chore: move to byorg framework

package.json

@@ -41,7 +41,11 @@
     "registry": "https://registry.npmjs.org/"
   },
   "dependencies": {
+    "@ai-sdk/anthropic": "^0.0.54",
+    "@ai-sdk/mistral": "^0.0.46",
+    "@ai-sdk/openai": "^0.0.71",
     "@anthropic-ai/sdk": "^0.26.1",
+    "@callstack/byorg-core": "0.1.2",
     "@inkjs/ui": "^1.0.0",
     "@mistralai/mistralai": "^1.0.2",
     "chalk": "^5.3.0",

src/commands/chat/providers.tsx

@@ -2,10 +2,14 @@ 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';
-import { getProvider, type Provider, type ProviderName } from '../../engine/providers/provider.js';
+import {
+  getProvider,
+  type ProviderInfo,
+  type ProviderName,
+} from '../../engine/providers/provider-info.js';
 import type { ConfigFile } from '../../config-file.js';
 
-export const providerOptionMapping: Record<string, Provider> = {
+export const providerOptionMapping: Record<string, ProviderInfo> = {
   openai: openAi,
   anthropic,
   anth: anthropic,
@@ -16,7 +20,7 @@ export const providerOptionMapping: Record<string, Provider> = {
 
 export const providerOptions = Object.keys(providerOptionMapping);
 
-export function resolveProviderFromOption(providerOption: string): Provider {
+export function resolveProviderInfoFromOption(providerOption: string): ProviderInfo {
   const provider = providerOptionMapping[providerOption];
   if (!provider) {
     throw new Error(`Provider not found: ${providerOption}.`);
@@ -25,7 +29,7 @@ export function resolveProviderFromOption(providerOption: string): Provider {
   return provider;
 }
 
-export function getDefaultProvider(config: ConfigFile): Provider {
+export function getDefaultProviderInfo(config: ConfigFile): ProviderInfo {
   const providerNames = Object.keys(config.providers) as ProviderName[];
   const providerName = providerNames ? providerNames[0] : undefined;
 

src/commands/chat/state/actions.ts

@@ -1,39 +1,36 @@
-import type { ModelResponse } from '../../../engine/inference.js';
+import { type AssistantMessage, type AssistantResponse } from '@callstack/byorg-core';
 import { useChatState, type MessageLevel } from './state.js';
 
-export function addUserMessage(text: string) {
+export function addUserMessage(content: string) {
   useChatState.setState((state) => {
     return {
       activeView: null,
-      contextMessages: [...state.contextMessages, { role: 'user', content: text }],
-      chatMessages: [...state.chatMessages, { type: 'user', text }],
+      contextMessages: [...state.contextMessages, { role: 'user', content }],
+      chatMessages: [...state.chatMessages, { role: 'user', content }],
     };
   });
 }
 
-export function addAiResponse(response: ModelResponse) {
+export function addAssistantResponse(response: AssistantResponse) {
   useChatState.setState((state) => {
-    const outputMessages = {
-      type: 'ai',
-      text: response.message.content,
-      responseTime: response.responseTime,
-      usage: response.usage,
-      data: response.data,
-    } as const;
+    const message: AssistantMessage = {
+      role: 'assistant',
+      content: response.content,
+    };
 
     return {
       activeView: null,
-      contextMessages: [...state.contextMessages, response.message],
-      chatMessages: [...state.chatMessages, outputMessages],
+      contextMessages: [...state.contextMessages, message],
+      chatMessages: [...state.chatMessages, response],
     };
   });
 }
 
-export function addProgramMessage(text: string, level: MessageLevel = 'info') {
+export function addProgramMessage(content: string, level: MessageLevel = 'info') {
   useChatState.setState((state) => {
     return {
       activeView: null,
-      chatMessages: [...state.chatMessages, { type: 'program', level, text }],
+      chatMessages: [...state.chatMessages, { role: 'program', level, content }],
     };
   });
 }
@@ -57,7 +54,7 @@ export function forgetContextMessages() {
       contextMessages: [],
       chatMessages: [
         ...state.chatMessages,
-        { type: 'program', level: 'info', text: 'AI will forget previous messages.' },
+        { role: 'program', level: 'info', content: 'AI will forget previous messages.' },
       ],
     };
   });
@@ -92,7 +89,7 @@ export function triggerExit() {
     return {
       activeView: null,
       shouldExit: true,
-      chatMessages: [...state.chatMessages, { type: 'program', level: 'info', text: 'Bye! 👋' }],
+      chatMessages: [...state.chatMessages, { role: 'program', level: 'info', content: 'Bye! 👋' }],
     };
   });
 }
@@ -103,7 +100,7 @@ export function setVerbose(verbose: boolean) {
       verbose,
       chatMessages: [
         ...state.chatMessages,
-        { type: 'program', level: 'info', text: `Verbose mode: ${verbose ? 'on' : 'off'}` },
+        { role: 'program', level: 'info', content: `Verbose mode: ${verbose ? 'on' : 'off'}` },
       ],
     };
   });

src/commands/chat/state/init.ts

@@ -1,9 +1,9 @@
+import { createApp, loggingPlugin, type Message } from '@callstack/byorg-core';
 import { type ConfigFile } from '../../../config-file.js';
 import { DEFAULT_SYSTEM_PROMPT } from '../../../default-config.js';
 import type { ResponseStyle } from '../../../engine/providers/config.js';
-import type { Message } from '../../../engine/inference.js';
 import type { PromptOptions } from '../prompt-options.js';
-import { getDefaultProvider, resolveProviderFromOption } from '../providers.js';
+import { getDefaultProviderInfo, resolveProviderInfoFromOption } from '../providers.js';
 import { filterOutApiKey, handleInputFile } from '../utils.js';
 import { useChatState, type ChatMessage, type ChatState } from './state.js';
 
