feat(image-recognition): independent key/url/model defaults

Make the HTTP vision provider's configuration independent of the main chat
model: the API key default moves from DEEPSEEK_API_KEY to IMAGE_RECOGNITION_API_KEY,
and the URL and model get image-recognition-specific defaults (Aliyun DashScope
compatible-mode and qwen3-vl-flash) instead of the chat model's deepseek-v4-flash.
The provider fails loud when no model is configured.

Co-Authored-By: Claude <noreply@anthropic.com>

packages/vision/image-recognition-http/README.md

@@ -8,10 +8,10 @@ URL and returns the recognized text.
 
 | Key | Type | Meaning |
 |---|---|---|
-| `baseURL` | string | Endpoint base; `/chat/completions` is appended. Blank inherits `$DSH_IMAGE_RECOGNITION_BASE_URL`, else unavailable. |
+| `baseURL` | string | Endpoint base; `/chat/completions` is appended. Defaults to `https://dashscope.aliyuncs.com/compatible-mode/v1` (Aliyun DashScope compatible-mode); `$DSH_IMAGE_RECOGNITION_BASE_URL` overrides. |
 | `apiKey` | string (secret) | Literal key; prefer `apiKeyEnv`. |
-| `apiKeyEnv` | string (credential-ref) | Credential reference resolved per recognition; defaults to `DEEPSEEK_API_KEY`. |
+| `apiKeyEnv` | string (credential-ref) | Credential reference resolved per recognition; defaults to `IMAGE_RECOGNITION_API_KEY` (distinct from the main chat model's `DEEPSEEK_API_KEY`). |
-| `model` | string | Vision model name; defaults to `deepseek-v4-flash`. |
+| `model` | string | Vision model name; defaults to `qwen3-vl-flash` (an image-recognition-specific default, independent of the chat model). |
 | `maxTokens` | number | Generated-token bound; defaults to 2048. |
 
 The endpoint and key are editable live through the `image-recognition-http`

packages/vision/image-recognition-http/src/index.ts

@@ -16,6 +16,7 @@ import type {} from '@deepseek-ai/dsh-session'
 import type {} from '@deepseek-ai/dsh-image-recognition'
 import {
   ImageRecognitionHttpProvider,
+  IMAGE_RECOGNITION_DEFAULT_BASE_URL,
   IMAGE_RECOGNITION_DEFAULT_MAX_TOKENS,
   IMAGE_RECOGNITION_DEFAULT_MODEL,
 } from './provider.ts'
@@ -23,6 +24,7 @@ import type { ImageRecognitionHttpProviderOptions } from './provider.ts'
 
 export {
   ImageRecognitionHttpProvider,
+  IMAGE_RECOGNITION_DEFAULT_BASE_URL,
   IMAGE_RECOGNITION_DEFAULT_MAX_TOKENS,
   IMAGE_RECOGNITION_DEFAULT_MODEL,
   IMAGE_RECOGNITION_HTTP_PROVIDER_ID,
@@ -35,17 +37,19 @@ export const name = 'image-recognition-http'
 /** The image-recognition seam this provider registers into. */
 export const inject = ['imageRecognition']
 
-const DEFAULT_API_KEY_ENV = 'DEEPSEEK_API_KEY'
+// Distinct from the main chat model's key: image recognition may use a
+// different provider (e.g. Aliyun DashScope) than the conversation LLM.
+const DEFAULT_API_KEY_ENV = 'IMAGE_RECOGNITION_API_KEY'
 
