opencode-see-image 0.9.0 → 0.9.2

package/.claude/settings.local.json

@@ -0,0 +1,18 @@
+{
+  "permissions": {
+    "allow": [
+      "Bash(bun build *)",
+      "Bash(npm view *)",
+      "Bash(FILTER_BRANCH_SQUELCH_WARNING=1 git filter-branch -f --msg-filter 'sed \"/Co-Authored-By: Claude/d\" | sed -e :a -e \"/^\\\\n*$/{\\\\$d;N;ba\" -e \"}\"' HEAD~2..HEAD)",
+      "Bash(echo \"--- created, exit $? ---\")",
+      "Bash(node -p \"require\\('./package.json'\\).version\")",
+      "Bash(echo \"local package.json version: $\\(node -p \"require\\('./package.json'\\).version\" \\)\")",
+      "Bash(bun --version)",
+      "Bash(bun pm *)",
+      "Read(//Users/alfa/Documents/opencodeprojects/opencode-see-image/bun-types/**)",
+      "Bash(bun run *)",
+      "WebFetch(domain:docs.z.ai)",
+      "Bash(npm publish *)"
+    ]
+  }
+}

package/README.md

@@ -42,7 +42,7 @@ You need a connected vision-capable provider. The plugin auto-detects whichever
 2. Select **opencode** (OpenCode Zen)
 3. Paste your API key from [opencode.ai/auth](https://opencode.ai/auth)
 
-The plugin falls back to **big-pickle** (~12000ms). No subscription needed.
+The plugin falls back to **mimo-v2.5-free**. No subscription needed.
 
 ### Paid, w/ OpenCode Go
 1. Run `/connect` in opencode
@@ -55,7 +55,7 @@ The plugin prefers **minimax-m3** via opencode-go (~3000ms) when available.
 
 Set the `SEE_IMAGE_*` env vars to point at any Anthropic-Messages-compatible endpoint. See [Configuration](#configuration) below.
 
-**Resolution order:** explicit `SEE_IMAGE_API_KEY` env → configured `SEE_IMAGE_PROVIDER` → `opencode-go` (MiniMax M3) → `opencode` (big-pickle, free).
+**Resolution order:** explicit `SEE_IMAGE_API_KEY` env → configured `SEE_IMAGE_PROVIDER` → `opencode-go` (MiniMax M3) → `opencode` (mimo-v2.5-free, free).
 
 ## How it works
 
@@ -126,7 +126,8 @@ export SEE_IMAGE_MODEL="kimi-k2.7-code"
 
 | Model | Speed | Notes |
 |---|---|---|
-| `big-pickle` | ~12000ms | Free. Accurate. Default fallback when only Zen is connected. |
+| `mimo-v2.5-free` | — | Free. Default fallback when only Zen is connected (routed via CLI). |
+| `big-pickle` | ~12000ms | Free. Accurate. Alternative Zen fallback. |
 
 **Paid (OpenCode Go):**
 
@@ -139,7 +140,7 @@ export SEE_IMAGE_MODEL="kimi-k2.7-code"
 
 ## Updating
 
-**Auto-update (built in):** the plugin checks npm for a newer version on every opencode startup. If one exists, it updates itself via `opencode plugin --force` (uses opencode's bundled bun, no global bun needed) and shows a toast: *"opencode-see-image updated to X.Y.Z, restart opencode to apply"*. You just need to restart opencode to load the new version. Nothing to configure.
+**Auto-update (built in):** the plugin checks npm for a newer version on startup. If one exists, it updates itself via `opencode plugin --force` (uses opencode's bundled bun, no global bun needed) and shows a toast: *"opencode-see-image updated to X.Y.Z, restart opencode to apply"*. You just need to restart opencode to load the new version. Nothing to configure.
 