@@ -12,19 +12,19 @@ export function initChatState(
   configFile: ConfigFile,
   initialPrompt: string,
 ) {
-  const provider = options.provider
-    ? resolveProviderFromOption(options.provider)
-    : getDefaultProvider(configFile);
+  const providerInfo = options.provider
+    ? resolveProviderInfoFromOption(options.provider)
+    : getDefaultProviderInfo(configFile);
 
-  const providerFileConfig = configFile.providers[provider.name];
+  const providerFileConfig = configFile.providers[providerInfo.name];
   if (!providerFileConfig) {
-    throw new Error(`Provider config not found: ${provider.name}.`);
+    throw new Error(`Provider config not found: ${providerInfo.name}.`);
   }
 
   const modelOrAlias = options.model ?? providerFileConfig.model;
   const model = modelOrAlias
-    ? (provider.modelAliases[modelOrAlias] ?? modelOrAlias)
-    : provider.defaultModel;
+    ? (providerInfo.modelAliases[modelOrAlias] ?? modelOrAlias)
+    : providerInfo.defaultModel;
 
   const systemPrompt = providerFileConfig.systemPrompt ?? DEFAULT_SYSTEM_PROMPT;
 
@@ -40,47 +40,55 @@ export function initChatState(
 
   if (modelOrAlias != null && modelOrAlias !== model) {
     outputMessages.push({
-      type: 'program',
+      role: 'program',
       level: 'debug',
-      text: `Resolved model alias "${modelOrAlias}" to "${model}".`,
+      content: `Resolved model alias "${modelOrAlias}" to "${model}".`,
     });
   }
 
   outputMessages.push({
-    type: 'program',
+    role: 'program',
     level: 'debug',
-    text: `Loaded config: ${JSON.stringify(providerConfig, filterOutApiKey, 2)}`,
+    content: `Loaded config: ${JSON.stringify(providerConfig, filterOutApiKey, 2)}`,
   });
 
   if (options.file) {
-    const { systemMessage, costWarning, costInfo } = handleInputFile(
-      options.file,
-      providerConfig,
-      provider,
-    );
+    const {
+      systemPrompt: fileSystemPrompt,
+      costWarning,
+      costInfo,
+    } = handleInputFile(options.file, providerConfig, providerInfo);
+
+    providerConfig.systemPrompt += `\n\n${fileSystemPrompt}`;
 
-    contextMessages.push(systemMessage);
     if (costWarning) {
-      outputMessages.push({ type: 'program', level: 'warning', text: costWarning });
+      outputMessages.push({ role: 'program', level: 'warning', content: costWarning });
     } else if (costInfo) {
-      outputMessages.push({ type: 'program', level: 'info', text: costInfo });
+      outputMessages.push({ role: 'program', level: 'info', content: costInfo });
     }
   }
 
   if (initialPrompt) {
     contextMessages.push({ role: 'user', content: initialPrompt });
-    outputMessages.push({ type: 'user', text: initialPrompt });
+    outputMessages.push({ role: 'user', content: initialPrompt });
   }
 
+  const app = createApp({
+    chatModel: providerInfo.getChatModel(providerConfig),
+    systemPrompt: () => providerConfig.systemPrompt,
+    plugins: [loggingPlugin],
+  });
+
   const state: ChatState = {
     activeView: null,
-    provider,
+    provider: providerInfo,
     providerConfig,
     contextMessages,
     chatMessages: outputMessages,
     verbose: options.verbose ?? false,
     shouldExit: false,
     stream: options.stream ?? true,
+    app,
   };
 
   useChatState.setState(state);

src/commands/chat/state/state.ts

@@ -1,43 +1,30 @@
 import { create } from 'zustand';
+import type { Application, AssistantResponse, Message, UserMessage } from '@callstack/byorg-core';
 import type { ProviderConfig } from '../../../engine/providers/config.js';
-import type { Provider } from '../../../engine/providers/provider.js';
-import type { Message, ModelUsage } from '../../../engine/inference.js';
+import type { ProviderInfo } from '../../../engine/providers/provider-info.js';
 
 export interface ChatState {
-  provider: Provider;
+  provider: ProviderInfo;
   providerConfig: ProviderConfig;
   contextMessages: Message[];
   chatMessages: ChatMessage[];
   activeView: ActiveView;
   verbose: boolean;
   shouldExit: boolean;
   stream: boolean;
+  app: Application;
 }
 
-export type ChatMessage = UserChatMessage | AiChatMessage | ProgramChatMessage;
+export type ChatMessage = UserMessage | AssistantResponse | ProgramOutput;
 
-export interface UserChatMessage {
-  type: 'user';
-  text: string;
-}
-
-export interface AiChatMessage {
-  type: 'ai';
-  text: string;
-  responseTime: number;
-  usage: ModelUsage;
-  cost?: number;
-  data?: unknown;
+export interface ProgramOutput {
+  role: 'program';
+  content: string;
+  level: MessageLevel;
 }
 
 export type MessageLevel = 'debug' | 'info' | 'warning' | 'error';
 
-export interface ProgramChatMessage {
-  type: 'program';
-  level: MessageLevel;
-  text: string;
-}
-
 type ActiveView = 'info' | 'help' | null;
 
