fix(ollama): send chat requests to the /api path

The Ollama SDK appends /chat to the base URL, and Ollama serves chat at
/api/chat. The model list already added the missing /api, so with
"http://localhost:11434" (the address our docs showed) models were listed
but every chat request went to /chat and got a 404. An OpenAI-style
".../v1" address went to /v1/chat.

ollamaApiUrl() turns the server address, ".../v1" and ".../api" into
".../api". Chat, the model list and the settings dialog's request URL
hint all use it. The docs now show http://localhost:11434/api, which also
works on released versions.

components/chat-example-panel.tsx

@@ -77,6 +77,7 @@ export default function ExamplePanel({
     minimal?: boolean
 }) {
     const dict = useDictionary()
+    const isSelfHosted = process.env.NEXT_PUBLIC_SELFHOSTED === "true"
 
     const handleReplicateFlowchart = async () => {
         setInput("Replicate this flowchart.")
@@ -125,29 +126,31 @@ export default function ExamplePanel({
         <div className={minimal ? "" : "py-6 px-2 animate-fade-in"}>
             {!minimal && (
                 <>
-                    {/* MCP Server Notice */}
+                    {/* MCP Server Notice, hidden on self-hosted deployments */}
-                    <a
+                    {!isSelfHosted && (
-                        href="https://github.com/DayuanJiang/next-ai-draw-io/tree/main/packages/mcp-server"
+                        <a
-                        target="_blank"
+                            href="https://github.com/DayuanJiang/next-ai-draw-io/tree/main/packages/mcp-server"
-                        rel="noopener noreferrer"
+                            target="_blank"
-                        className="block mb-4 p-3 rounded-xl bg-gradient-to-r from-purple-500/10 to-blue-500/10 border border-purple-500/20 hover:border-purple-500/40 transition-colors group"
+                            rel="noopener noreferrer"
-                    >
+                            className="block mb-4 p-3 rounded-xl bg-gradient-to-r from-purple-500/10 to-blue-500/10 border border-purple-500/20 hover:border-purple-500/40 transition-colors group"
-                        <div className="flex items-center gap-3">
+                        >
-                            <div className="w-8 h-8 rounded-lg bg-purple-500/20 flex items-center justify-center shrink-0">
+                            <div className="flex items-center gap-3">
-                                <Terminal className="w-4 h-4 text-purple-500" />
+                                <div className="w-8 h-8 rounded-lg bg-purple-500/20 flex items-center justify-center shrink-0">
-                            </div>
+                                    <Terminal className="w-4 h-4 text-purple-500" />
-                            <div className="min-w-0">
+                                </div>
-                                <div className="flex items-center gap-2">
+                                <div className="min-w-0">
-                                    <span className="text-sm font-medium text-foreground group-hover:text-purple-500 transition-colors">
+                                    <div className="flex items-center gap-2">
-                                        {dict.examples.mcpServer}
+                                        <span className="text-sm font-medium text-foreground group-hover:text-purple-500 transition-colors">
-                                    </span>
+                                            {dict.examples.mcpServer}
+                                        </span>
+                                    </div>
+                                    <p className="text-xs text-muted-foreground">
+                                        {dict.examples.mcpDescription}
+                                    </p>
                                 </div>
-                                <p className="text-xs text-muted-foreground">
-                                    {dict.examples.mcpDescription}
-                                </p>
                             </div>
-                        </div>
+                        </a>
-                    </a>
+                    )}
 
                     {/* Welcome section */}
                     <div className="text-center mb-6">

docs/cn/ai-providers.md

@@ -192,7 +192,7 @@ MODELSCOPE_BASE_URL=https://your-custom-endpoint
 可选的自定义 URL：
 
 ```bash
-OLLAMA_BASE_URL=http://localhost:11434
+OLLAMA_BASE_URL=http://localhost:11434/api
 ```
 
 ### Vercel AI Gateway

docs/en/ai-providers.md

@@ -194,7 +194,7 @@ AI_MODEL=llama3.2
 Optional custom URL:
 
 ```bash
-OLLAMA_BASE_URL=http://localhost:11434
+OLLAMA_BASE_URL=http://localhost:11434/api
 ```
 
