@showlotus/opencode-image-vision 1.0.0 → 1.0.2

package/README.md

@@ -1,5 +1,6 @@
 # opencode-image-vision
 
+[![npm version](https://img.shields.io/npm/v/@showlotus/opencode-image-vision.svg)](https://www.npmjs.com/package/@showlotus/opencode-image-vision)
 [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](LICENSE)
 [![MCP](https://img.shields.io/badge/MCP-Server-blue.svg)](https://modelcontextprotocol.io/)
 [![Node](https://img.shields.io/badge/Node.js-22%2B-green.svg)](https://nodejs.org/)
@@ -57,7 +58,7 @@ No clone or install needed. Just add to your `opencode.jsonc`:
   "mcp": {
     "image-vision": {
       "type": "local",
-      "command": ["npx", "-y", "opencode-image-vision"],
+      "command": ["npx", "-y", "@showlotus/opencode-image-vision"],
       "environment": {
         "model": "zhipuai-coding-plan/glm-4.6v",
       },
@@ -211,12 +212,30 @@ The screenshot shows a terminal with the following error message...
 
 ### Supported providers
 
-| Provider ID           | Base URL                               | Models     |
-| --------------------- | -------------------------------------- | ---------- |
-| `zhipuai-coding-plan` | `https://open.bigmodel.cn/api/paas/v4` | `glm-4.6v` |
-| `zai-coding-plan`     | `https://open.bigmodel.cn/api/paas/v4` | `glm-4.6v` |
-| `z-ai`                | `https://open.bigmodel.cn/api/paas/v4` | `glm-4.6v` |
-| `zhipuai`             | `https://open.bigmodel.cn/api/paas/v4` | `glm-4.6v` |
+**OpenAI-compatible** (reuse `OpenAICompatibleProvider`):
+
+| Provider ID           | Base URL                                                 | Example Models                              |
+| --------------------- | -------------------------------------------------------- | ------------------------------------------- |
+| `zhipuai-coding-plan` | `https://open.bigmodel.cn/api/paas/v4`                   | `glm-4.6v`                                  |
+| `zai-coding-plan`     | `https://open.bigmodel.cn/api/paas/v4`                   | `glm-4.6v`                                  |
+| `z-ai`                | `https://open.bigmodel.cn/api/paas/v4`                   | `glm-4.6v`                                  |
+| `zhipuai`             | `https://open.bigmodel.cn/api/paas/v4`                   | `glm-4.6v`                                  |
+| `moonshot` / `kimi`   | `https://api.moonshot.cn/v1`                             | `moonshot-v1-32k-vision-preview`            |
+| `minimax` / `minimax-cn-coding-plan` | `https://api.minimaxi.chat/v1`             | `MiniMax-Text-01`                           |
+| `openai`              | `https://api.openai.com/v1`                              | `gpt-4o`, `gpt-4o-mini`                     |
+| `qwen` / `dashscope`  | `https://dashscope.aliyuncs.com/compatible-mode/v1`      | `qwen-vl-max`, `qwen-vl-plus`               |
+| `doubao` / `volcengine` | `https://ark.cn-beijing.volces.com/api/v3`             | `doubao-vision-pro-32k`                     |
+| `yi` / `lingyiwanwu`  | `https://api.lingyiwanwu.com/v1`                         | `yi-vision-v2`                              |
+| `gemini` / `google`   | `https://generativelanguage.googleapis.com/v1beta/openai` | `gemini-2.0-flash`, `gemini-1.5-pro`      |
+| `stepfun`             | `https://api.stepfun.com/v1`                             | `step-1v-32k`                               |
+| `baichuan`            | `https://api.baichuan-ai.com/v1`                         | `Baichuan4-Vision`                          |
+| `hunyuan`             | `https://api.hunyuan.cloud.tencent.com/v1`               | `hunyuan-vision`                            |
+
+**Custom API format**:
+
+| Provider ID           | Base URL                           | Provider Class   | Example Models                           |
+| --------------------- | ---------------------------------- | ---------------- | ---------------------------------------- |
+| `anthropic` / `claude` | `https://api.anthropic.com/v1`    | `ClaudeProvider` | `claude-3-5-sonnet-20241022`             |
 
 ---
 
@@ -231,49 +250,56 @@ Model: [calls analyze_images with session_id]
         PostgreSQL isn't running on port 5432. Start it with: brew services start postgresql"
 ```
 
-The text-only model never sees pixels — it reads the description returned by GLM-4.6V and reasons over it.
+The text-only model never sees pixels — it reads the description returned by the vision model and reasons over it.
 
 ---
 
 ## Extending with new providers
 
-Adding a new vision provider takes 3 steps:
+Most vision model providers use the **OpenAI-compatible chat completions API** — you only need to add 2 registry entries (no code). Only providers with a **different API format** need a custom class.
 
-**1. Add base URL to the registry** (`src/opencode.js`):
+### Adding an OpenAI-compatible provider (e.g. OpenAI, Qwen, Doubao)
+
+**1. Add base URL** (`src/opencode.js` → `PROVIDER_REGISTRY`):
 