 // @ts-expect-error lazy init

src/commands/chat/ui/AiResponseLoader.tsx → src/commands/chat/ui/AssistantResponseLoader.tsx

@@ -4,11 +4,11 @@ import { TextSpinner } from '../../../components/TextSpinner.js';
 import { colors } from '../../../theme/colors.js';
 import { texts } from '../texts.js';
 
-interface AiResponseLoaderProps {
+interface AssistantResponseLoaderProps {
   text?: string;
 }
 
-export function AiResponseLoader({ text }: AiResponseLoaderProps) {
+export function AssistantResponseLoader({ text }: AssistantResponseLoaderProps) {
   return (
     <Text color={colors.assistant}>
       <Text color={colors.assistant}>{texts.assistantLabel}</Text>

src/commands/chat/ui/ChatUi.tsx

@@ -1,11 +1,10 @@
 import React, { useEffect, useState } from 'react';
 import { Box } from 'ink';
 import { ExitApp } from '../../../components/ExitApp.js';
-import type { ModelResponseUpdate } from '../../../engine/inference.js';
 import { extractErrorMessage } from '../../../output.js';
 import { processChatCommand } from '../chat-commands.js';
 import {
-  addAiResponse,
+  addAssistantResponse,
   addProgramMessage,
   addUserMessage,
   hideActiveView,
@@ -18,7 +17,7 @@ import { HelpOutput } from './HelpOutput.js';
 import { StatusBar } from './StatusBar.js';
 import { InfoOutput } from './InfoOutput.js';
 import { ChatMessageList } from './list/ChatMessageList.js';
-import { AiResponseLoader } from './AiResponseLoader.js';
+import { AssistantResponseLoader } from './AssistantResponseLoader.js';
 
 export function ChatUi() {
   const contextMessages = useChatState((state) => state.contextMessages);
@@ -59,7 +58,7 @@ export function ChatUi() {
       {activeView === 'help' && <HelpOutput />}
       {activeView === 'info' && <InfoOutput />}
 
-      {isLoading ? <AiResponseLoader text={loadedResponse} /> : null}
+      {isLoading ? <AssistantResponseLoader text={loadedResponse} /> : null}
       {showInput && <UserMessageInput onSubmit={handleSubmit} />}
 
       <StatusBar />
@@ -69,8 +68,7 @@ export function ChatUi() {
 }
 
 function useAiResponse() {
-  const provider = useChatState((state) => state.provider);
-  const providerConfig = useChatState((state) => state.providerConfig);
+  const app = useChatState((state) => state.app);
   const stream = useChatState((state) => state.stream);
 
   const [isLoading, setLoading] = useState(false);
@@ -82,16 +80,20 @@ function useAiResponse() {
       setLoadedResponse(undefined);
 
       const messages = useChatState.getState().contextMessages;
-      if (stream && provider.getChatCompletionStream != null) {
-        const response = await provider.getChatCompletionStream(
-          providerConfig,
-          messages,
-          (update: ModelResponseUpdate) => setLoadedResponse(update.content),
-        );
-        addAiResponse(response);
+      if (stream) {
+        const { response } = await app.processMessages(messages, {
+          onPartialResponse: (update) => setLoadedResponse(update),
+        });
+
+        if (response.role == 'assistant') {
+          addAssistantResponse(response);
+        } else {
+          addProgramMessage(response.content);
+        }
       } else {
-        const response = await provider.getChatCompletion(providerConfig, messages);
-        addAiResponse(response);
+        throw new Error('Non-Stream mode is not supported yet');
+        // const response = await provider.getChatCompletion(providerConfig, messages);
+        // addAiResponse(response);
       }
     } catch (error) {
       // We cannot leave unanswered user message in context, as there is no AI response for it.

src/commands/chat/ui/StatusBar.tsx

@@ -1,57 +1,65 @@
 import React, { useMemo } from 'react';
 import { Box, Text } from 'ink';
-import type { ModelUsage } from '../../../engine/inference.js';
 import { formatCost, formatSpeed, formatTokenCount } from '../../../format.js';
 import { calculateUsageCost } from '../../../engine/session.js';
 import { useChatState, type ChatMessage } from '../state/state.js';
 
+type TotalUsage = {
+  inputTokens: number;
+  outputTokens: number;
+  requests: number;
+  responseTime: number;
+  model: string | undefined;
+};
+
 export function StatusBar() {
   const verbose = useChatState((state) => state.verbose);
   const items = useChatState((state) => state.chatMessages);
   const provider = useChatState((state) => state.provider);
   const providerConfig = useChatState((state) => state.providerConfig);
 
   const totalUsage = useMemo(() => calculateTotalUsage(items), [items]);
-  const totalTime = useMemo(() => calculateTotalResponseTime(items), [items]);
 
-  const modelPricing = provider.modelPricing[providerConfig.model];
+  const model = totalUsage.model || providerConfig.model;
+  const modelPricing = provider.modelPricing[model] ?? provider.modelPricing[providerConfig.model];
   const totalCost = calculateUsageCost(totalUsage, modelPricing) ?? 0;
 
   return (
     <Box flexDirection="row" marginTop={1}>
       <Text color={'gray'}>
-        LLM: {provider.label}/{providerConfig.model} - Total Cost:{' '}
-        {verbose ? formatVerboseStats(totalCost, totalUsage, totalTime) : formatCost(totalCost)}
+        LLM: {provider.label}/{model} - Total Cost:{' '}
+        {verbose ? formatVerboseStats(totalCost, totalUsage) : formatCost(totalCost)}
       </Text>
     </Box>
   );
 }
 
