viho 0.1.1 → 0.2.0

package/README.md

@@ -13,11 +13,14 @@
 
 ## Features
 
-- Multiple AI model management
+- Multiple AI model management (OpenAI, Google Gemini)
+- Support for OpenAI-compatible APIs
+- Support for Google Gemini AI (API and Vertex AI)
 - Interactive Q&A with streaming responses
 - Continuous chat sessions for multi-turn conversations
-- Support for thinking mode (enabled/disabled/auto)
-- Configurable API endpoints (OpenAI, Anthropic, custom providers)
+- Support for thinking mode (for compatible models)
+- Expert mode with domain-specific knowledge
+- Configurable API endpoints
 - Default model configuration
 - Simple and intuitive CLI interface
 - Persistent configuration storage
@@ -62,19 +65,40 @@ viho ask
 
 Add a new AI model configuration interactively.
 
-You'll be prompted to enter:
+First, you'll select the model type:
+
+- **openai** - For OpenAI and OpenAI-compatible APIs
+- **gemini api** - For Google Gemini AI Studio
+- **gemini vertex** - For Google Gemini Vertex AI
+
+Then you'll be prompted to enter the required information based on the model type:
+
+**For OpenAI:**
 
 - Model name (a custom identifier)
 - API key
 - Base URL (e.g., https://api.openai.com/v1)
-- Model ID (e.g., gpt-4, claude-3-5-sonnet-20241022)
-- Thinking mode (enabled/disabled/auto)
+- Model ID (e.g., gpt-4, gpt-4o)
+- Thinking mode (enabled/disabled)
+
+**For Gemini API:**
+
+- Model name (a custom identifier)
+- API key (from Google AI Studio)
+- Model ID (e.g., gemini-pro, gemini-1.5-flash, gemini-1.5-pro)
+
+**For Gemini Vertex:**
+
+- Model name (a custom identifier)
+- Project ID (your GCP project)
+- Location (e.g., us-east1, us-central1)
+- Model ID (e.g., gemini-1.5-flash-002, gemini-1.5-pro-002)
 
 ```bash
 viho model add
 ```
 
-After adding a model, it will be available for use with `viho ask` and `viho chat` commands.
+After adding a model, it will be available for use with `viho ask`, `viho chat`, and `viho expert` commands.
 
 #### `viho model list`
 
@@ -87,24 +111,30 @@ viho model list
 This command displays:
 
 - Model name with a `(default)` tag for the default model
-- Model ID
-- Base URL
-- Thinking mode setting
+- Model type (openai, gemini api, or gemini vertex)
+- Type-specific configuration details
 
 **Example output:**
 
 ```
 Configured models:
 
-  • deepseek (default)
-    Model ID: ep-20250822181529-2gg27
-    Base URL: https://ark.cn-beijing.volces.com/api/v3
-    Thinking: auto
-
-  • kimi
-    Model ID: kimi-k2-thinking
-    Base URL: https://api.moonshot.cn/v1
+  • gpt4 (default)
+    Type: openai
+    Model ID: gpt-4o
+    Base URL: https://api.openai.com/v1
     Thinking: enabled
+
+  • gemini
+    Type: gemini api
+    Model ID: gemini-1.5-flash
+    API Key: ***
+
+  • gemini-pro
+    Type: gemini vertex
+    Model ID: gemini-1.5-pro-002
+    Project ID: my-project-123
+    Location: us-east1
 ```
 
 #### `viho model remove`
@@ -177,6 +207,49 @@ The chat session runs in a loop, allowing you to ask multiple questions continuo
 - `viho ask` - Single question, exits after receiving the answer
 - `viho chat` - Continuous loop, keeps asking for new questions until manually stopped (Ctrl+C)
 
+### Expert Mode
+
+Expert mode allows you to chat with an AI model that has access to domain-specific documentation, making it more knowledgeable about particular libraries or frameworks.
+
+#### `viho expert list`
+
+List all available expert resources:
+
+```bash
+viho expert list
+```
+
+This displays available experts like:
+
+- `antd` - Ant Design documentation
+- `daisyui` - DaisyUI documentation
+
+#### `viho expert <name> [modelName]`
+
+Start an expert chat session with domain-specific knowledge:
+
+```bash
+viho expert antd
+```
+
+Or specify a model explicitly:
+
+```bash
+viho expert daisyui mymodel
+```
+
+The expert mode works similarly to `viho chat` but includes the relevant documentation as context, making the AI more accurate when answering questions about that specific library or framework.
+
+**Example:**
+
+```bash
+# Get help with Ant Design
+viho expert antd
+
+# Ask: "How do I create a responsive table with sorting?"
+# The AI will use Ant Design documentation to provide accurate answers
+```
+
 ## Configuration
 
