mirror of
https://github.com/enricoros/big-AGI.git
synced 2026-05-11 14:10:15 -07:00
Enrico Ros
950 lines
34 KiB
TypeScript
950 lines
34 KiB
TypeScript
import { LLM_IF_OAI_Chat, LLM_IF_OAI_Complete, LLM_IF_OAI_Fn, LLM_IF_OAI_Json, LLM_IF_OAI_Vision } from '../../store-llms';
|
|
|
|
import type { ModelDescriptionSchema } from '../llm.server.types';
|
|
import type { OpenAIWire } from './openai.wiretypes';
|
|
import { wireGroqModelsListOutputSchema } from './groq.wiretypes';
|
|
import { wireMistralModelsListOutputSchema } from './mistral.wiretypes';
|
|
import { wireOpenrouterModelsListOutputSchema } from './openrouter.wiretypes';
|
|
import { wireTogetherAIListOutputSchema } from './togetherai.wiretypes';
|
|
|
|
|
|
// [Azure] / [OpenAI]
|
|
const _knownOpenAIChatModels: ManualMappings = [
|
|
|
|
// GPT4 Turbo with Vision -> 2024-04-09
|
|
{
|
|
idPrefix: 'gpt-4-turbo',
|
|
label: 'GPT-4 Turbo',
|
|
description: 'GPT-4 Turbo with Vision. The latest GPT-4 Turbo model with vision capabilities. Vision requests can now use JSON mode and function calling. Currently points to gpt-4-turbo-2024-04-09.',
|
|
symLink: 'gpt-4-turbo-2024-04-09',
|
|
hidden: true,
|
|
// copied from symlinked
|
|
contextWindow: 128000,
|
|
maxCompletionTokens: 4096,
|
|
trainingDataCutoff: 'Dec 2023',
|
|
interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Fn, LLM_IF_OAI_Json],
|
|
pricing: { chatIn: 10, chatOut: 30 },
|
|
benchmark: { cbaElo: 1261 },
|
|
},
|
|
{
|
|
isLatest: true,
|
|
idPrefix: 'gpt-4-turbo-2024-04-09',
|
|
label: 'GPT-4 Turbo (2024-04-09)',
|
|
description: 'GPT-4 Turbo with Vision model. Vision requests can now use JSON mode and function calling. gpt-4-turbo currently points to this version.',
|
|
contextWindow: 128000,
|
|
maxCompletionTokens: 4096,
|
|
trainingDataCutoff: 'Dec 2023',
|
|
interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Fn, LLM_IF_OAI_Json],
|
|
pricing: { chatIn: 10, chatOut: 30 },
|
|
benchmark: { cbaElo: 1261 },
|
|
},
|
|
|
|
// GPT4 Turbo Previews
|
|
{
|
|
idPrefix: 'gpt-4-turbo-preview', // GPT-4 Turbo preview model -> 0125
|
|
label: 'GPT-4 Preview Turbo',
|
|
description: 'GPT-4 Turbo preview model. Currently points to gpt-4-0125-preview.',
|
|
symLink: 'gpt-4-0125-preview',
|
|
hidden: true,
|
|
// copied from symlinked
|
|
isPreview: true,
|
|
contextWindow: 128000,
|
|
maxCompletionTokens: 4096,
|
|
trainingDataCutoff: 'Dec 2023',
|
|
interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn, LLM_IF_OAI_Json],
|
|
pricing: { chatIn: 10, chatOut: 30 },
|
|
benchmark: { cbaElo: 1251 },
|
|
},
|
|
{
|
|
idPrefix: 'gpt-4-0125-preview', // GPT-4 Turbo preview model
|
|
label: 'GPT-4 Turbo (0125)',
|
|
description: 'GPT-4 Turbo preview model intended to reduce cases of "laziness" where the model doesn\'t complete a task. Returns a maximum of 4,096 output tokens.',
|
|
isPreview: true,
|
|
contextWindow: 128000,
|
|
maxCompletionTokens: 4096,
|
|
trainingDataCutoff: 'Dec 2023',
|
|
interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn, LLM_IF_OAI_Json],
|
|
pricing: { chatIn: 10, chatOut: 30 },
|
|
benchmark: { cbaElo: 1251 },
|
|
},
|
|
{
|
|
idPrefix: 'gpt-4-1106-preview', // GPT-4 Turbo preview model
|
|
label: 'GPT-4 Turbo (1106)',
|
|
description: 'GPT-4 Turbo preview model featuring improved instruction following, JSON mode, reproducible outputs, parallel function calling, and more. Returns a maximum of 4,096 output tokens.',
|
|
isPreview: true,
|
|
contextWindow: 128000,
|
|
maxCompletionTokens: 4096,
|
|
trainingDataCutoff: 'Apr 2023',
|
|
interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
|
|
pricing: { chatIn: 10, chatOut: 30 },
|
|
benchmark: { cbaElo: 1255 },
|
|
hidden: true,
|
|
},
|
|
|
|
// GPT4 Vision Previews
|
|
{
|
|
idPrefix: 'gpt-4-vision-preview', // GPT-4 Turbo vision preview
|
|
label: 'GPT-4 Preview Vision',
|
|
description: 'GPT-4 model with the ability to understand images, in addition to all other GPT-4 Turbo capabilities. This is a preview model, we recommend developers to now use gpt-4-turbo which includes vision capabilities. Currently points to gpt-4-1106-vision-preview.',
|
|
symLink: 'gpt-4-1106-vision-preview',
|
|
// copied from symlinked
|
|
isPreview: true,
|
|
contextWindow: 128000,
|
|
maxCompletionTokens: 4096,
|
|
trainingDataCutoff: 'Apr 2023',
|
|
interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Fn],
|
|
pricing: { chatIn: 10, chatOut: 30 },
|
|
hidden: true, // Deprecated in favor of gpt-4-turbo
|
|
},
|
|
{
|
|
idPrefix: 'gpt-4-1106-vision-preview',