-function formatVerboseStats(cost: number, usage: ModelUsage, time: number) {
+function formatVerboseStats(cost: number, usage: TotalUsage) {
   const usageOutput = usage
-    ? ` (tokens: ${formatTokenCount(usage.inputTokens)} in + ${formatTokenCount(usage.outputTokens)} out, requests: ${usage.requests}, speed: ${formatSpeed(usage.outputTokens, time)})`
+    ? ` (tokens: ${formatTokenCount(usage.inputTokens)} in + ${formatTokenCount(
+        usage.outputTokens,
+      )} out, requests: ${usage.requests}, speed: ${formatSpeed(usage.outputTokens, usage.responseTime)})`
     : '';
   return `${formatCost(cost)}${usageOutput}`;
 }
 
-function calculateTotalUsage(messages: ChatMessage[]) {
-  const usage: ModelUsage = { inputTokens: 0, outputTokens: 0, requests: 0 };
+function calculateTotalUsage(messages: ChatMessage[]): TotalUsage {
+  const usage: TotalUsage = {
+    inputTokens: 0,
+    outputTokens: 0,
+    requests: 0,
+    responseTime: 0,
+    model: undefined,
+  };
   messages.forEach((message) => {
-    if (message.type === 'ai') {
+    if (message.role === 'assistant') {
       usage.inputTokens += message.usage?.inputTokens ?? 0;
       usage.outputTokens += message.usage?.outputTokens ?? 0;
       usage.requests += message.usage?.requests ?? 0;
+      usage.responseTime += message.usage?.responseTime ?? 0;
+      usage.model = message.usage?.model ?? usage.model;
     }
   });
-  return usage;
-}
 
-function calculateTotalResponseTime(messages: ChatMessage[]) {
-  let total = 0;
-  messages.forEach((message) => {
-    if (message.type === 'ai') {
-      total += message.responseTime ?? 0;
-    }
-  });
-  return total;
+  return usage;
 }

src/commands/chat/ui/list/AiChatMessageItem.tsx → src/commands/chat/ui/list/AssistantResponseItem.tsx

@@ -1,32 +1,32 @@
 import React from 'react';
 import { Text } from 'ink';
+import type { AssistantResponse } from '@callstack/byorg-core';
 import { formatSpeed, formatTime } from '../../../../format.js';
 import { colors } from '../../../../theme/colors.js';
-import { useChatState, type AiChatMessage } from '../../state/state.js';
+import { useChatState } from '../../state/state.js';
 import { texts } from '../../texts.js';
 
-interface AiChatMessageItemProps {
-  message: AiChatMessage;
+interface AssistantResponseItemProps {
+  message: AssistantResponse;
 }
 
-export function AiChatMessageItem({ message }: AiChatMessageItemProps) {
+export function AssistantResponseItem({ message }: AssistantResponseItemProps) {
   const verbose = useChatState((state) => state.verbose);
 
   return (
     <>
       <Text color={colors.assistant}>
         <Text>{texts.assistantLabel}</Text>
-        <Text>{message.text}</Text>
+        <Text>{message.content}</Text>
 
-        {verbose && message.responseTime != null ? (
+        {verbose && message.usage.responseTime != null ? (
           <Text color={colors.info}>
             {' '}
-            ({formatTime(message.responseTime)},{' '}
-            {formatSpeed(message.usage?.outputTokens, message.responseTime)})
+            ({formatTime(message.usage.responseTime)},{' '}
+            {formatSpeed(message.usage?.outputTokens, message.usage.responseTime)})
           </Text>
         ) : null}
       </Text>
-      {verbose ? <Text color={colors.debug}>{JSON.stringify(message.data, null, 2)}</Text> : null}
       <Text> {/* Add a newline */}</Text>
     </>
   );

src/commands/chat/ui/list/ChatMessageList.tsx

@@ -1,9 +1,9 @@
 import * as React from 'react';
 import { Box, Static } from 'ink';
 import { useChatState, type ChatMessage } from '../../state/state.js';
-import { ProgramChatMessageItem } from './ProgramChatMessageItem.js';
-import { UserChatMessageItem } from './UserChatMessageItem.js';
-import { AiChatMessageItem } from './AiChatMessageItem.js';
+import { UserMessageItem } from './UserChatMessageItem.js';
+import { AssistantResponseItem } from './AssistantResponseItem.js';
+import { ProgramOutputItem } from './ProgramChatMessageItem.js';
 
 export function ChatMessageList() {
   const messages = useChatState((state) => state.chatMessages);
@@ -21,16 +21,16 @@ export function ChatMessageList() {
 }
 
 function renderMessage(message: ChatMessage, index: number): React.ReactNode {
-  if (message.type === 'user') {
-    return <UserChatMessageItem key={index} message={message} />;
+  if (message.role === 'user') {
+    return <UserMessageItem key={index} message={message} />;
   }
 
-  if (message.type === 'ai') {
-    return <AiChatMessageItem key={index} message={message} />;
+  if (message.role === 'assistant') {
+    return <AssistantResponseItem key={index} message={message} />;
   }
 