 /** Plugin config (all optional — `apply` fills env-var and constant defaults). */
 export interface Config {
   /** Literal API key; prefer {@link apiKeyEnv} so no secret enters configuration files. */
   apiKey?: string
-  /** Credential reference resolved for each recognition; defaults to `DEEPSEEK_API_KEY`. */
+  /** Credential reference resolved per recognition; defaults to `IMAGE_RECOGNITION_API_KEY`. */
   apiKeyEnv?: string
-  /** OpenAI-compatible endpoint base; `/chat/completions` is appended. */
+  /** OpenAI-compatible endpoint base; `/chat/completions` is appended. Defaults to Aliyun DashScope compatible-mode. */
   baseURL?: string
-  /** Vision model name. Defaults to `deepseek-v4-flash`. */
+  /** Vision model name; defaults to `qwen3-vl-flash` (an image-recognition-specific default, not the chat model). */
   model?: string
   /** Upper bound on generated tokens. Defaults to 2048. */
   maxTokens?: number
@@ -54,7 +58,7 @@ export interface Config {
 export const Config: z<Config> = z.object({
   apiKey: z.string().role('secret'),
   apiKeyEnv: z.string().role('credential-ref').default(DEFAULT_API_KEY_ENV),
-  baseURL: z.string(),
+  baseURL: z.string().default(IMAGE_RECOGNITION_DEFAULT_BASE_URL),
   model: z.string().default(IMAGE_RECOGNITION_DEFAULT_MODEL),
   maxTokens: z.number().step(1).min(1).default(IMAGE_RECOGNITION_DEFAULT_MAX_TOKENS),
 })
@@ -89,7 +93,7 @@ function resolveOptions(ctx: Context, config: Config): ImageRecognitionHttpProvi
     },
     baseURL: config.baseURL
       ?? launchEnvironmentOf(ctx).get(BASE_URL_ENV)?.value
-      ?? '',
+      ?? IMAGE_RECOGNITION_DEFAULT_BASE_URL,
     model: config.model ?? IMAGE_RECOGNITION_DEFAULT_MODEL,
     maxTokens: config.maxTokens ?? IMAGE_RECOGNITION_DEFAULT_MAX_TOKENS,
     recordRequest: (request) => {

packages/vision/image-recognition-http/src/provider.ts

@@ -32,8 +32,9 @@ declare module '@deepseek-ai/dsh-session/types' {
   }
 }
 
-export const IMAGE_RECOGNITION_DEFAULT_MODEL = 'deepseek-v4-flash'
+export const IMAGE_RECOGNITION_DEFAULT_MODEL = 'qwen3-vl-flash'
 export const IMAGE_RECOGNITION_DEFAULT_MAX_TOKENS = 2048
+export const IMAGE_RECOGNITION_DEFAULT_BASE_URL = 'https://dashscope.aliyuncs.com/compatible-mode/v1'
 
 const USER_AGENT = 'deepseek-harness/0.0.1'
 
@@ -115,6 +116,12 @@ export class ImageRecognitionHttpProvider implements ImageRecognitionProvider {
         'IMAGE_RECOGNITION_PROVIDER_CREDENTIAL_MISSING',
       )
     }
+    if (options.model.length === 0) {
+      throw new ImageRecognitionError(
+        'image-recognition provider has no model configured; set one in the plugin settings',
+        'IMAGE_RECOGNITION_PROVIDER_CREDENTIAL_MISSING',
+      )
+    }
     const source = await imageSource(request.image)
     const payload = {
       model: options.model,

packages/vision/image-recognition-http/tests/provider.spec.ts

@@ -2,7 +2,6 @@ import { afterEach, describe, expect, it, vi } from 'vitest'
 import type { ImageRecognitionRequest } from '@deepseek-ai/dsh-image-recognition'
 import {
   ImageRecognitionHttpProvider,
-  IMAGE_RECOGNITION_DEFAULT_MODEL,
   type ImageRecognitionHttpProviderOptions,
 } from '../src/provider.ts'
 
@@ -15,7 +14,7 @@ function makeProvider(overrides: Partial<ImageRecognitionHttpProviderOptions> =
   return new ImageRecognitionHttpProvider(() => ({
     resolveApiKey: async () => 'key',
     baseURL: 'https://vision.example.com/v1',
-    model: IMAGE_RECOGNITION_DEFAULT_MODEL,
+    model: 'vision-model',
     maxTokens: 128,
     ...overrides,
   }))
@@ -60,6 +59,11 @@ describe('ImageRecognitionHttpProvider.recognize', () => {
       .rejects.toThrow(expect.objectContaining({ code: 'IMAGE_RECOGNITION_PROVIDER_CREDENTIAL_MISSING' }))
   })
 
+  it('throws IMAGE_RECOGNITION_PROVIDER_CREDENTIAL_MISSING without a model', async () => {
+    await expect(makeProvider({ model: '' }).recognize(request))
+      .rejects.toThrow(expect.objectContaining({ code: 'IMAGE_RECOGNITION_PROVIDER_CREDENTIAL_MISSING' }))
+  })
+
   it('maps a non-2xx response to IMAGE_RECOGNITION_PROVIDER_ERROR', async () => {
     stubFetch({ ok: false, status: 429, json: async () => ({ error: { message: 'rate limited' } }) })
     await expect(makeProvider().recognize(request))