send-prompt.ts•7.11 kB
import {
createAction,
Property,
} from '@activepieces/pieces-framework';
import OpenAI from 'openai';
import {
AuthenticationType,
httpClient,
HttpMethod,
} from '@activepieces/pieces-common';
import { localaiAuth } from '../..';
import { json } from 'stream/consumers';
const billingIssueMessage = `Error Occurred: 429 \n
1. Set LocalAI API Url. \n
2. Generate a new API key (optional). \n
3. Attempt the process again. \n
For guidance, visit: https://localai.io/`;
const unaurthorizedMessage = `Error Occurred: 401 \n
Ensure that your API key is valid. \n
`;
export const askLocalAI = createAction({
auth: localaiAuth,
name: 'ask_localai',
displayName: 'Ask LocalAI',
description: 'Ask LocalAI anything you want!',
props: {
model: Property.Dropdown({
displayName: 'Model',
required: true,
description:
'The model which will generate the completion. Some models are suitable for natural language tasks, others specialize in code.',
refreshers: [],
defaultValue: 'gpt-3.5-turbo',
options: async ({ auth }) => {
if (!auth) {
return {
disabled: true,
placeholder: 'Enter your api key first',
options: [],
};
}
try {
const response = await httpClient.sendRequest<{
data: { id: string }[];
}>({
url: (<any>auth).base_url + '/models',
method: HttpMethod.GET,
authentication: {
type: AuthenticationType.BEARER_TOKEN,
token: (<any>auth).access_token as string,
},
});
return {
disabled: false,
options: response.body.data.map((model) => {
return {
label: model.id,
value: model.id,
};
}),
};
} catch (error) {
return {
disabled: true,
options: [],
placeholder: "Couldn't Load Models",
};
}
},
}),
prompt: Property.LongText({
displayName: 'Question',
required: true,
}),
temperature: Property.Number({
displayName: 'Temperature',
required: false,
description:
'Controls randomness: Lowering results in less random completions. As the temperature approaches zero, the model will become deterministic and repetitive.',
}),
maxTokens: Property.Number({
displayName: 'Maximum Tokens',
required: false,
description:
"The maximum number of tokens to generate. Requests can use up to 2,048 or 4,096 tokens shared between prompt and completion, don't set the value to maximum and leave some tokens for the input. The exact limit varies by model. (One token is roughly 4 characters for normal English text)",
}),
topP: Property.Number({
displayName: 'Top P',
required: false,
description:
'An alternative to sampling with temperature, called nucleus sampling, where the model considers the results of the tokens with top_p probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered.',
}),
frequencyPenalty: Property.Number({
displayName: 'Frequency penalty',
required: false,
description:
"Number between -2.0 and 2.0. Positive values penalize new tokens based on their existing frequency in the text so far, decreasing the model's likelihood to repeat the same line verbatim.",
}),
presencePenalty: Property.Number({
displayName: 'Presence penalty',
required: false,
description:
"Number between -2.0 and 2.0. Positive values penalize new tokens based on whether they appear in the text so far, increasing the mode's likelihood to talk about new topics.",
}),
roles: Property.Json({
displayName: 'Roles',
required: false,
description: 'Array of roles to specify more accurate response',
defaultValue: [
{ role: 'system', content: 'You are a helpful assistant.' },
],
}),
},
async run({ auth, propsValue }) {
const openai = new OpenAI({
baseURL: auth.base_url,
apiKey: auth.access_token,
});
let billingIssue = false;
let unaurthorized = false;
let model = 'gpt-3.5-turbo';
if (propsValue.model) {
model = propsValue.model;
}
let temperature = 0.9;
if (propsValue.temperature) {
temperature = Number(propsValue.temperature);
}
let maxTokens = 2048;
if (propsValue.maxTokens) {
maxTokens = Number(propsValue.maxTokens);
}
let topP = 1;
if (propsValue.topP) {
topP = Number(propsValue.topP);
}
let frequencyPenalty = 0.0;
if (propsValue.frequencyPenalty) {
frequencyPenalty = Number(propsValue.frequencyPenalty);
}
let presencePenalty = 0.6;
if (propsValue.presencePenalty) {
presencePenalty = Number(propsValue.presencePenalty);
}
const rolesArray = propsValue.roles
? (propsValue.roles as unknown as any[])
: [];
const roles = rolesArray.map((item) => {
const rolesEnum = ['system', 'user', 'assistant'];
if (!rolesEnum.includes(item.role)) {
throw new Error(
'The only available roles are: [system, user, assistant]'
);
}
return {
role: item.role,
content: item.content,
};
});
const maxRetries = 4;
let retries = 0;
let response: string | undefined;
while (retries < maxRetries) {
try {
response = (
await openai.chat.completions.create({
model: model,
messages: [
...roles,
{
role: 'user',
content: propsValue['prompt'],
},
],
temperature: temperature,
max_tokens: maxTokens,
top_p: topP,
frequency_penalty: frequencyPenalty,
presence_penalty: presencePenalty,
})
)?.choices[0]?.message?.content?.trim();
break; // Break out of the loop if the request is successful
} catch (error: any) {
if (error?.message?.includes('code 429')) {
billingIssue = true;
if (retries + 1 === maxRetries) {
throw error;
}
// Calculate the time delay for the next retry using exponential backoff
const delay = Math.pow(6, retries) * 1000;
console.log(`Retrying in ${delay} milliseconds...`);
await sleep(delay); // Wait for the calculated delay
retries++;
break;
} else {
if (error?.message?.includes('code 401')) {
unaurthorized = true;
}
throw error;
}
}
}
if (billingIssue) {
throw new Error(billingIssueMessage);
}
if (unaurthorized) {
throw new Error(unaurthorizedMessage);
}
return response;
},
});
function sleep(ms: number) {
return new Promise((resolve) => setTimeout(resolve, ms));
}