chore: update models 2024-05-07 (#50)

* chore: update perplexity models * chore: update mistral models * chore: update default models
callstack · May 7, 2024 · 8bdf2b2 · 8bdf2b2
1 parent 2c5e4bf
commit 8bdf2b2
Show file tree

Hide file tree

Showing 6 changed files with 19 additions and 23 deletions.
diff --git a/src/commands/chat/providers.tsx b/src/commands/chat/providers.tsx
@@ -1,4 +1,4 @@
-import openAi from '../../engine/providers/openAi.js';
+import openAi from '../../engine/providers/open-ai.js';
 import anthropic from '../../engine/providers/anthropic.js';
 import perplexity from '../../engine/providers/perplexity.js';
 import mistral from '../../engine/providers/mistral.js';

diff --git a/src/commands/chat/state/init.ts b/src/commands/chat/state/init.ts
@@ -26,9 +26,7 @@ export function initChatState(
     ? provider.modelAliases[modelOrAlias] ?? modelOrAlias
     : provider.defaultModel;
 
-  const systemPrompt = !provider.skipSystemPrompt?.includes(model)
-    ? providerFileConfig.systemPrompt ?? DEFAULT_SYSTEM_PROMPT
-    : undefined;
+  const systemPrompt = providerFileConfig.systemPrompt ?? DEFAULT_SYSTEM_PROMPT;
 
   const providerConfig = {
     apiKey: providerFileConfig.apiKey,

diff --git a/src/engine/providers/mistral.ts b/src/engine/providers/mistral.ts
@@ -26,14 +26,15 @@ const Mistral: Provider = {
   modelPricing: {
     'open-mistral-7b': { inputTokensCost: 0.25, outputTokensCost: 0.25 },
     'open-mixtral-8x7b': { inputTokensCost: 0.7, outputTokensCost: 0.7 },
-    'mistral-small-latest': { inputTokensCost: 2, outputTokensCost: 6 },
+    'open-mixtral-8x22b': { inputTokensCost: 2, outputTokensCost: 6 },
+    'mistral-small-latest': { inputTokensCost: 1, outputTokensCost: 3 },
     'mistral-medium-latest': { inputTokensCost: 2.7, outputTokensCost: 8.1 },
-    'mistral-large-latest': { inputTokensCost: 8, outputTokensCost: 24 },
+    'mistral-large-latest': { inputTokensCost: 4, outputTokensCost: 12 },
   },
 
   modelAliases: {
     mistral: 'open-mistral-7b',
-    mixtral: 'open-mixtral-8x7b',
+    mixtral: 'open-mixtral-8x22b',
     small: 'mistral-small-latest',
     medium: 'mistral-medium-latest',
     large: 'mistral-large-latest',

diff --git a/src/engine/providers/openAi.ts → src/engine/providers/open-ai.ts b/src/engine/providers/openAi.ts → src/engine/providers/open-ai.ts
diff --git a/src/engine/providers/perplexity.ts b/src/engine/providers/perplexity.ts
@@ -10,38 +10,37 @@ const Perplexity: Provider = {
   apiKeyUrl: 'https://perplexity.ai/settings/api',
 
   // Perplexity models: https://docs.perplexity.ai/docs/model-cards
-  defaultModel: 'sonar-medium-chat',
+  defaultModel: 'llama-3-sonar-large-32k-chat',
 
   // Price per 1M tokens [input, output], per 1k requests.
   // Source: https://docs.perplexity.ai/docs/model-cards
   // Source: https://docs.perplexity.ai/docs/pricing
   modelPricing: {
-    'sonar-small-chat': { inputTokensCost: 0.2, outputTokensCost: 0.2 },
-    'sonar-medium-chat': { inputTokensCost: 0.6, outputTokensCost: 0.6 },
-    'sonar-small-online': {
+    'llama-3-sonar-small-32k-chat': { inputTokensCost: 0.2, outputTokensCost: 0.2 },
+    'llama-3-sonar-small-32k-online': {
       inputTokensCost: 0.2,
       outputTokensCost: 0.2,
       requestsCost: 5,
     },
-    'sonar-medium-online': {
+    'llama-3-sonar-large-32k-chat': { inputTokensCost: 0.6, outputTokensCost: 0.6 },
+    'llama-3-sonar-large-32k-online': {
       inputTokensCost: 0.6,
       outputTokensCost: 0.6,
       requestsCost: 5,
     },
-    'codellama-70b-instruct': { inputTokensCost: 1, outputTokensCost: 1 },
-    'mistral-7b-instruct': { inputTokensCost: 0.2, outputTokensCost: 0.2 },
+    'llama-3-8b-instruct': { inputTokensCost: 0.2, outputTokensCost: 0.2 },
+    'llama-3-70b-instruct': { inputTokensCost: 1, outputTokensCost: 1 },
     'mixtral-8x7b-instruct': { inputTokensCost: 0.6, outputTokensCost: 0.6 },
   },
 
   modelAliases: {
-    online: 'sonar-medium-online',
-    codellama: 'codellama-70b-instruct',
-    mistral: 'mistral-7b-instruct',
-    mixtral: 'mixtral-8x7b-instruct',
+    'small': 'llama-3-sonar-small-32k-chat',
+    'large': 'llama-3-sonar-large-32k-chat',
+    'online': 'llama-3-sonar-large-32k-online',
+    'llama-3': 'llama-3-70b-instruct',
+    'mixtral': 'mixtral-8x7b-instruct',
   },
 
-  skipSystemPrompt: ['sonar-small-online', 'sonar-medium-online'],
-
   getChatCompletion: async (config: ProviderConfig, messages: Message[]) => {
     const api = new OpenAI({
       apiKey: config.apiKey,

diff --git a/src/engine/providers/provider.ts b/src/engine/providers/provider.ts
@@ -1,6 +1,6 @@
 import type { Message, ModelResponse, ModelResponseUpdate } from '../inference.js';
 import type { ProviderConfig } from './config.js';
-import openAi from './openAi.js';
+import openAi from './open-ai.js';
 import perplexity from './perplexity.js';
 import anthropic from './anthropic.js';
 import mistral from './mistral.js';
@@ -17,8 +17,6 @@ export interface Provider {
   modelPricing: Record<string, ModelPricing>;
   modelAliases: Record<string, string>;
 
-  skipSystemPrompt?: string[];
-
   getChatCompletion: (config: ProviderConfig, messages: Message[]) => Promise<ModelResponse>;
   getChatCompletionStream?: (
     config: ProviderConfig,