 ```javascript
-const PROVIDER_REGISTRY = {
-  'zhipuai-coding-plan': { baseUrl: 'https://open.bigmodel.cn/api/paas/v4', format: 'openai' },
-  // Add new provider:
-  openai: { baseUrl: 'https://api.openai.com/v1', format: 'openai' },
-}
+'my-provider': { baseUrl: 'https://api.example.com/v1', format: 'openai' },
+```
+
+**2. Add provider mapping** (`src/providers/index.js` → `OPENAI_COMPATIBLE`):
+
+```javascript
+'my-provider': OpenAICompatibleProvider,
+```
+
+Done. Set `"model": "my-provider/my-vision-model"` in config.
+
+### Adding a custom-format provider (e.g. Anthropic Claude)
+
+**1. Add base URL** (`src/opencode.js` → `PROVIDER_REGISTRY`):
+
+```javascript
+'my-provider': { baseUrl: 'https://api.example.com/v1', format: 'custom' },
 ```
 
-**2. Create a provider class** (`src/providers/openai.js`) — only needed if the API format differs:
+**2. Create a provider class** (`src/providers/my-provider.js`):
 
 ```javascript
 import { VisionProvider } from './base.js'
 
-export class OpenAIProvider extends VisionProvider {
+export class MyProvider extends VisionProvider {
   async analyze(base64, mime, prompt) {
     // Implement provider-specific API call
   }
 }
 ```
 
-**3. Register the mapping** (`src/providers/index.js`):
+See `src/providers/claude.js` for a working example (Anthropic uses `x-api-key` auth and `/messages` endpoint).
 
-```javascript
-const PROVIDER_MAP = {
-  'zhipuai-coding-plan': GLMProvider,
-  openai: OpenAIProvider,
-}
-```
-
-Then set the `model` environment variable:
+**3. Add provider mapping** (`src/providers/index.js` → `PROVIDER_MAP`):
 
-```jsonc
-"environment": { "model": "openai/gpt-4o" }
+```javascript
+'my-provider': MyProvider,
 ```
 
 ---

package/package.json

@@ -1,8 +1,9 @@
 {
   "name": "@showlotus/opencode-image-vision",
-  "version": "1.0.0",
+  "version": "1.0.2",
   "description": "MCP server that reads images from OpenCode's SQLite database and analyzes them via vision AI providers",
   "type": "module",
+  "packageManager": "pnpm@8.11.0",
   "main": "src/index.js",
   "bin": {
     "opencode-image-vision": "./src/index.js"

package/src/concurrency.js

@@ -0,0 +1,19 @@
+// 并发执行任务，控制最大并发数，结果按原顺序返回
+export async function mapWithConcurrency(items, concurrency, worker) {
+  const results = new Array(items.length)
+  let cursor = 0
+
+  // 单个执行器：循环领取下一个未处理的任务，直到全部完成
+  const run = async () => {
+    while (cursor < items.length) {
+      const i = cursor++
+      results[i] = await worker(items[i], i)
+    }
+  }
+
+  // 启动不超过任务数量的并发执行器
+  await Promise.all(
+    Array.from({ length: Math.min(concurrency, items.length) }, run),
+  )
+  return results
+}

package/src/index.js

@@ -5,6 +5,7 @@ import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js'
 import { z } from 'zod'
 import { getDatabase, getImages } from './db.js'
 import { createProvider } from './providers/index.js'
+import { mapWithConcurrency } from './concurrency.js'
 
 const DEFAULT_PROMPT =
   process.env.prompt ||
@@ -17,6 +18,9 @@ const DEFAULT_PROMPT =
 const DEFAULT_LIMIT = Number(process.env.limit) || 5
 const MAX_LIMIT = Number(process.env.max_limit) || 20
 
+// 并发分析图片的最大并发数，可通过环境变量 concurrency 覆盖
+const DEFAULT_CONCURRENCY = Number(process.env.concurrency) || 5
+
 let provider
 try {
   provider = createProvider()
@@ -60,16 +64,19 @@ server.tool(
         }
       }
 