|
|
label: 'GPT-4 Preview Vision (1106)',
|
|
description: 'GPT-4 model with the ability to understand images, in addition to all other GPT-4 Turbo capabilities. This is a preview model, we recommend developers to now use gpt-4-turbo which includes vision capabilities. Returns a maximum of 4,096 output tokens.',
|
|
isPreview: true,
|
|
contextWindow: 128000,
|
|
maxCompletionTokens: 4096,
|
|
trainingDataCutoff: 'Apr 2023',
|
|
interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Fn],
|
|
pricing: { chatIn: 10, chatOut: 30 },
|
|
hidden: true, // Deprecated in favor of gpt-4-turbo
|
|
},
|
|
|
|
|
|
// GPT4-32k's
|
|
{
|
|
idPrefix: 'gpt-4-32k',
|
|
label: 'GPT-4 32k',
|
|
description: 'Currently points to gpt-4-32k-0613. This model was never rolled out widely in favor of GPT-4 Turbo.',
|
|
symLink: 'gpt-4-32k-0613',
|
|
// copied from symlinked
|
|
contextWindow: 32768,
|
|
trainingDataCutoff: 'Sep 2021',
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
pricing: { chatIn: 60, chatOut: 120 },
|
|
hidden: true,
|
|
},
|
|
{
|
|
idPrefix: 'gpt-4-32k-0613',
|
|
label: 'GPT-4 32k (0613)',
|
|
description: 'Snapshot of gpt-4-32k from June 13th 2023 with improved function calling support. This model was never rolled out widely in favor of GPT-4 Turbo.',
|
|
contextWindow: 32768,
|
|
trainingDataCutoff: 'Sep 2021',
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
pricing: { chatIn: 60, chatOut: 120 },
|
|
hidden: true,
|
|
},
|
|
{
|
|
idPrefix: 'gpt-4-32k-0314',
|
|
label: 'GPT-4 32k (0314)',
|
|
description: 'Snapshot of gpt-4-32 from March 14th 2023. Will be deprecated on June 13th 2024 at the earliest.',
|
|
contextWindow: 32768,
|
|
trainingDataCutoff: 'Sep 2021',
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
pricing: { chatIn: 60, chatOut: 120 },
|
|
hidden: true,
|
|
},
|
|
|
|
|
|
// GPT4's
|
|
{
|
|
isLatest: true,
|
|
idPrefix: 'gpt-4-0613',
|
|
label: 'GPT-4 (0613)',
|
|
description: 'Snapshot of gpt-4 from June 13th 2023 with improved function calling support. Data up to Sep 2021.',
|
|
contextWindow: 8192,
|
|
trainingDataCutoff: 'Sep 2021',
|
|
interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
|
|
pricing: { chatIn: 30, chatOut: 60 },
|
|
benchmark: { cbaElo: 1164 },
|
|
},
|
|
{
|
|
idPrefix: 'gpt-4-0314',
|
|
label: 'GPT-4 (0314)',
|
|
description: 'Snapshot of gpt-4 from March 14th 2023 with function calling data. Data up to Sep 2021.',
|
|
contextWindow: 8192,
|
|
trainingDataCutoff: 'Sep 2021',
|
|
interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
|
|
pricing: { chatIn: 30, chatOut: 60 },
|
|
benchmark: { cbaElo: 1189 },
|
|
hidden: true,
|
|
},
|
|
{
|
|
idPrefix: 'gpt-4',
|
|
label: 'GPT-4',
|
|
description: 'Currently points to gpt-4-0613.',
|
|
symLink: 'gpt-4-0613',
|
|
hidden: true,
|
|
// copied from symlinked
|
|
contextWindow: 8192,
|
|
trainingDataCutoff: 'Sep 2021',
|
|
interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
|
|
pricing: { chatIn: 30, chatOut: 60 },
|
|
benchmark: { cbaElo: 1164 },
|
|
},
|
|
|
|
|
|
// 3.5-Turbo-Instruct (Not for Chat)
|
|
{
|
|
idPrefix: 'gpt-3.5-turbo-instruct',
|
|
label: '3.5-Turbo Instruct',
|
|
description: 'Similar capabilities as GPT-3 era models. Compatible with legacy Completions endpoint and not Chat Completions.',
|
|
contextWindow: 4097,
|
|
trainingDataCutoff: 'Sep 2021',
|
|
interfaces: [/* NO: LLM_IF_OAI_Chat,*/ LLM_IF_OAI_Complete],
|
|
pricing: { chatIn: 1.5, chatOut: 2 },
|
|
hidden: true,
|
|
},
|
|
|
|
|
|
// 3.5-Turbo-16k's
|
|
{
|
|
isLatest: true,
|
|
idPrefix: 'gpt-3.5-turbo-0125',
|
|
label: '3.5-Turbo (0125)',
|
|
description: 'The latest GPT-3.5 Turbo model with higher accuracy at responding in requested formats and a fix for a bug which caused a text encoding issue for non-English language function calls. Returns a maximum of 4,096 output tokens.',
|
|
contextWindow: 16385,
|
|
maxCompletionTokens: 4096,
|
|
trainingDataCutoff: 'Sep 2021',
|
|
interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
|
|
pricing: { chatIn: 0.5, chatOut: 1.5 },
|
|
benchmark: { cbaElo: 1104 },
|
|
},
|
|
{
|
|
idPrefix: 'gpt-3.5-turbo-1106',
|
|
label: '3.5-Turbo (1106)',
|
|
description: 'The latest GPT-3.5 Turbo model with improved instruction following, JSON mode, reproducible outputs, parallel function calling, and more.',
|
|
contextWindow: 16385,
|
|
maxCompletionTokens: 4096,
|
|
trainingDataCutoff: 'Sep 2021',
|
|
interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
|
|
pricing: { chatIn: 1, chatOut: 2 },
|
|