-  if (message.type === 'program') {
-    return <ProgramChatMessageItem key={index} output={message} />;
+  if (message.role === 'program') {
+    return <ProgramOutputItem key={index} output={message} />;
   }
 
   return null;

src/commands/chat/ui/list/ProgramChatMessageItem.tsx

@@ -1,29 +1,29 @@
 import React from 'react';
 import { Text } from 'ink';
 import { colors } from '../../../../theme/colors.js';
-import { useChatState, type ProgramChatMessage } from '../../state/state.js';
+import { useChatState, type ProgramOutput } from '../../state/state.js';
 
-interface ProgramChatMessageItemProps {
-  output: ProgramChatMessage;
+interface ProgramOutputItemProps {
+  output: ProgramOutput;
 }
 
-export function ProgramChatMessageItem({ output }: ProgramChatMessageItemProps) {
+export function ProgramOutputItem({ output }: ProgramOutputItemProps) {
   const verbose = useChatState((state) => state.verbose);
 
   if (output.level === 'error') {
-    return <Text color={colors.error}>{output.text}</Text>;
+    return <Text color={colors.error}>{output.content}</Text>;
   }
 
   if (output.level === 'warning') {
-    return <Text color={colors.warning}>{output.text}</Text>;
+    return <Text color={colors.warning}>{output.content}</Text>;
   }
 
   if (output.level === 'info') {
-    return <Text color={colors.info}>{output.text}</Text>;
+    return <Text color={colors.info}>{output.content}</Text>;
   }
 
   if (verbose && output.level === 'debug') {
-    return <Text color={colors.debug}>{output.text}</Text>;
+    return <Text color={colors.debug}>{output.content}</Text>;
   }
 
   return null;

src/commands/chat/ui/list/UserChatMessageItem.tsx

@@ -1,18 +1,18 @@
 import React from 'react';
 import { Text } from 'ink';
+import type { UserMessage } from '@callstack/byorg-core';
 import { colors } from '../../../../theme/colors.js';
-import { type UserChatMessage } from '../../state/state.js';
 import { texts } from '../../texts.js';
 
 interface UserChatMessageItemProps {
-  message: UserChatMessage;
+  message: UserMessage;
 }
 
-export function UserChatMessageItem({ message }: UserChatMessageItemProps) {
+export function UserMessageItem({ message }: UserChatMessageItemProps) {
   return (
     <Text color={colors.user}>
       <Text>{texts.userLabel}</Text>
-      <Text>{message.text}</Text>
+      <Text>{message.content}</Text>
     </Text>
   );
 }

src/commands/chat/utils.ts

@@ -1,33 +1,34 @@
 import * as fs from 'fs';
 import * as path from 'path';
 import * as os from 'os';
+import type { Message } from '@callstack/byorg-core';
 import {
   DEFAULT_FILE_PROMPT,
   FILE_COST_WARNING,
   FILE_TOKEN_COUNT_WARNING,
 } from '../../default-config.js';
 import { calculateUsageCost } from '../../engine/session.js';
 import { getTokensCount } from '../../engine/tokenizer.js';
-import type { Message, SystemMessage } from '../../engine/inference.js';
 import type { ProviderConfig } from '../../engine/providers/config.js';
-import type { Provider } from '../../engine/providers/provider.js';
+import type { ProviderInfo } from '../../engine/providers/provider-info.js';
 import { formatCost, formatTokenCount } from '../../format.js';
 import {
   getConversationStoragePath,
   getDefaultFilename,
   getUniqueFilename,
 } from '../../file-utils.js';
+import { texts } from './texts.js';
 
 interface HandleInputFileResult {
-  systemMessage: SystemMessage;
+  systemPrompt: string;
   costWarning: string | null;
   costInfo: string | null;
 }
 
 export function handleInputFile(
   inputFile: string,
   config: ProviderConfig,
-  provider: Provider,
+  provider: ProviderInfo,
 ): HandleInputFileResult {
   const filePath = path.resolve(inputFile.replace('~', os.homedir()));
 
@@ -59,7 +60,7 @@ export function handleInputFile(
   );
 
   return {
-    systemMessage: { role: 'system', content },
+    systemPrompt: content,
     costWarning,
     costInfo,
   };
@@ -92,11 +93,9 @@ export function saveConversation(messages: Message[]) {
 function roleToLabel(role: Message['role']): string {
   switch (role) {
     case 'user':
-      return 'me';
+      return texts.userLabel;
     case 'assistant':
-      return 'ai';
-    case 'system':
-      return 'system';
+      return texts.assistantLabel;
     default:
       return role;
   }

src/commands/init/ui/InitUi.tsx

@@ -1,6 +1,6 @@
 import React, { useState } from 'react';
 import { Box, Newline, Text } from 'ink';
-import { type Provider } from '../../../engine/providers/provider.js';
+import { type ProviderInfo } from '../../../engine/providers/provider-info.js';
 import { writeConfigFile } from '../../../config-file.js';
 import { colors } from '../../../theme/colors.js';
 import { ExitApp } from '../../../components/ExitApp.js';
@@ -17,10 +17,10 @@ export function InitUi({ hasConfig }: InitUiProps) {
   const [step, setStep] = useState(0);
 
   const [overwriteConfig, setOverwriteConfig] = useState(true);
-  const [provider, setProvider] = useState<Provider>();
+  const [provider, setProvider] = useState<ProviderInfo>();
   const [hasKey, setHasKey] = useState<boolean>();
 
