
||||
return this.provider.handleStream(
|
||||
{
|
||||
systemPrompt: ctx.systemPrompt,
|
||||
preprocessedContext: ctx.preprocessedContext,
|
||||
maxTokens: ctx.maxTokens,
|
||||
messages: Utils.normalizeWithBlanks(ctx.messages),
|
||||
model: ctx.model,
|
||||
env: ctx.env,
|
||||
},
|
||||
dataCallback,
|
||||
);
|
||||
}
|
||||
}
|
||||
|
Reference in New Issue
Block a user