benchmark: { cbaElo: 1072 },
|
|
hidden: true,
|
|
},
|
|
{
|
|
idPrefix: 'gpt-3.5-turbo',
|
|
label: '3.5-Turbo',
|
|
description: 'Currently points to gpt-3.5-turbo-0125.',
|
|
symLink: 'gpt-3.5-turbo-0125',
|
|
hidden: true,
|
|
// copied
|
|
contextWindow: 16385,
|
|
maxCompletionTokens: 4096,
|
|
trainingDataCutoff: 'Sep 2021',
|
|
interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
|
|
pricing: { chatIn: 0.5, chatOut: 1.5 },
|
|
benchmark: { cbaElo: 1104 },
|
|
},
|
|
|
|
|
|
// Azure variants - because someone forgot the dot
|
|
{
|
|
idPrefix: 'gpt-35-turbo-16k',
|
|
label: '3.5-Turbo 16k',
|
|
description: 'Fair speed and smarts, large context',
|
|
contextWindow: 16384,
|
|
interfaces: [LLM_IF_OAI_Chat], // as azure doesn't version model id's (in the deployments), let's assume no function calling
|
|
},
|
|
{
|
|
idPrefix: 'gpt-35-turbo',
|
|
label: '3.5-Turbo',
|
|
contextWindow: 4097,
|
|
description: 'Fair speed and smarts',
|
|
interfaces: [LLM_IF_OAI_Chat], // as azure doesn't version model id's (in the deployments), let's assume no function calling
|
|
},
|
|
|
|
// Fallback - unknown
|
|
{
|
|
idPrefix: '',
|
|
label: '?:',
|
|
description: 'Unknown, please let us know the ID. Assuming a 4097 context window size and Chat capabilities.',
|
|
contextWindow: 4097,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
hidden: true,
|
|
},
|
|
];
|
|
|
|
const openAIModelsDenyList: string[] = [
|
|
/* /v1/audio/speech */
|
|
'tts-1-hd', 'tts-1',
|
|
/* /v1/embeddings */
|
|
'text-embedding-3-small', 'text-embedding-3-large', 'text-embedding-ada-002',
|
|
/* /v1/audio/transcriptions, /v1/audio/translations */
|
|
'whisper-1',
|
|
/* /v1/images/generations */
|
|
'dall-e-3', 'dall-e-2',
|
|
/* /v1/completions (Legacy) */
|
|
'-turbo-instruct', 'davinci-', 'babbage-',
|
|
|
|
// just Legacy models, that we should drop
|
|
'gpt-3.5-turbo-16k-0613', 'gpt-3.5-turbo-0613', 'gpt-3.5-turbo-0301', 'gpt-3.5-turbo-16k',
|
|
];
|
|
|
|
export function openAIModelFilter(model: OpenAIWire.Models.ModelDescription) {
|
|
return !openAIModelsDenyList.some(deny => model.id.includes(deny));
|
|
}
|
|
|
|
export function openAIModelToModelDescription(modelId: string, modelCreated: number, modelUpdated?: number): ModelDescriptionSchema {
|
|
return fromManualMapping(_knownOpenAIChatModels, modelId, modelCreated, modelUpdated);
|
|
}
|
|
|
|
export function azureModelToModelDescription(azureDeploymentRef: string, openAIModelIdBase: string, modelCreated: number, modelUpdated?: number): ModelDescriptionSchema {
|
|
// if the deployment name mataches an OpenAI model prefix, use that
|
|
const known = _knownOpenAIChatModels.find(base => azureDeploymentRef == base.idPrefix);
|
|
return fromManualMapping(_knownOpenAIChatModels, known ? azureDeploymentRef : openAIModelIdBase, modelCreated, modelUpdated);
|
|
}
|
|
|
|
|
|
// [LM Studio]
|
|
export function lmStudioModelToModelDescription(modelId: string): ModelDescriptionSchema {
|
|
|
|
// LM Studio model ID's are the file names of the model files
|
|
function getFileName(filePath: string): string {
|
|
const normalizedPath = filePath.replace(/\\/g, '/');
|
|
return normalizedPath.split('/').pop() || '';
|
|
}
|
|
|
|
return fromManualMapping([], modelId, undefined, undefined, {
|
|
idPrefix: modelId,
|
|
label: getFileName(modelId)
|
|
.replace('.gguf', '')
|
|
.replace('.bin', ''),
|
|
// .replaceAll('-', ' '),
|
|
description: `Unknown LM Studio model. File: ${modelId}`,
|
|
contextWindow: null, // 'not provided'
|
|
interfaces: [LLM_IF_OAI_Chat], // assume..
|
|
});
|
|
}
|
|
|
|
|
|
// [LocalAI]
|
|
const _knownLocalAIChatModels: ManualMappings = [
|
|
{
|
|
idPrefix: 'ggml-gpt4all-j',
|
|
label: 'GPT4All-J',
|
|
description: 'GPT4All-J on LocalAI',
|
|
contextWindow: 2048,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
},
|
|
{
|
|
idPrefix: 'luna-ai-llama2',
|
|
label: 'Luna AI Llama2 Uncensored',
|
|
description: 'Luna AI Llama2 on LocalAI',
|
|
contextWindow: 4096,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
},
|
|
];
|
|
|
|
export function localAIModelToModelDescription(modelId: string): ModelDescriptionSchema {
|
|
return fromManualMapping(_knownLocalAIChatModels, modelId, undefined, undefined, {
|
|
idPrefix: modelId,
|
|
label: modelId
|
|
.replace('ggml-', '')
|
|
.replace('.bin', '')
|
|
.replaceAll('-', ' '),
|
|
description: 'Unknown localAI model. Please update `models.data.ts` with this ID',
|
|
contextWindow: null, // 'not provided'
|
|
interfaces: [LLM_IF_OAI_Chat], // assume..