 **Manual update** (if you want to force it now):
 ```bash

package/index.ts

@@ -11,7 +11,7 @@ const ENDPOINT =
   "https://opencode.ai/zen/go/v1/messages"
 const MODEL = process.env.SEE_IMAGE_MODEL || "minimax-m3"
 const PROVIDER_ID = process.env.SEE_IMAGE_PROVIDER || "opencode-go"
-const TIMEOUT = parseInt(process.env.SEE_IMAGE_TIMEOUT || "10000", 10)
+const TIMEOUT = parseInt(process.env.SEE_IMAGE_TIMEOUT || "30000", 10)
 const API_VERSION = process.env.SEE_IMAGE_API_VERSION || "2023-06-01"
 const USER_AGENT =
   process.env.SEE_IMAGE_USER_AGENT ||
@@ -47,8 +47,9 @@ function resolveFromDb(
   const dbPath = opencodeDbPath()
   if (!fs.existsSync(dbPath)) return null
 
+  let db: Database | undefined
   try {
-    const db = new Database(dbPath, { readonly: true })
+    db = new Database(dbPath, { readonly: true })
     let rows: Array<{ data: string }>
 
     if (!filename || filename === "clipboard") {
@@ -98,8 +99,6 @@ function resolveFromDb(
       }
     }
 
-    db.close()
-
     if (!rows.length) return null
     const part = JSON.parse(rows[0].data)
     const url: string = part.url || ""
@@ -112,6 +111,8 @@ function resolveFromDb(
     }
   } catch {
     return null
+  } finally {
+    db?.close()
   }
 }
 
@@ -217,7 +218,6 @@ function readProviderKey(providerID: string): string | null {
 
 async function seeImageViaSDK(
   client: any,
-  $: any,
   dataUrl: string,
   mediaType: string,
   prompt: string,
@@ -225,21 +225,54 @@ async function seeImageViaSDK(
 ): Promise<{ text: string; model: string; provider: string }> {
   const errors: string[] = []
 
-  // Write image to a temp file so the server can read it directly
   const b64 = dataUrl.split(",")[1] || ""
-  const tmpPath = path.join(os.tmpdir(), `see-image-${Date.now()}.png`)
-  try {
-    fs.writeFileSync(tmpPath, Buffer.from(b64, "base64"))
-  } catch {}
+  const ext =
+    Object.entries(EXT_MEDIA).find(([, m]) => m === mediaType)?.[0] || "png"
+
+  // The free CLI fallback needs the image on disk. Write it lazily and only
+  // once, so the common SDK/dataURL path never touches the filesystem. Use the
+  // real extension so the CLI can sniff the type correctly.
+  let tmpPath: string | null = null
+  const ensureTmpFile = (): string | null => {
+    if (tmpPath) return tmpPath
+    const p = path.join(os.tmpdir(), `see-image-${Date.now()}.${ext}`)
+    try {
+      fs.writeFileSync(p, Buffer.from(b64, "base64"))
+      tmpPath = p
+    } catch {
+      return null
+    }
+    return tmpPath
+  }
 
-  // For free opencode models, use CLI instead of SDK (SDK returns empty)
+  // For free opencode models, use CLI instead of SDK (SDK returns empty).
+  // Use Bun.spawn (not $) so we get a killable handle: Bun's $ ShellPromise
+  // has no .kill(), so racing it against a timeout would leak the process.
+  // We kill the child on both timeout and external abort.
   const freeFallback = async (modelID: string, userPrompt: string): Promise<string | null> => {
+    const filePath = ensureTmpFile()
+    if (!filePath) return null
+    const proc = Bun.spawn(
+      [
+        "opencode",
+        "run",
+        "-f",
+        filePath,
+        "-m",
+        `opencode/${modelID}`,
+        userPrompt,
+        "--format",
+        "json",
+        "--dangerously-skip-permissions",
+      ],
+      { stdout: "pipe", stderr: "ignore" },
+    )
+    const timer = setTimeout(() => proc.kill(), TIMEOUT)
+    const onAbort = () => proc.kill()
+    abort?.addEventListener("abort", onAbort)
     try {
-      const controller = new AbortController()
-      const timer = setTimeout(() => controller.abort(), TIMEOUT)
-      const proc = $`opencode run -f ${tmpPath} -m opencode/${modelID} ${userPrompt} --format json --dangerously-skip-permissions`
-      const out = await proc.text()
-      clearTimeout(timer)
+      const out = await new Response(proc.stdout).text()
+      await proc.exited
       for (const line of out.split("\n").filter(Boolean)) {
         try {
           const parsed = JSON.parse(line)
@@ -248,11 +281,13 @@ async function seeImageViaSDK(
           }
         } catch {}
       }
-    } catch {}
+    } catch {} finally {
+      clearTimeout(timer)
+      abort?.removeEventListener("abort", onAbort)
+    }
     return null