 Configuration is stored in `~/viho.json`. You can manage all settings through the CLI commands.
@@ -200,34 +273,58 @@ Configuration is stored in `~/viho.json`. You can manage all settings through th
 
 ## Supported Providers
 
-viho works with any OpenAI-compatible API, including:
+viho supports multiple AI providers:
+
+### OpenAI-Compatible APIs
 
-- OpenAI (GPT-4, GPT-3.5, etc.)
-- Anthropic Claude (via compatible endpoints)
+- OpenAI (GPT-4, GPT-4o, GPT-3.5, etc.)
+- Any OpenAI-compatible API endpoints
 - Custom LLM providers with OpenAI-compatible APIs
 
+### Google Gemini
+
+- **Gemini API** (via Google AI Studio)
+  - Ideal for personal development and prototyping
+  - Get API key from [Google AI Studio](https://makersuite.google.com/app/apikey)
+
+- **Gemini Vertex AI** (via Google Cloud)
+  - Enterprise-grade with advanced features
+  - Requires Google Cloud project with Vertex AI enabled
+  - Supports context caching for cost optimization
+
 ## Examples
 
 ### Adding an OpenAI Model
 
 ```bash
 viho model add
+# Select model type: openai
 # Enter model name: gpt4
 # Enter API key: sk-...
 # Enter base URL: https://api.openai.com/v1
-# Enter model ID: gpt-4
+# Enter model ID: gpt-4o
 # Thinking mode: disabled
 ```
 
-### Adding a Claude Model
+### Adding a Gemini API Model
+
+```bash
+viho model add
+# Select model type: gemini api
+# Enter model name: gemini
+# Enter API key: your-google-ai-api-key
+# Enter model ID: gemini-1.5-flash
+```
+
+### Adding a Gemini Vertex AI Model
 
 ```bash
 viho model add
-# Enter model name: claude
-# Enter API key: your-anthropic-key
-# Enter base URL: https://api.anthropic.com
-# Enter model ID: claude-3-5-sonnet-20241022
-# Thinking mode: auto
+# Select model type: gemini vertex
+# Enter model name: gemini-pro
+# Enter projectId: my-gcp-project
+# Enter location: us-east1
+# Enter model ID: gemini-1.5-pro-002
 ```
 
 ### Setting Up for First Use
@@ -247,7 +344,8 @@ viho ask
 
 - [qiao-cli](https://www.npmjs.com/package/qiao-cli) - CLI utilities
 - [qiao-config](https://www.npmjs.com/package/qiao-config) - Configuration management
-- [qiao-llm](https://www.npmjs.com/package/qiao-llm) - LLM integration
+- [qiao-file](https://www.npmjs.com/package/qiao-file) - File utilities
+- [viho-llm](https://www.npmjs.com/package/viho-llm) - Multi-provider LLM integration (OpenAI, Gemini)
 
 ## License
 

package/bin/viho-ask.js

@@ -1,12 +1,9 @@
 // qiao
 const cli = require('qiao-cli');
 
-// llm
-const LLM = require('qiao-llm');
-
 // util
 const { ask } = require('../src/llm.js');
-const { getDB, preLLMAsk } = require('../src/util.js');
+const { getDB, preLLMAsk, initLLM } = require('../src/util.js');
 const db = getDB();
 
 // cmd
@@ -18,10 +15,7 @@ cli.cmd
     const model = await preLLMAsk('ask', db, modelName);
 
     // llm
-    const llm = LLM({
-      apiKey: model.apiKey,
-      baseURL: model.baseURL,
-    });
+    const llm = initLLM(model);
 
     // ask
     await ask(llm, model);

package/bin/viho-chat.js

@@ -1,12 +1,9 @@
 // qiao
 const cli = require('qiao-cli');
 