|
|
});
|
|
}
|
|
|
|
|
|
// [Mistral]
|
|
|
|
const _knownMistralChatModels: ManualMappings = [
|
|
// Large
|
|
{
|
|
idPrefix: 'mistral-large-2402',
|
|
label: 'Mistral Large (2402)',
|
|
description: 'Top-tier reasoning for high-complexity tasks.',
|
|
contextWindow: 32768,
|
|
interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
|
|
pricing: { chatIn: 8, chatOut: 24 },
|
|
benchmark: { cbaElo: 1159 },
|
|
isLatest: true,
|
|
},
|
|
{
|
|
idPrefix: 'mistral-large-latest',
|
|
label: 'Mistral Large (latest)',
|
|
symLink: 'mistral-large-2402',
|
|
hidden: true,
|
|
// copied
|
|
description: 'Top-tier reasoning for high-complexity tasks.',
|
|
contextWindow: 32768,
|
|
interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
|
|
pricing: { chatIn: 8, chatOut: 24 },
|
|
benchmark: { cbaElo: 1159 },
|
|
},
|
|
{
|
|
idPrefix: 'mistral-large',
|
|
label: 'Mistral Large (?)',
|
|
description: 'Flagship model, with top-tier reasoning capabilities and language support (English, French, German, Italian, Spanish, and Code)',
|
|
contextWindow: 32768,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
hidden: true,
|
|
},
|
|
|
|
// Medium - not updated on 2024-02-26
|
|
{
|
|
idPrefix: 'mistral-medium-2312',
|
|
label: 'Mistral Medium (2312)',
|
|
description: 'Mistral internal prototype model.',
|
|
contextWindow: 32768,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
pricing: { chatIn: 2.7, chatOut: 8.1 },
|
|
benchmark: { cbaElo: 1148 },
|
|
},
|
|
{
|
|
idPrefix: 'mistral-medium-latest',
|
|
label: 'Mistral Medium (latest)',
|
|
symLink: 'mistral-medium-2312',
|
|
hidden: true,
|
|
// copied
|
|
description: 'Mistral internal prototype model.',
|
|
contextWindow: 32768,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
pricing: { chatIn: 2.7, chatOut: 8.1 },
|
|
benchmark: { cbaElo: 1148 },
|
|
},
|
|
{
|
|
idPrefix: 'mistral-medium',
|
|
label: 'Mistral Medium',
|
|
description: 'Mistral internal prototype model.',
|
|
contextWindow: 32768,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
pricing: { chatIn: 2.7, chatOut: 8.1 },
|
|
benchmark: { cbaElo: 1148 },
|
|
hidden: true,
|
|
},
|
|
|
|
// Small (8x7B)
|
|
{
|
|
idPrefix: 'mistral-small-2402',
|
|
label: 'Mistral Small (2402)',
|
|
description: 'Optimized endpoint. Cost-efficient reasoning for low-latency workloads. Mistral Small outperforms Mixtral 8x7B and has lower latency',
|
|
contextWindow: 32768,
|
|
interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
|
|
pricing: { chatIn: 2, chatOut: 6 },
|
|
isLatest: true,
|
|
},
|
|
{
|
|
idPrefix: 'mistral-small-2312',
|
|
label: 'Mistral Small (2312)',
|
|
description: 'Aka open-mixtral-8x7b. Cost-efficient reasoning for low-latency workloads. Mistral Small outperforms Mixtral 8x7B and has lower latency',
|
|
contextWindow: 32768,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
pricing: { chatIn: 2, chatOut: 6 },
|
|
hidden: true,
|
|
},
|
|
{
|
|
idPrefix: 'mistral-small-latest',
|
|
label: 'Mistral Small (latest)',
|
|
symLink: 'mistral-small-2402',
|
|
hidden: true,
|
|
// copied
|
|
description: 'Cost-efficient reasoning for low-latency workloads. Mistral Small outperforms Mixtral 8x7B and has lower latency',
|
|
contextWindow: 32768,
|
|
interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
|
|
pricing: { chatIn: 2, chatOut: 6 },
|
|
},
|
|
{
|
|
idPrefix: 'mistral-small',
|
|
label: 'Mistral Small',
|
|
description: 'Cost-efficient reasoning for low-latency workloads.',
|
|
contextWindow: 32768,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
pricing: { chatIn: 2, chatOut: 6 },
|
|
hidden: true,
|
|
},
|
|
|
|
// Open Mixtral (8x7B)
|
|
{
|
|
idPrefix: 'open-mixtral-8x7b',
|
|
label: 'Open Mixtral (8x7B)',
|
|
description: 'Mixtral 8x7B model, aka mistral-small-2312',
|
|
// symLink: 'mistral-small-2312',
|
|
// copied
|
|
contextWindow: 32768,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