-  const writeConfig = (provider?: Provider, apiKey?: string) => {
+  const writeConfig = (provider?: ProviderInfo, apiKey?: string) => {
     if (!provider || !apiKey) {
       return;
     }

src/commands/init/ui/SelectProviderStep.tsx

@@ -4,20 +4,20 @@ import { Select } from '@inkjs/ui';
 import {
   getProvider,
   providers,
-  type Provider,
+  type ProviderInfo,
   type ProviderName,
-} from '../../../engine/providers/provider.js';
+} from '../../../engine/providers/provider-info.js';
 import { colors } from '../../../theme/colors.js';
 
 interface SelectProviderStepProps {
   label: string;
-  onSelect: (provider: Provider) => void;
+  onSelect: (provider: ProviderInfo) => void;
 }
 
 const providerItems = providers.map((p) => ({ label: p.label, value: p.name }));
 
 export function SelectProviderStep({ label, onSelect }: SelectProviderStepProps) {
-  const [value, setValue] = useState<Provider>();
+  const [value, setValue] = useState<ProviderInfo>();
 
   const handleSelect = (name: ProviderName) => {
     const provider = getProvider(name);

src/config-file.ts

@@ -2,7 +2,7 @@ import * as fs from 'fs';
 import * as os from 'os';
 import * as path from 'path';
 import { z } from 'zod';
-import { providerNames } from './engine/providers/provider.js';
+import { providerNames } from './engine/providers/provider-info.js';
 
 const LEGACY_CONFIG_FILENAME = '.airc';
 const CONFIG_FILENAME = '.airc.json';

src/engine/providers/anthropic.ts

@@ -1,16 +1,9 @@
-import AnthropicAPI from '@anthropic-ai/sdk';
-import {
-  type AiMessage,
-  type Message,
-  type ModelResponseUpdate,
-  type UserMessage,
-} from '../inference.js';
-import { responseStyles, type ProviderConfig } from './config.js';
-import type { Provider } from './provider.js';
+import { createAnthropic } from '@ai-sdk/anthropic';
+import { VercelChatModelAdapter } from '@callstack/byorg-core';
+import { type ProviderConfig } from './config.js';
+import type { ProviderInfo } from './provider-info.js';
 
-type ModelMessage = UserMessage | AiMessage;
-
-const Anthropic: Provider = {
+const Anthropic: ProviderInfo = {
   label: 'Anthropic',
   name: 'anthropic',
   apiKeyUrl: 'https://console.anthropic.com/settings/keys',
@@ -39,93 +32,11 @@ const Anthropic: Provider = {
     opus: 'claude-3-opus-20240229',
   },
 
-  getChatCompletion: async (config: ProviderConfig, messages: Message[]) => {
-    const api = new AnthropicAPI({
-      apiKey: config.apiKey,
-    });
-
-    const nonSystemMessages = messages.filter((m) => m.role !== 'system') as ModelMessage[];
-
-    const systemMessagesContent = messages.filter((m) => m.role === 'system').map((m) => m.content);
-    const systemPrompt = [config.systemPrompt, ...systemMessagesContent].join('\n\n');
-
-    const startTime = performance.now();
-    const response = await api.messages.create({
-      messages: nonSystemMessages,
-      model: config.model,
-      max_tokens: 1024,
-      system: systemPrompt,
-      ...responseStyles[config.responseStyle],
+  getChatModel: (config: ProviderConfig) => {
+    const client = createAnthropic({ apiKey: config.apiKey });
+    return new VercelChatModelAdapter({
+      languageModel: client.languageModel(config.model),
     });
-    const responseTime = performance.now() - startTime;
-
-    const firstContent = response.content[0];
-    if (firstContent?.type !== 'text') {
-      throw new Error(`Received unexpected content type from Anthropic API: ${firstContent?.type}`);
-    }
-
-    return {
-      message: {
-        role: 'assistant',
-        content: firstContent.text,
-      },
-      usage: {
-        inputTokens: response.usage.input_tokens,
-        outputTokens: response.usage.output_tokens,
-        requests: 1,
-      },
-      responseTime,
-      responseModel: response.model,
-      data: response,
-    };
-  },
-
-  getChatCompletionStream: async function (
-    config: ProviderConfig,
-    messages: Message[],
-    onResponseUpdate: (update: ModelResponseUpdate) => void,
-  ) {
-    const api = new AnthropicAPI({
-      apiKey: config.apiKey,
-    });
-
-    const nonSystemMessages = messages.filter((m) => m.role !== 'system') as ModelMessage[];
-
-    const systemMessagesContent = messages.filter((m) => m.role === 'system').map((m) => m.content);
-    const systemPrompt = [config.systemPrompt, ...systemMessagesContent].join('\n\n');
-
-    const startTime = performance.now();
-    let content = '';
-    const stream = api.messages
-      .stream({
-        messages: nonSystemMessages,
-        model: config.model,
-        max_tokens: 1024,
-        system: systemPrompt,
-        ...responseStyles[config.responseStyle],
-      })
-      .on('text', (text) => {
-        content += text;
-        onResponseUpdate({ content });
-      });
-
-    const response = await stream.finalMessage();
-    const responseTime = performance.now() - startTime;
-
-    return {
-      message: {
-        role: 'assistant',
-        content,
-      },
-      usage: {
-        inputTokens: response.usage.input_tokens,
-        outputTokens: response.usage.output_tokens,
-        requests: 1,
-      },
-      responseTime,
-      responseModel: response.model,
-      data: response,
-    };
   },
 };
 

src/engine/providers/mistral.ts

@@ -1,11 +1,8 @@
-import { Mistral as MistralClient } from '@mistralai/mistralai';
-import type { CompletionEvent } from '@mistralai/mistralai/models/components/completionevent.js';
-import { type Message, type ModelResponseUpdate } from '../inference.js';
-import { estimateInputTokens, estimateOutputTokens } from '../tokenizer.js';
-import { responseStyles, type ProviderConfig } from './config.js';
-import type { Provider } from './provider.js';
+import { createMistral } from '@ai-sdk/mistral';
+import { VercelChatModelAdapter } from '@callstack/byorg-core';
+import type { ProviderInfo } from './provider-info.js';
 