-// llm
-const LLM = require('qiao-llm');
-
 // util
 const { ask } = require('../src/llm.js');
-const { getDB, preLLMAsk } = require('../src/util.js');
+const { getDB, preLLMAsk, initLLM } = require('../src/util.js');
 const db = getDB();
 
 // cmd
@@ -18,10 +15,7 @@ cli.cmd
     const model = await preLLMAsk('chat', db, modelName);
 
     // init
-    const llm = LLM({
-      apiKey: model.apiKey,
-      baseURL: model.baseURL,
-    });
+    const llm = initLLM(model);
 
     // chat
     let keepChatting = true;

package/bin/viho-expert.js

@@ -1,13 +1,10 @@
 // qiao
 const cli = require('qiao-cli');
 
-// llm
-const LLM = require('qiao-llm');
-
 // util
 const { expertAsk } = require('../src/llm.js');
 const { experts } = require('../src/experts/experts.js');
-const { getDB, printLogo, preLLMAsk } = require('../src/util.js');
+const { getDB, printLogo, preLLMAsk, initLLM } = require('../src/util.js');
 const db = getDB();
 
 // actions
@@ -65,10 +62,7 @@ async function expertAskByName(expertName) {
   if (!model) return;
 
   // init
-  const llm = LLM({
-    apiKey: model.apiKey,
-    baseURL: model.baseURL,
-  });
+  const llm = initLLM(model);
 
   // chat
   let keepChatting = true;

package/bin/viho-model.js

@@ -32,8 +32,19 @@ cli.cmd
  */
 async function modelAdd() {
   try {
-    // q a
-    const questions = [
+    // model type
+    const modelTypeQuestion = [
+      {
+        type: 'list',
+        name: 'modelType',
+        message: 'openai vs gemini',
+        choices: ['openai', 'gemini api', 'gemini vertex'],
+      },
+    ];
+    const modelTypeAnswer = await cli.ask(modelTypeQuestion);
+
+    // questions
+    const openAIQuestion = [
       {
         type: 'input',
         name: 'modelName',
@@ -61,7 +72,55 @@ async function modelAdd() {
         choices: ['enabled', 'disabled'],
       },
     ];
-    const answers = await cli.ask(questions);
+    const geminiAPIQuestion = [
+      {
+        type: 'input',
+        name: 'modelName',
+        message: 'Enter model name:',
+      },
+      {
+        type: 'input',
+        name: 'apiKey',
+        message: 'Enter API key:',
+      },
+      {
+        type: 'input',
+        name: 'modelID',
+        message: 'Enter model ID (e.g., gemini-pro, gemini-1.5-flash):',
+      },
+    ];
+    const geminiVertexQuestion = [
+      {
+        type: 'input',
+        name: 'modelName',
+        message: 'Enter model name:',
+      },
+      {
+        type: 'input',
+        name: 'projectId',
+        message: 'Enter projectId:',
+      },
+      {
+        type: 'input',
+        name: 'location',
+        message: 'Enter location:',
+      },
+      {
+        type: 'input',
+        name: 'modelID',
+        message: 'Enter model ID (e.g., gemini-1.5-flash-002, gemini-1.5-pro-002):',
+      },
+    ];
+
+    // final question
+    let finalQuestion;
+    if (modelTypeAnswer.modelType === 'openai') finalQuestion = openAIQuestion;
+    if (modelTypeAnswer.modelType === 'gemini api') finalQuestion = geminiAPIQuestion;
+    if (modelTypeAnswer.modelType === 'gemini vertex') finalQuestion = geminiVertexQuestion;
+
+    // answers
+    const answers = await cli.ask(finalQuestion);
+    answers.modelType = modelTypeAnswer.modelType;
     console.log();
 