pricing: { chatIn: 0.7, chatOut: 0.7 },
|
|
},
|
|
|
|
// Tiny (7B)
|
|
{
|
|
idPrefix: 'mistral-tiny-2312',
|
|
label: 'Mistral Tiny (2312)',
|
|
description: 'Aka open-mistral-7b. Used for large batch processing tasks where cost is a significant factor but reasoning capabilities are not crucial',
|
|
contextWindow: 32768,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
hidden: true,
|
|
},
|
|
{
|
|
idPrefix: 'mistral-tiny',
|
|
label: 'Mistral Tiny',
|
|
description: 'Used for large batch processing tasks where cost is a significant factor but reasoning capabilities are not crucial',
|
|
contextWindow: 32768,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
hidden: true,
|
|
},
|
|
// Open Mistral (7B)
|
|
{
|
|
idPrefix: 'open-mistral-7b',
|
|
label: 'Open Mistral (7B)',
|
|
description: 'Mistral 7B model, aka mistral-tiny-2312',
|
|
// symLink: 'mistral-tiny-2312',
|
|
// copied
|
|
contextWindow: 32768,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
pricing: { chatIn: 0.25, chatOut: 0.25 },
|
|
},
|
|
|
|
|
|
{
|
|
idPrefix: 'mistral-embed',
|
|
label: 'Mistral Embed',
|
|
description: 'State-of-the-art semantic for extracting representation of text extracts.',
|
|
// output: 1024 dimensions
|
|
maxCompletionTokens: 1024, // HACK - it's 1024 dimensions, but those are not 'completion tokens'
|
|
contextWindow: 32768, // actually unknown, assumed from the other models
|
|
interfaces: [],
|
|
hidden: true,
|
|
},
|
|
];
|
|
|
|
|
|
const mistralModelFamilyOrder = [
|
|
'mistral-large', 'mistral-medium', 'mistral-small', 'open-mixtral-8x7b', 'mistral-tiny', 'open-mistral-7b', 'mistral-embed', '🔗',
|
|
];
|
|
|
|
export function mistralModelToModelDescription(_model: unknown): ModelDescriptionSchema {
|
|
const model = wireMistralModelsListOutputSchema.parse(_model);
|
|
return fromManualMapping(_knownMistralChatModels, model.id, model.created, undefined, {
|
|
idPrefix: model.id,
|
|
label: model.id.replaceAll(/[_-]/g, ' '),
|
|
description: 'New Mistral Model',
|
|
contextWindow: 32768,
|
|
interfaces: [LLM_IF_OAI_Chat], // assume..
|
|
hidden: true,
|
|
});
|
|
}
|
|
|
|
export function mistralModelsSort(a: ModelDescriptionSchema, b: ModelDescriptionSchema): number {
|
|
const aPrefixIndex = mistralModelFamilyOrder.findIndex(prefix => a.id.startsWith(prefix));
|
|
const bPrefixIndex = mistralModelFamilyOrder.findIndex(prefix => b.id.startsWith(prefix));
|
|
if (aPrefixIndex !== -1 && bPrefixIndex !== -1) {
|
|
if (aPrefixIndex !== bPrefixIndex)
|
|
return aPrefixIndex - bPrefixIndex;
|
|
if (a.label.startsWith('🔗') && !b.label.startsWith('🔗')) return 1;
|
|
if (!a.label.startsWith('🔗') && b.label.startsWith('🔗')) return -1;
|
|
return b.label.localeCompare(a.label);
|
|
}
|
|
return aPrefixIndex !== -1 ? 1 : -1;
|
|
}
|
|
|
|
|
|
// [Oobabooga]
|
|
const _knownOobaboogaChatModels: ManualMappings = [];
|
|
|
|
const _knownOobaboogaNonChatModels: string[] = [
|
|
'None', 'text-curie-001', 'text-davinci-002', 'all-mpnet-base-v2', 'text-embedding-ada-002',
|
|
/* 'gpt-3.5-turbo' // used to be here, but now it's the way to select the activly loaded ooababooga model */
|
|
];
|
|
|
|
export function oobaboogaModelToModelDescription(modelId: string, created: number): ModelDescriptionSchema {
|
|
let label = modelId.replaceAll(/[_-]/g, ' ').split(' ').map(word => word[0].toUpperCase() + word.slice(1)).join(' ');
|
|
if (label.endsWith('.bin'))
|
|
label = label.slice(0, -4);
|
|
|
|
// special case for the default (and only 'chat') model
|
|
if (modelId === 'gpt-3.5-turbo')
|
|
label = 'Oobabooga Model';
|
|
|
|
return fromManualMapping(_knownOobaboogaChatModels, modelId, created, undefined, {
|
|
idPrefix: modelId,
|
|
label: label,
|
|
description: 'Oobabooga model',
|
|
contextWindow: 4096, // FIXME: figure out how to the context window size from Oobabooga
|
|
interfaces: [LLM_IF_OAI_Chat], // assume..