-const Mistral: Provider = {
+const Mistral: ProviderInfo = {
   label: 'Mistral',
   name: 'mistral',
   apiKeyUrl: 'https://console.mistral.ai/api-keys/',
@@ -40,105 +37,12 @@ const Mistral: Provider = {
     codestral: 'codestral-latest',
   },
 
-  getChatCompletion: async (config: ProviderConfig, messages: Message[]) => {
-    const api = new MistralClient({ apiKey: config.apiKey });
-    const allMessages = getMessages(config, messages);
-
-    const startTime = performance.now();
-    const response = await api.chat.complete({
-      model: config.model,
-      messages: allMessages,
-      ...getMistralResponseStyle(config),
+  getChatModel(config) {
+    const client = createMistral({ apiKey: config.apiKey });
+    return new VercelChatModelAdapter({
+      languageModel: client.languageModel(config.model),
     });
-    const responseTime = performance.now() - startTime;
-
-    return {
-      message: {
-        role: 'assistant',
-        content: response.choices?.[0]?.message?.content ?? '',
-      },
-      usage: {
-        inputTokens: response.usage.promptTokens,
-        outputTokens: response.usage.completionTokens,
-        requests: 1,
-      },
-      responseTime,
-      responseModel: response.model,
-      data: response,
-    };
-  },
-
-  getChatCompletionStream: async function (
-    config: ProviderConfig,
-    messages: Message[],
-    onResponseUpdate: (update: ModelResponseUpdate) => void,
-  ) {
-    const api = new MistralClient({ apiKey: config.apiKey });
-    const allMessages = getMessages(config, messages);
-
-    const startTime = performance.now();
-    const stream = await api.chat.stream({
-      messages: allMessages,
-      model: config.model,
-      ...getMistralResponseStyle(config),
-    });
-
-    let lastChunk: CompletionEvent | null = null;
-    let content = '';
-    for await (const chunk of stream) {
-      lastChunk = chunk;
-      content += chunk.data.choices[0]?.delta?.content || '';
-      onResponseUpdate({ content });
-    }
-
-    const responseTime = performance.now() - startTime;
-
-    return {
-      message: {
-        role: 'assistant',
-        content,
-      },
-      usage: {
-        inputTokens: lastChunk?.data.usage?.promptTokens ?? estimateInputTokens(allMessages),
-        outputTokens: lastChunk?.data.usage?.completionTokens ?? estimateOutputTokens(content),
-        requests: 1,
-      },
-      responseTime,
-      responseModel: lastChunk?.data.model || 'unknown',
-      data: lastChunk,
-    };
   },
 };
 
-function getMessages(config: ProviderConfig, messages: Message[]): Message[] {
-  if (!config.systemPrompt) {
-    return messages;
-  }
-
-  const systemMessage: Message = {
-    role: 'system',
-    content: config.systemPrompt,
-  };
-  return [systemMessage, ...messages];
-}
-
-interface MistralResponseStyle {
-  temperature?: number;
-  topP?: number;
-}
-
-function getMistralResponseStyle(config: ProviderConfig): MistralResponseStyle {
-  const style = responseStyles[config.responseStyle];
-
-  const result: MistralResponseStyle = {};
-  if ('temperature' in style) {
-    result.temperature = style.temperature;
-  }
-  if ('top_p' in style) {
-    result.topP = style.top_p;
-  }
-
-  return result;
-}
-
 export default Mistral;

src/engine/providers/open-ai.ts

@@ -1,10 +1,9 @@
-import OpenAI from 'openai';
-import { type Message, type ModelResponseUpdate } from '../inference.js';
+import { createOpenAI } from '@ai-sdk/openai';
+import { VercelChatModelAdapter } from '@callstack/byorg-core';
 import { type ProviderConfig } from './config.js';
-import type { Provider } from './provider.js';
-import { getChatCompletion, getChatCompletionStream } from './utils/open-ai-api.js';
+import type { ProviderInfo } from './provider-info.js';
 
-const OpenAi: Provider = {
+const OpenAi: ProviderInfo = {
   label: 'OpenAI',
   name: 'openAi',
   apiKeyUrl: 'https://platform.openai.com/api-keys',
@@ -31,24 +30,11 @@ const OpenAi: Provider = {
 
   modelAliases: {},
 
-  getChatCompletion: async (config: ProviderConfig, messages: Message[]) => {
-    const api = new OpenAI({
-      apiKey: config.apiKey,
+  getChatModel: (config: ProviderConfig) => {
+    const client = createOpenAI({ apiKey: config.apiKey, compatibility: 'strict' });
+    return new VercelChatModelAdapter({
+      languageModel: client.languageModel(config.model),
     });
-
-    return await getChatCompletion(api, config, messages);
-  },
-
-  getChatCompletionStream: async function (
-    config: ProviderConfig,
-    messages: Message[],
-    onResponseUpdate: (update: ModelResponseUpdate) => void,
-  ) {
-    const api = new OpenAI({
-      apiKey: config.apiKey,
-    });
-
-    return await getChatCompletionStream(api, config, messages, onResponseUpdate);
   },
 };
 

src/engine/providers/perplexity.ts

@@ -1,10 +1,8 @@
-import OpenAI from 'openai';
-import { type Message, type ModelResponseUpdate } from '../inference.js';
-import { type ProviderConfig } from './config.js';
-import type { Provider } from './provider.js';
-import { getChatCompletion, getChatCompletionStream } from './utils/open-ai-api.js';
+import { createOpenAI } from '@ai-sdk/openai';
+import { VercelChatModelAdapter } from '@callstack/byorg-core';
+import type { ProviderInfo } from './provider-info.js';
 