-      const results = []
-      for (let i = 0; i < images.length; i++) {
-        const img = images[i]
-        try {
-          const desc = await provider.analyze(img.base64, img.mime, analysisPrompt)
-          results.push(`### Image ${i + 1}: ${img.filename}\n\n${desc}`)
-        } catch (e) {
-          results.push(`### Image ${i + 1}: ${img.filename}\n\n[Analysis failed: ${e.message}]`)
-        }
-      }
+      // 并发分析图片，单张失败不影响其他图片，结果保持原顺序
+      const results = await mapWithConcurrency(
+        images,
+        DEFAULT_CONCURRENCY,
+        async (img, i) => {
+          try {
+            const desc = await provider.analyze(img.base64, img.mime, analysisPrompt)
+            return `### Image ${i + 1}: ${img.filename}\n\n${desc}`
+          } catch (e) {
+            return `### Image ${i + 1}: ${img.filename}\n\n[Analysis failed: ${e.message}]`
+          }
+        },
+      )
 
       return {
         content: [

package/src/opencode.js

@@ -5,10 +5,28 @@ import { join } from 'node:path';
 // Provider ID → base URL mapping
 // Future providers can be added here
 const PROVIDER_REGISTRY = {
-  'zhipuai-coding-plan': { baseUrl: 'https://open.bigmodel.cn/api/paas/v4', format: 'openai' },
-  'zai-coding-plan':     { baseUrl: 'https://open.bigmodel.cn/api/paas/v4', format: 'openai' },
-  'z-ai':                { baseUrl: 'https://open.bigmodel.cn/api/paas/v4', format: 'openai' },
-  'zhipuai':             { baseUrl: 'https://open.bigmodel.cn/api/paas/v4', format: 'openai' },
+  'zhipuai-coding-plan':    { baseUrl: 'https://open.bigmodel.cn/api/paas/v4', format: 'openai' },
+  'zai-coding-plan':        { baseUrl: 'https://open.bigmodel.cn/api/paas/v4', format: 'openai' },
+  'z-ai':                   { baseUrl: 'https://open.bigmodel.cn/api/paas/v4', format: 'openai' },
+  'zhipuai':                { baseUrl: 'https://open.bigmodel.cn/api/paas/v4', format: 'openai' },
+  'moonshot':               { baseUrl: 'https://api.moonshot.cn/v1', format: 'openai' },
+  'kimi':                   { baseUrl: 'https://api.moonshot.cn/v1', format: 'openai' },
+  'minimax':                { baseUrl: 'https://api.minimaxi.chat/v1', format: 'openai' },
+  'minimax-cn-coding-plan': { baseUrl: 'https://api.minimaxi.chat/v1', format: 'openai' },
+  'openai':                 { baseUrl: 'https://api.openai.com/v1', format: 'openai' },
+  'qwen':                   { baseUrl: 'https://dashscope.aliyuncs.com/compatible-mode/v1', format: 'openai' },
+  'dashscope':              { baseUrl: 'https://dashscope.aliyuncs.com/compatible-mode/v1', format: 'openai' },
+  'doubao':                 { baseUrl: 'https://ark.cn-beijing.volces.com/api/v3', format: 'openai' },
+  'volcengine':             { baseUrl: 'https://ark.cn-beijing.volces.com/api/v3', format: 'openai' },
+  'yi':                     { baseUrl: 'https://api.lingyiwanwu.com/v1', format: 'openai' },
+  'lingyiwanwu':            { baseUrl: 'https://api.lingyiwanwu.com/v1', format: 'openai' },
+  'gemini':                 { baseUrl: 'https://generativelanguage.googleapis.com/v1beta/openai', format: 'openai' },
+  'google':                 { baseUrl: 'https://generativelanguage.googleapis.com/v1beta/openai', format: 'openai' },
+  'stepfun':                { baseUrl: 'https://api.stepfun.com/v1', format: 'openai' },
+  'baichuan':               { baseUrl: 'https://api.baichuan-ai.com/v1', format: 'openai' },
+  'hunyuan':                { baseUrl: 'https://api.hunyuan.cloud.tencent.com/v1', format: 'openai' },
+  'anthropic':              { baseUrl: 'https://api.anthropic.com/v1', format: 'anthropic' },
+  'claude':                 { baseUrl: 'https://api.anthropic.com/v1', format: 'anthropic' },
 };
 