|
|
hidden: _knownOobaboogaNonChatModels.includes(modelId),
|
|
});
|
|
}
|
|
|
|
|
|
// [OpenRouter]
|
|
|
|
const orOldModelIDs = [
|
|
'openai/gpt-3.5-turbo-0301', 'openai/gpt-4-0314', 'openai/gpt-4-32k-0314', 'openai/text-davinci-002',
|
|
'anthropic/claude-2.1', 'anthropic/claude-2.0', 'anthropic/claude-v1', 'anthropic/claude-1.2',
|
|
'anthropic/claude-instant-v1-100k', 'anthropic/claude-v1-100k', 'anthropic/claude-instant-1.0',
|
|
];
|
|
|
|
const orModelFamilyOrder = [
|
|
// great models (pickes by hand, they're free)
|
|
'mistralai/mistral-7b-instruct', 'nousresearch/nous-capybara-7b',
|
|
// great orgs
|
|
'huggingfaceh4/', 'openchat/', 'anthropic/', 'google/', 'mistralai/', 'openai/', 'meta-llama/', 'phind/',
|
|
];
|
|
|
|
export function openRouterModelFamilySortFn(a: { id: string }, b: { id: string }): number {
|
|
const aPrefixIndex = orModelFamilyOrder.findIndex(prefix => a.id.startsWith(prefix));
|
|
const bPrefixIndex = orModelFamilyOrder.findIndex(prefix => b.id.startsWith(prefix));
|
|
|
|
// If both have a prefix, sort by prefix first, and then alphabetically
|
|
if (aPrefixIndex !== -1 && bPrefixIndex !== -1)
|
|
return aPrefixIndex !== bPrefixIndex ? aPrefixIndex - bPrefixIndex : a.id.localeCompare(b.id);
|
|
|
|
// If one has a prefix and the other doesn't, prioritize the one with prefix
|
|
return aPrefixIndex !== -1 ? -1 : 1;
|
|
}
|
|
|
|
export function openRouterModelToModelDescription(wireModel: object): ModelDescriptionSchema {
|
|
|
|
// parse the model
|
|
const model = wireOpenrouterModelsListOutputSchema.parse(wireModel);
|
|
|
|
// parse pricing
|
|
const pricing: ModelDescriptionSchema['pricing'] = {
|
|
chatIn: parseFloat(model.pricing.prompt) * 1000,
|
|
chatOut: parseFloat(model.pricing.completion),
|
|
};
|
|
const seemsFree = pricing.chatIn === 0 && pricing.chatOut === 0;
|
|
|
|
// openrouter provides the fields we need as part of the model object
|
|
let label = model.name || model.id.replace('/', ' · ');
|
|
if (seemsFree)
|
|
label += ' · 🎁'; // Free? Discounted?
|
|
|
|
// hidden: hide by default older models or models not in known families
|
|
const hidden = orOldModelIDs.includes(model.id) || !orModelFamilyOrder.some(prefix => model.id.startsWith(prefix));
|
|
|
|
return fromManualMapping([], model.id, undefined, undefined, {
|
|
idPrefix: model.id,
|
|
// latest: ...
|
|
label,
|
|
// created: ...
|
|
// updated: ...
|
|
description: model.description,
|
|
contextWindow: model.context_length || 4096,
|
|
maxCompletionTokens: model.top_provider.max_completion_tokens || undefined,
|
|
// trainingDataCutoff: ...
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
// benchmark: ...
|
|
pricing,
|
|
hidden,
|
|
});
|
|
}
|
|
|
|
|
|
// [Together AI]
|
|
|
|
const _knownTogetherAIChatModels: ManualMappings = [
|
|
{
|
|
idPrefix: 'NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO',
|
|
label: 'Nous Hermes 2 - Mixtral 8x7B-DPO',
|
|
description: 'Nous Hermes 2 Mixtral 7bx8 DPO is the new flagship Nous Research model trained over the Mixtral 7bx8 MoE LLM. The model was trained on over 1,000,000 entries of primarily GPT-4 generated data, as well as other high quality data from open datasets across the AI landscape, achieving state of the art performance on a variety of tasks.',
|
|
contextWindow: 32768,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
},
|
|
{
|
|
idPrefix: 'NousResearch/Nous-Hermes-2-Mixtral-8x7B-SFT',
|
|
label: 'Nous Hermes 2 - Mixtral 8x7B-SFT',
|
|
description: 'Nous Hermes 2 Mixtral 7bx8 SFT is the new flagship Nous Research model trained over the Mixtral 7bx8 MoE LLM. The model was trained on over 1,000,000 entries of primarily GPT-4 generated data, as well as other high quality data from open datasets across the AI landscape, achieving state of the art performance on a variety of tasks.',
|
|
contextWindow: 32768,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
},
|
|
{
|
|
idPrefix: 'mistralai/Mixtral-8x7B-Instruct-v0.1',
|
|
label: 'Mixtral-8x7B Instruct',
|
|
description: 'The Mixtral-8x7B Large Language Model (LLM) is a pretrained generative Sparse Mixture of Experts.',
|
|
contextWindow: 32768,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
},
|
|
{
|
|
idPrefix: 'mistralai/Mistral-7B-Instruct-v0.2',
|
|
label: 'Mistral (7B) Instruct v0.2',
|
|
description: 'The Mistral-7B-Instruct-v0.2 Large Language Model (LLM) is an improved instruct fine-tuned version of Mistral-7B-Instruct-v0.1.',
|
|
contextWindow: 32768,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
},
|
|
{
|
|
idPrefix: 'NousResearch/Nous-Hermes-2-Yi-34B',
|
|
label: 'Nous Hermes-2 Yi (34B)',
|
|
description: 'Nous Hermes 2 - Yi-34B is a state of the art Yi Fine-tune',
|
|
contextWindow: 4097,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
},
|
|
] as const;
|
|
|
|
export function togetherAIModelsToModelDescriptions(wireModels: unknown): ModelDescriptionSchema[] {
|
|
|
|
function togetherAIModelToModelDescription(model: { id: string, created: number }) {
|
|
return fromManualMapping(_knownTogetherAIChatModels, model.id, model.created, undefined, {
|
|
idPrefix: model.id,
|
|
label: model.id.replaceAll('/', ' · ').replaceAll(/[_-]/g, ' '),
|
|
description: 'New Togehter AI Model',
|
|
contextWindow: null, // unknown
|
|
interfaces: [LLM_IF_OAI_Chat], // assume..