-const Perplexity: Provider = {
+const Perplexity: ProviderInfo = {
   label: 'Perplexity',
   name: 'perplexity',
   apiKeyUrl: 'https://perplexity.ai/settings/api',
@@ -43,26 +41,14 @@ const Perplexity: Provider = {
     huge: 'llama-3.1-sonar-huge-128k-online',
   },
 
-  getChatCompletion: async (config: ProviderConfig, messages: Message[]) => {
-    const api = new OpenAI({
+  getChatModel(config) {
+    const client = createOpenAI({
       apiKey: config.apiKey,
-      baseURL: 'https://api.perplexity.ai',
+      baseURL: 'https://api.perplexity.ai/',
     });
-
-    return await getChatCompletion(api, config, messages);
-  },
-
-  getChatCompletionStream: async function (
-    config: ProviderConfig,
-    messages: Message[],
-    onResponseUpdate: (update: ModelResponseUpdate) => void,
-  ) {
-    const api = new OpenAI({
-      apiKey: config.apiKey,
-      baseURL: 'https://api.perplexity.ai',
+    return new VercelChatModelAdapter({
+      languageModel: client.languageModel(config.model),
     });
-
-    return await getChatCompletionStream(api, config, messages, onResponseUpdate);
   },
 };
 

src/engine/providers/provider.ts → src/engine/providers/provider-info.ts

@@ -1,4 +1,4 @@
-import type { Message, ModelResponse, ModelResponseUpdate } from '../inference.js';
+import type { ChatModel } from '@callstack/byorg-core';
 import type { ProviderConfig } from './config.js';
 import openAi from './open-ai.js';
 import perplexity from './perplexity.js';
@@ -8,7 +8,7 @@ import mistral from './mistral.js';
 export const providerNames = ['openAi', 'anthropic', 'perplexity', 'mistral'] as const;
 export type ProviderName = (typeof providerNames)[number];
 
-export interface Provider {
+export interface ProviderInfo {
   name: ProviderName;
   label: string;
   apiKeyUrl: string;
@@ -17,12 +17,7 @@ export interface Provider {
   modelPricing: Record<string, ModelPricing>;
   modelAliases: Record<string, string>;
 
-  getChatCompletion: (config: ProviderConfig, messages: Message[]) => Promise<ModelResponse>;
-  getChatCompletionStream?: (
-    config: ProviderConfig,
-    messages: Message[],
-    onStreamUpdate: (update: ModelResponseUpdate) => void,
-  ) => Promise<ModelResponse>;
+  getChatModel: (config: ProviderConfig) => ChatModel;
 }
 
 export interface ModelPricing {
@@ -36,7 +31,7 @@ export interface ModelPricing {
   requestsCost?: number;
 }
 
-const providersMap: Record<ProviderName, Provider> = {
+const providersMap: Record<ProviderName, ProviderInfo> = {
   openAi,
   anthropic,
   perplexity,
@@ -45,7 +40,7 @@ const providersMap: Record<ProviderName, Provider> = {
 
 export const providers = Object.values(providersMap);
 
-export function getProvider(providerName: ProviderName): Provider {
+export function getProvider(providerName: ProviderName): ProviderInfo {
   const provider = providersMap[providerName];
   if (!provider) {
     throw new Error(`Provider not found: ${providerName}.`);

src/engine/session.ts

@@ -1,5 +1,5 @@
-import type { ModelUsage } from './inference.js';
-import type { ModelPricing } from './providers/provider.js';
+import type { ModelUsage } from '@callstack/byorg-core';
+import type { ModelPricing } from './providers/provider-info.js';
 
 export interface SessionUsage {
   total: ModelUsage;

src/engine/tokenizer.ts

@@ -1,6 +1,6 @@
 import { createRequire } from 'module';
 import { Tiktoken } from 'tiktoken/lite';
-import type { Message } from './inference.js';
+import type { Message } from '@callstack/byorg-core';
 
 // Workaround for JSON loading in ESM
 // See: https://www.stefanjudis.com/snippets/how-to-import-json-files-in-es-modules-node-js/#option-2%3A-leverage-the-commonjs-%60require%60-function-to-load-json-files

src/file-utils.ts

@@ -2,7 +2,7 @@ import * as fs from 'fs';
 import * as os from 'os';
 import * as path from 'path';
 import { format } from 'date-fns';
-import type { Message } from './engine/inference.js';
+import type { Message } from '@callstack/byorg-core';
 
 export const CHATS_SAVE_DIRECTORY = '~/ai-chats';
 

tsconfig.json

@@ -4,9 +4,8 @@
       "ai-cli": ["./src/index.js"]
     },
     "target": "ES2020",
-    "lib": ["ES2020"],
-    "module": "Node16",
-    "moduleResolution": "Node16",
+    "module": "ESNext",
+    "moduleResolution": "Bundler",
     "outDir": "build",
     "jsx": "react",
     "esModuleInterop": true,