 export function resolveProviderConfig(providerId, modelId) {

package/src/providers/claude.js

@@ -0,0 +1,64 @@
+import { VisionProvider } from './base.js'
+
+export class ClaudeProvider extends VisionProvider {
+  constructor(config) {
+    super(config)
+    this.apiKey = config.apiKey
+    this.baseUrl = config.baseUrl
+    this.model = config.model
+    this.timeout = config.timeout || 60_000
+
+    if (!this.apiKey) {
+      throw new Error('Anthropic API key not configured.')
+    }
+    if (!this.baseUrl) {
+      throw new Error('Anthropic base URL not configured.')
+    }
+    if (!this.model) {
+      throw new Error('Anthropic model not configured.')
+    }
+  }
+
+  async analyze(base64, mime, prompt) {
+    const ctrl = new AbortController()
+    const timer = setTimeout(() => ctrl.abort(), this.timeout)
+
+    try {
+      const res = await fetch(`${this.baseUrl}/messages`, {
+        method: 'POST',
+        headers: {
+          'Content-Type': 'application/json',
+          'x-api-key': this.apiKey,
+          'anthropic-version': '2023-06-01',
+        },
+        body: JSON.stringify({
+          model: this.model,
+          max_tokens: 1024,
+          messages: [
+            {
+              role: 'user',
+              content: [
+                {
+                  type: 'image',
+                  source: { type: 'base64', media_type: mime, data: base64 },
+                },
+                { type: 'text', text: prompt },
+              ],
+            },
+          ],
+        }),
+        signal: ctrl.signal,
+      })
+
+      if (!res.ok) {
+        const t = await res.text().catch(() => '')
+        throw new Error(`Anthropic API ${res.status}: ${t.slice(0, 200)}`)
+      }
+
+      const json = await res.json()
+      return json.content?.[0]?.text?.trim() || '[No content returned]'
+    } finally {
+      clearTimeout(timer)
+    }
+  }
+}

package/src/providers/index.js

@@ -1,12 +1,34 @@
-import { GLMProvider } from './glm.js'
+import { OpenAICompatibleProvider } from './openai-compatible.js'
+import { ClaudeProvider } from './claude.js'
 import { resolveProviderConfig } from '../opencode.js'
 
-// Provider ID → provider class mapping
+const OPENAI_COMPATIBLE = {
+  'zhipuai-coding-plan': OpenAICompatibleProvider,
+  'zai-coding-plan': OpenAICompatibleProvider,
+  'z-ai': OpenAICompatibleProvider,
+  'zhipuai': OpenAICompatibleProvider,
+  'moonshot': OpenAICompatibleProvider,
+  'kimi': OpenAICompatibleProvider,
+  'minimax': OpenAICompatibleProvider,
+  'minimax-cn-coding-plan': OpenAICompatibleProvider,
+  'openai': OpenAICompatibleProvider,
+  'qwen': OpenAICompatibleProvider,
+  'dashscope': OpenAICompatibleProvider,
+  'doubao': OpenAICompatibleProvider,
+  'volcengine': OpenAICompatibleProvider,
+  'yi': OpenAICompatibleProvider,
+  'lingyiwanwu': OpenAICompatibleProvider,
+  'gemini': OpenAICompatibleProvider,
+  'google': OpenAICompatibleProvider,
+  'stepfun': OpenAICompatibleProvider,
+  'baichuan': OpenAICompatibleProvider,
+  'hunyuan': OpenAICompatibleProvider,
+}
+
 const PROVIDER_MAP = {
-  'zhipuai-coding-plan': GLMProvider,
-  'zai-coding-plan': GLMProvider,
-  'z-ai': GLMProvider,
-  'zhipuai': GLMProvider,
+  ...OPENAI_COMPATIBLE,
+  'anthropic': ClaudeProvider,
+  'claude': ClaudeProvider,
 }
 
 export function createProvider() {
@@ -14,7 +36,7 @@ export function createProvider() {
   const slashIdx = raw.indexOf('/')
   if (slashIdx === -1) {
     throw new Error(
-      `Invalid VISION_MODEL format: "${raw}". Expected "provider/model", e.g. "zhipuai-coding-plan/glm-4.6v"`,
+      `Invalid model format: "${raw}". Expected "provider/model", e.g. "zhipuai-coding-plan/glm-4.6v"`,
     )
   }
   const providerId = raw.slice(0, slashIdx)

package/src/providers/{glm.js → openai-compatible.js}

@@ -1,6 +1,6 @@
 import { VisionProvider } from './base.js'
 
-export class GLMProvider extends VisionProvider {
+export class OpenAICompatibleProvider extends VisionProvider {
   constructor(config) {
     super(config)
     this.apiKey = config.apiKey
@@ -9,13 +9,13 @@ export class GLMProvider extends VisionProvider {
     this.timeout = config.timeout || 60_000
 
     if (!this.apiKey) {
-      throw new Error('GLM API key not configured.')
+      throw new Error('API key not configured.')
     }
     if (!this.baseUrl) {
-      throw new Error('GLM base URL not configured.')
+      throw new Error('Base URL not configured.')
     }
     if (!this.model) {
-      throw new Error('GLM model not configured.')
+      throw new Error('Model not configured.')
     }
   }
 
@@ -48,7 +48,7 @@ export class GLMProvider extends VisionProvider {
 
       if (!res.ok) {
         const t = await res.text().catch(() => '')
-        throw new Error(`GLM API ${res.status}: ${t.slice(0, 200)}`)
+        throw new Error(`Vision API ${res.status}: ${t.slice(0, 200)}`)
       }
 
       const json = await res.json()