|
|
hidden: true,
|
|
});
|
|
}
|
|
|
|
function togetherAIModelsSort(a: ModelDescriptionSchema, b: ModelDescriptionSchema): number {
|
|
if (a.hidden && !b.hidden)
|
|
return 1;
|
|
if (!a.hidden && b.hidden)
|
|
return -1;
|
|
if (a.created !== b.created)
|
|
return (b.created || 0) - (a.created || 0);
|
|
return a.id.localeCompare(b.id);
|
|
}
|
|
|
|
return wireTogetherAIListOutputSchema.parse(wireModels)
|
|
.map(togetherAIModelToModelDescription)
|
|
.sort(togetherAIModelsSort);
|
|
}
|
|
|
|
|
|
// Perplexity
|
|
|
|
const _knownPerplexityChatModels: ModelDescriptionSchema[] = [
|
|
// {
|
|
// id: 'codellama-34b-instruct',
|
|
// label: 'Codellama 34B Instruct (deprecated)',
|
|
// description: 'Will be removed on March 15th, 2024. Try Codellama 70B Instruct as a replacement.',
|
|
// contextWindow: 16384,
|
|
// interfaces: [LLM_IF_OAI_Chat],
|
|
// hidden: true,
|
|
// },
|
|
{
|
|
id: 'codellama-70b-instruct',
|
|
label: 'Codellama 70B Instruct',
|
|
description: 'Code Llama is a collection of pretrained and fine-tuned generative text models. This model is designed for general code synthesis and understanding.',
|
|
contextWindow: 16384,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
},
|
|
// {
|
|
// id: 'llama-2-70b-chat',
|
|
// label: 'Llama 2 70B Chat (deprecated)',
|
|
// description: 'Will be removed on March 15th, 2024. Try mixtral-8x7b-instruct as a replacement.',
|
|
// contextWindow: 4096,
|
|
// interfaces: [LLM_IF_OAI_Chat],
|
|
// hidden: true,
|
|
// },
|
|
{
|
|
id: 'mistral-7b-instruct',
|
|
label: 'Mistral 7B Instruct',
|
|
description: 'The Mistral-7B-Instruct-v0.1 Large Language Model (LLM) is a instruct fine-tuned version of the Mistral-7B-v0.1 generative text model using a variety of publicly available conversation datasets.',
|
|
contextWindow: 16384,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
},
|
|
{
|
|
id: 'mixtral-8x7b-instruct',
|
|
label: 'Mixtral 8x7B Instruct',
|
|
description: 'The Mixtral-8x7B Large Language Model (LLM) is a pretrained generative Sparse Mixture of Experts.',
|
|
contextWindow: 16384,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
},
|
|
// {
|
|
// id: 'pplx-7b-online',
|
|
// label: 'Perplexity 7B Online (deprecated)',
|
|
// description: 'Will be removed on March 15th, 2024. Try Sonar Small Online as a replacement.',
|
|
// contextWindow: 4096,
|
|
// interfaces: [LLM_IF_OAI_Chat],
|
|
// hidden: true,
|
|
// },
|
|
// {
|
|
// id: 'pplx-70b-online',
|
|
// label: 'Perplexity 70B Online (deprecated)',
|
|
// description: 'Will be removed on March 15th, 2024. Try Sonar Medium Online as a replacement.',
|
|
// contextWindow: 4096,
|
|
// interfaces: [LLM_IF_OAI_Chat],
|
|
// hidden: true,
|
|
// },
|
|
// {
|
|
// id: 'pplx-8x7b-online',
|
|
// label: 'Perplexity 8x7B Online (deprecated)',
|
|
// description: 'Will be removed on March 15th, 2024. Try Sonar Medium Online as a replacement.',
|
|
// contextWindow: 4096,
|
|
// interfaces: [LLM_IF_OAI_Chat],
|
|
// hidden: true,
|
|
// },
|
|
// {
|
|
// id: 'pplx-7b-chat',
|
|
// label: 'Perplexity 7B Chat (deprecated)',
|
|
// description: 'Will be removed on March 15th, 2024. Try Sonar Small Chat as a replacement.',
|
|
// contextWindow: 8192,
|
|
// interfaces: [LLM_IF_OAI_Chat],
|
|
// hidden: true,
|
|
// },
|
|
// {
|
|
// id: 'pplx-70b-chat',
|
|
// label: 'Perplexity 70B Chat (deprecated)',
|
|
// description: 'Will be removed on March 15th, 2024. Try Sonar Medium Chat as a replacement.',
|
|
// contextWindow: 4096,
|
|
// interfaces: [LLM_IF_OAI_Chat],
|
|
// hidden: true,
|
|
// },
|
|
// {
|
|
// id: 'pplx-8x7b-chat',
|
|
// label: 'Perplexity 8x7B Chat (deprecated)',
|
|
// description: 'Will be removed on March 15th, 2024. Try Sonar Medium Chat as a replacement.',
|
|
// contextWindow: 4096,
|
|
// interfaces: [LLM_IF_OAI_Chat],
|
|
// hidden: true,
|
|
// },
|
|
{
|
|
id: 'sonar-small-chat',
|
|
label: 'Sonar Small Chat',
|
|
description: 'Sonar Small Chat',
|
|
contextWindow: 16384,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
},
|
|
{
|
|
id: 'sonar-medium-chat',
|
|
label: 'Sonar Medium Chat',
|
|
description: 'Sonar Medium Chat',
|
|
contextWindow: 16384,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
},
|
|
{
|
|
id: 'sonar-small-online',
|
|
label: 'Sonar Small Online 🌐',
|
|
description: 'Sonar Small Online',
|
|
contextWindow: 12000,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
},
|
|
{
|
|
id: 'sonar-medium-online',
|
|
label: 'Sonar Medium Online 🌐',
|
|
description: 'Sonar Medium Online',
|
|
contextWindow: 12000,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
},
|
|
];
|
|
|
|
const perplexityAIModelFamilyOrder = [
|
|
'sonar-medium', 'sonar-small', 'mixtral', 'mistral', 'codellama', 'llama-2', '',
