File size: 5,471 Bytes
f8b5d42 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 | const { NativeEmbedder } = require("../../EmbeddingEngines/native");
const {
handleDefaultStreamResponseV2,
formatChatHistory,
} = require("../../helpers/chat/responses");
const {
LLMPerformanceMonitor,
} = require("../../helpers/chat/LLMPerformanceMonitor");
class TextGenWebUILLM {
constructor(embedder = null) {
const { OpenAI: OpenAIApi } = require("openai");
if (!process.env.TEXT_GEN_WEB_UI_BASE_PATH)
throw new Error(
"TextGenWebUI must have a valid base path to use for the api."
);
this.basePath = process.env.TEXT_GEN_WEB_UI_BASE_PATH;
this.openai = new OpenAIApi({
baseURL: this.basePath,
apiKey: process.env.TEXT_GEN_WEB_UI_API_KEY ?? null,
});
this.model = null;
this.limits = {
history: this.promptWindowLimit() * 0.15,
system: this.promptWindowLimit() * 0.15,
user: this.promptWindowLimit() * 0.7,
};
this.embedder = embedder ?? new NativeEmbedder();
this.defaultTemp = 0.7;
this.log(`Inference API: ${this.basePath} Model: ${this.model}`);
}
log(text, ...args) {
console.log(`\x1b[36m[${this.constructor.name}]\x1b[0m ${text}`, ...args);
}
#appendContext(contextTexts = []) {
if (!contextTexts || !contextTexts.length) return "";
return (
"\nContext:\n" +
contextTexts
.map((text, i) => {
return `[CONTEXT ${i}]:\n${text}\n[END CONTEXT ${i}]\n\n`;
})
.join("")
);
}
streamingEnabled() {
return "streamGetChatCompletion" in this;
}
static promptWindowLimit(_modelName) {
const limit = process.env.TEXT_GEN_WEB_UI_MODEL_TOKEN_LIMIT || 4096;
if (!limit || isNaN(Number(limit)))
throw new Error("No token context limit was set.");
return Number(limit);
}
// Ensure the user set a value for the token limit
// and if undefined - assume 4096 window.
promptWindowLimit() {
const limit = process.env.TEXT_GEN_WEB_UI_MODEL_TOKEN_LIMIT || 4096;
if (!limit || isNaN(Number(limit)))
throw new Error("No token context limit was set.");
return Number(limit);
}
// Short circuit since we have no idea if the model is valid or not
// in pre-flight for generic endpoints
isValidChatCompletionModel(_modelName = "") {
return true;
}
/**
* Generates appropriate content array for a message + attachments.
* @param {{userPrompt:string, attachments: import("../../helpers").Attachment[]}}
* @returns {string|object[]}
*/
#generateContent({ userPrompt, attachments = [] }) {
if (!attachments.length) {
return userPrompt;
}
const content = [{ type: "text", text: userPrompt }];
for (let attachment of attachments) {
content.push({
type: "image_url",
image_url: {
url: attachment.contentString,
},
});
}
return content.flat();
}
/**
* Construct the user prompt for this model.
* @param {{attachments: import("../../helpers").Attachment[]}} param0
* @returns
*/
constructPrompt({
systemPrompt = "",
contextTexts = [],
chatHistory = [],
userPrompt = "",
attachments = [],
}) {
const prompt = {
role: "system",
content: `${systemPrompt}${this.#appendContext(contextTexts)}`,
};
return [
prompt,
...formatChatHistory(chatHistory, this.#generateContent),
{
role: "user",
content: this.#generateContent({ userPrompt, attachments }),
},
];
}
async getChatCompletion(messages = null, { temperature = 0.7 }) {
const result = await LLMPerformanceMonitor.measureAsyncFunction(
this.openai.chat.completions
.create({
model: this.model,
messages,
temperature,
})
.catch((e) => {
throw new Error(e.message);
})
);
if (
!result.output.hasOwnProperty("choices") ||
result.output.choices.length === 0
)
return null;
return {
textResponse: result.output.choices[0].message.content,
metrics: {
prompt_tokens: result.output.usage?.prompt_tokens || 0,
completion_tokens: result.output.usage?.completion_tokens || 0,
total_tokens: result.output.usage?.total_tokens || 0,
outputTps: result.output.usage?.completion_tokens / result.duration,
duration: result.duration,
},
};
}
async streamGetChatCompletion(messages = null, { temperature = 0.7 }) {
const measuredStreamRequest = await LLMPerformanceMonitor.measureStream(
this.openai.chat.completions.create({
model: this.model,
stream: true,
messages,
temperature,
}),
messages
);
return measuredStreamRequest;
}
handleStream(response, stream, responseProps) {
return handleDefaultStreamResponseV2(response, stream, responseProps);
}
// Simple wrapper for dynamic embedder & normalize interface for all LLM implementations
async embedTextInput(textInput) {
return await this.embedder.embedTextInput(textInput);
}
async embedChunks(textChunks = []) {
return await this.embedder.embedChunks(textChunks);
}
async compressMessages(promptArgs = {}, rawHistory = []) {
const { messageArrayCompressor } = require("../../helpers/chat");
const messageArray = this.constructPrompt(promptArgs);
return await messageArrayCompressor(this, messageArray, rawHistory);
}
}
module.exports = {
TextGenWebUILLM,
};
|