   }
 
-  const fileUrl = tmpPath
   let result: { text: string; model: string; provider: string } | undefined
 
   try {
@@ -279,7 +314,15 @@ async function seeImageViaSDK(
 
       let sessionID: string | undefined
       try {
-        const sessionRes = await client.session.create({ body: {} })
+        const sessionRes = await Promise.race([
+          client.session.create({ body: {} }),
+          new Promise<never>((_, reject) =>
+            setTimeout(
+              () => reject(new Error(`session.create timed out after ${TIMEOUT}ms`)),
+              TIMEOUT,
+            ),
+          ),
+        ])
         sessionID = sessionRes.data?.id
         if (!sessionID) {
           errors.push(`${providerID}/${modelID}: no session ID`)
@@ -287,22 +330,29 @@ async function seeImageViaSDK(
         }
 
         const controller = new AbortController()
+        const onAbort = () => controller.abort()
+        abort?.addEventListener("abort", onAbort)
         const timer = setTimeout(() => controller.abort(), TIMEOUT)
-        const res = await client.session.prompt({
-          path: { id: sessionID },
-          body: {
-            model: { providerID, modelID },
-            parts: [
-              { type: "file", mime: mediaType, url: providerID === "opencode" ? fileUrl : dataUrl },
-              { type: "text", text: prompt },
-            ],
-            tools: {},
-            system:
-              "You are a vision assistant. Describe the image accurately and concisely. Answer with text only.",
-          },
-          signal: controller.signal,
-        })
-        clearTimeout(timer)
+        let res
+        try {
+          res = await client.session.prompt({
+            path: { id: sessionID },
+            body: {
+              model: { providerID, modelID },
+              parts: [
+                { type: "file", mime: mediaType, url: dataUrl },
+                { type: "text", text: prompt },
+              ],
+              tools: {},
+              system:
+                "You are a vision assistant. Describe the image accurately and concisely. Answer with text only.",
+            },
+            signal: controller.signal,
+          })
+        } finally {
+          clearTimeout(timer)
+          abort?.removeEventListener("abort", onAbort)
+        }
 
         const parts = res.data?.parts ?? []
         const text = (parts as any[])
@@ -330,7 +380,10 @@ async function seeImageViaSDK(
 
     if (!result) {
       const apiKey =
-        process.env.SEE_IMAGE_API_KEY || readProviderKey("opencode-go")
+        process.env.SEE_IMAGE_API_KEY ||
+        (process.env.SEE_IMAGE_PROVIDER &&
+          readProviderKey(process.env.SEE_IMAGE_PROVIDER)) ||
+        readProviderKey("opencode-go")
       if (apiKey) {
         try {
           result = await seeImageViaHTTP(b64, mediaType, prompt, abort, apiKey)
@@ -344,13 +397,15 @@ async function seeImageViaSDK(
 
     const errMsg = errors.join("; ")
     const hint = errMsg.includes("usage limit")
-      ? ` Enable usage from your balance at https://opencode.ai/workspace/wrk_01KVARG0A0Y87XV5JYBNJ0WRXB/go`
+      ? ` Enable usage from your balance in your opencode workspace at https://opencode.ai/workspace`
       : ""
     throw new Error(
       `see_image: SDK vision call failed for all candidates. ${errMsg}.${hint}`,
     )
   } finally {
-    try { fs.unlinkSync(tmpPath) } catch {}
+    if (tmpPath) {
+      try { fs.unlinkSync(tmpPath) } catch {}
+    }
   }
 }
 
@@ -486,7 +541,6 @@ const SeeImagePlugin: Plugin = async (ctx) => {
       } else {
         result = await seeImageViaSDK(
           client,
-          $,
           resolved.dataUrl,
           resolved.mediaType,
           prompt,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "opencode-see-image",
-  "version": "0.9.0",
+  "version": "0.9.2",
   "description": "Give non-vision opencode models the ability to see images/screenshots by routing them to a vision-capable model (MiniMax M3 via opencode-go by default).",
   "type": "module",
   "main": "index.ts",
@@ -23,6 +23,6 @@
   "license": "MIT",
   "dependencies": {
     "@opencode-ai/plugin": "^1.15.0",
-    "opencode-plugin-update-kit": "^0.1.0"
+    "opencode-plugin-update-kit": "^0.2.0"
   }
 }