|
|
];
|
|
|
|
export function perplexityAIModelDescriptions() {
|
|
// change this implementation once upstream implements some form of models listing
|
|
return _knownPerplexityChatModels;
|
|
}
|
|
|
|
export function perplexityAIModelSort(a: ModelDescriptionSchema, b: ModelDescriptionSchema): number {
|
|
const aPrefixIndex = perplexityAIModelFamilyOrder.findIndex(prefix => a.id.startsWith(prefix));
|
|
const bPrefixIndex = perplexityAIModelFamilyOrder.findIndex(prefix => b.id.startsWith(prefix));
|
|
// sort by family
|
|
if (aPrefixIndex !== -1 && bPrefixIndex !== -1)
|
|
if (aPrefixIndex !== bPrefixIndex)
|
|
return aPrefixIndex - bPrefixIndex;
|
|
// then by reverse label
|
|
return b.label.localeCompare(a.label);
|
|
}
|
|
|
|
|
|
// Groq - https://console.groq.com/docs/models
|
|
|
|
const _knownGroqModels: ManualMappings = [
|
|
{
|
|
isLatest: true,
|
|
idPrefix: 'llama3-70b-8192',
|
|
label: 'Llama 3 · 70B',
|
|
description: 'LLaMA3 70b developed by Meta with a context window of 8,192 tokens.',
|
|
contextWindow: 8192,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
},
|
|
{
|
|
// isLatest: true,
|
|
idPrefix: 'llama3-8b-8192',
|
|
label: 'Llama 3 · 8B',
|
|
description: 'LLaMA3 8b developed by Meta with a context window of 8,192 tokens.',
|
|
contextWindow: 8192,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
},
|
|
{
|
|
idPrefix: 'llama2-70b-4096',
|
|
label: 'Llama 2 · 70B',
|
|
description: 'LLaMA2 70b developed by Meta with a context window of 4,096 tokens.',
|
|
contextWindow: 4096,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
hidden: true,
|
|
},
|
|
{
|
|
idPrefix: 'mixtral-8x7b-32768',
|
|
label: 'Mixtral 8x7B',
|
|
description: 'Mixtral 8x7b developed by Mistral with a context window of 32,768 tokens.',
|
|
contextWindow: 32768,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
},
|
|
{
|
|
idPrefix: 'gemma-7b-it',
|
|
label: 'Gemma 1.1 · 7B Instruct',
|
|
description: 'Gemma 7b developed by Google with a context window of 8,192 tokens.',
|
|
contextWindow: 8192,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
},
|
|
];
|
|
|
|
export function groqModelToModelDescription(_model: unknown): ModelDescriptionSchema {
|
|
const model = wireGroqModelsListOutputSchema.parse(_model);
|
|
return fromManualMapping(_knownGroqModels, model.id, model.created, undefined, {
|
|
idPrefix: model.id,
|
|
label: model.id.replaceAll(/[_-]/g, ' '),
|
|
description: 'New Model',
|
|
contextWindow: 32768,
|
|
interfaces: [LLM_IF_OAI_Chat],
|
|
hidden: true,
|
|
});
|
|
}
|
|
|
|
export function groqModelSortFn(a: ModelDescriptionSchema, b: ModelDescriptionSchema): number {
|
|
// sort as per their order in the known models
|
|
const aIndex = _knownGroqModels.findIndex(base => a.id.startsWith(base.idPrefix));
|
|
const bIndex = _knownGroqModels.findIndex(base => b.id.startsWith(base.idPrefix));
|
|
if (aIndex !== -1 && bIndex !== -1)
|
|
return aIndex - bIndex;
|
|
return a.id.localeCompare(b.id);
|
|
}
|
|
|
|
|
|
// Helpers
|
|
|
|
type ManualMapping = ({ idPrefix: string, isLatest?: boolean, isPreview?: boolean, isLegacy?: boolean, symLink?: string } & Omit<ModelDescriptionSchema, 'id' | 'created' | 'updated'>);
|
|
type ManualMappings = ManualMapping[];
|
|
|
|
function fromManualMapping(mappings: ManualMappings, id: string, created?: number, updated?: number, fallback?: ManualMapping): ModelDescriptionSchema {
|
|
|
|
// find the closest known model, or fall back, or take the last
|
|
const known = mappings.find(base => id === base.idPrefix)
|
|
|| mappings.find(base => id.startsWith(base.idPrefix))
|
|
|| fallback
|
|
|| mappings[mappings.length - 1];
|
|
|
|
// label for symlinks
|
|
let label = known.label;
|
|
if (known.symLink && id === known.idPrefix)
|
|
label = `🔗 ${known.label} → ${known.symLink/*.replace(known.idPrefix, '')*/}`;
|
|
|
|
// check whether this is a partial map, which indicates an unknown/new variant
|
|
const suffix = id.slice(known.idPrefix.length).trim();
|
|
|
|
// return the model description sheet
|
|
return {
|
|
id,
|
|
label: label
|
|
+ (suffix ? ` [${suffix.replaceAll('-', ' ').trim()}]` : '')
|
|
+ (known.isLatest ? ' 🌟' : '')
|
|
+ (known.isLegacy ? /*' 💩'*/ ' [legacy]' : ''),
|
|
created: created || 0,
|
|
updated: updated || created || 0,
|
|
description: known.description,
|
|
contextWindow: known.contextWindow,
|
|
...(!!known.maxCompletionTokens && { maxCompletionTokens: known.maxCompletionTokens }),
|
|
...(!!known.trainingDataCutoff && { trainingDataCutoff: known.trainingDataCutoff }),
|
|
interfaces: known.interfaces,
|
|
...(!!known.benchmark && { benchmark: known.benchmark }),
|
|
...(!!known.pricing && { pricing: known.pricing }),
|
|
...(!!known.hidden && { hidden: known.hidden }),
|
|
};
|
|
} |