     // check
@@ -110,9 +169,21 @@ async function modelList() {
       const isDefault = model.modelName === defaultModel;
       const defaultTag = isDefault ? cli.colors.green(' (default)') : '';
       console.log(cli.colors.white(`  • ${model.modelName}${defaultTag}`));
-      console.log(cli.colors.gray(`    Model ID: ${model.modelID}`));
-      console.log(cli.colors.gray(`    Base URL: ${model.baseURL}`));
-      console.log(cli.colors.gray(`    Thinking: ${model.modelThinking}`));
+      console.log(cli.colors.gray(`    Type: ${model.modelType || 'openai'}`));
+
+      if (model.modelType === 'openai') {
+        console.log(cli.colors.gray(`    Model ID: ${model.modelID}`));
+        console.log(cli.colors.gray(`    Base URL: ${model.baseURL}`));
+        console.log(cli.colors.gray(`    Thinking: ${model.modelThinking}`));
+      } else if (model.modelType === 'gemini api') {
+        console.log(cli.colors.gray(`    Model ID: ${model.modelID}`));
+        console.log(cli.colors.gray(`    API Key: ${model.apiKey ? '***' : 'Not set'}`));
+      } else if (model.modelType === 'gemini vertex') {
+        console.log(cli.colors.gray(`    Model ID: ${model.modelID}`));
+        console.log(cli.colors.gray(`    Project ID: ${model.projectId}`));
+        console.log(cli.colors.gray(`    Location: ${model.location}`));
+      }
+
       console.log();
     });
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "viho",
-  "version": "0.1.1",
+  "version": "0.2.0",
   "description": "A lightweight CLI tool for managing and interacting with AI models",
   "keywords": [
     "ai",
@@ -42,7 +42,7 @@
     "qiao-cli": "^5.0.0",
     "qiao-config": "^5.0.1",
     "qiao-file": "^5.0.1",
-    "qiao-llm": "^0.2.5"
+    "viho-llm": "^0.2.0"
   },
-  "gitHead": "30ba092c8b2ab0e52894506eb93fdd77a26e4e36"
+  "gitHead": "38f9238d5c59fe322424ee30e32037b44f6be7d8"
 }

package/src/llm.js

@@ -31,17 +31,36 @@ exports.ask = async (llm, model, systemPrompt) => {
   console.log(cli.colors.gray(answers.content));
   console.log();
 
-  // chat
-  const chatOptions = {
-    model: model.modelID,
-    messages: [
-      { role: 'system', content: systemPrompt || defaultSystemPrompt },
-      { role: 'user', content: answers.content },
-    ],
-    thinking: {
-      type: model.modelThinking,
-    },
-  };
+  // model type
+  const modelType = model.modelType || 'openai';
+
+  // chat options based on model type
+  let chatOptions;
+  if (modelType === 'openai') {
+    chatOptions = {
+      model: model.modelID,
+      messages: [
+        { role: 'system', content: systemPrompt || defaultSystemPrompt },
+        { role: 'user', content: answers.content },
+      ],
+      thinking: {
+        type: model.modelThinking,
+      },
+    };
+  } else if (modelType === 'gemini api' || modelType === 'gemini vertex') {
+    chatOptions = {
+      contents: [
+        {
+          role: 'user',
+          parts: [{ text: answers.content }],
+        },
+      ],
+    };
+    // Add system instruction if provided
+    if (systemPrompt && systemPrompt !== defaultSystemPrompt) {
+      chatOptions.systemInstruction = systemPrompt;
+    }
+  }
 
   // callback options
   const callbackOptions = {

package/src/util.js

@@ -10,6 +10,9 @@ const cli = require('qiao-cli');
 // db
 const DB = require('qiao-config');
 
+// llm
+const { OpenAIAPI, GeminiAPI, GeminiVertex } = require('viho-llm');
+
 // model
 const { getModelByName } = require('./model.js');
 
@@ -58,3 +61,37 @@ exports.preLLMAsk = async (type, db, modelName) => {
   // r
   return model;
 };
+
+/**
+ * initLLM
+ * @param {*} model
+ * @returns
+ */
+exports.initLLM = (model) => {
+  const modelType = model.modelType || 'openai';
+
+  if (modelType === 'openai') {
+    return OpenAIAPI({
+      apiKey: model.apiKey,
+      baseURL: model.baseURL,
+    });
+  } else if (modelType === 'gemini api') {
+    return GeminiAPI({
+      apiKey: model.apiKey,
+      modelName: model.modelID, // Use modelID for API calls
+    });
+  } else if (modelType === 'gemini vertex') {
+    return GeminiVertex({
+      projectId: model.projectId,
+      location: model.location,
+      modelName: model.modelID, // Use modelID for API calls
+    });
+  }
+
+  // fallback to openai
+  console.log(cli.colors.yellow(`Unknown model type: ${modelType}, falling back to OpenAI`));
+  return OpenAIAPI({
+    apiKey: model.apiKey,
+    baseURL: model.baseURL,
+  });
+};