 ### ModelScope

docs/ja/ai-providers.md

@@ -179,7 +179,7 @@ AI_MODEL=llama3.2
 任意のカスタム URL:
 
 ```bash
-OLLAMA_BASE_URL=http://localhost:11434
+OLLAMA_BASE_URL=http://localhost:11434/api
 ```
 
 ### ModelScope

env.example

@@ -1,6 +1,6 @@
 # AI Provider Configuration
 # AI_PROVIDER: Which provider to use
-# Options: bedrock, openai, anthropic, google, vertexai, azure, ollama, openrouter, aihubmix, deepseek, siliconflow, gateway, novita
+# Options: bedrock, openai, anthropic, google, vertexai, azure, ollama, openrouter, aihubmix, deepseek, siliconflow, sglang, gateway, edgeone, doubao, modelscope, glm, qwen, qiniu, kimi, minimax, novita, mimo, atlascloud
 # Default: bedrock
 AI_PROVIDER=bedrock
 
@@ -70,7 +70,7 @@ AI_MODEL=global.anthropic.claude-sonnet-4-5-20250929-v1:0
 # AZURE_REASONING_SUMMARY=detailed
 
 # Ollama Configuration (Local or Cloud)
-# OLLAMA_BASE_URL=https://ollama.com/api  # Optional: Ollama Cloud; defaults to local Ollama (http://127.0.0.1:11434)
+# OLLAMA_BASE_URL=https://ollama.com/api  # Optional: Ollama Cloud; defaults to local Ollama (http://127.0.0.1:11434/api)
 # OLLAMA_API_KEY=your-ollama-cloud-api-key    # Optional: For Ollama Cloud or authenticated remote instances
 # OLLAMA_ENABLE_THINKING=true                 # Optional: Enable thinking for models that support it (e.g., qwen3)
 
@@ -168,6 +168,7 @@ AI_MODEL=global.anthropic.claude-sonnet-4-5-20250929-v1:0
 # which triggers the client UI to display messages suggesting self-hosting or sponsorship.
 # This switch allows self-hosted users to provide custom messages in response to a 429 code,
 # in messageTokenSelfHosted, messageApiSelfHosted, and tipSelfHosted translation strings.
+# It also turns off the MCP server advertisement in the chat examples.
 # NEXT_PUBLIC_SELFHOSTED=true
 
 # Minimax Configuration (Optional)

lib/admin/settings.ts

@@ -125,7 +125,12 @@ let writableCache: boolean | null = null
 export function isSettingsWritable(): boolean {
     if (writableCache !== null) return writableCache
     try {
-        const dir = path.dirname(getSettingsPath())
+        const filePath = getSettingsPath()
+        if (fs.existsSync(filePath) && !fs.statSync(filePath).isFile()) {
+            writableCache = false
+            return writableCache
+        }
+        const dir = path.dirname(filePath)
         fs.mkdirSync(dir, { recursive: true })
         fs.accessSync(dir, fs.constants.W_OK)
         writableCache = true

lib/ai-providers.ts

@@ -25,6 +25,7 @@ import { getEnvFallback } from "@/lib/admin/settings"
 import { isPrivateUrl, redirectGuardedFetch } from "@/lib/ssrf-protection"
 import {
     normalizeBaseUrl,
+    ollamaApiUrl,
     PROVIDER_INFO,
     type ProviderName,
 } from "@/lib/types/model-config"
@@ -1025,7 +1026,7 @@ export function getAIModel(clientOverrides?: ClientOverrides): ModelConfig {
                     ? PROVIDER_INFO.ollama.defaultBaseUrl
                     : resolveBaseUrlEnv(overrides, "OLLAMA_BASE_URL"))
             model = createOllama({
-                ...(baseURL && { baseURL }),
+                ...(baseURL && { baseURL: ollamaApiUrl(baseURL) }),
                 ...(apiKey && {
                     headers: { Authorization: `Bearer ${apiKey}` },
                 }),

lib/provider-models.ts

@@ -3,6 +3,7 @@ import { getModelInfo } from "@/lib/model-catalog"
 import { readLimitedBody } from "@/lib/read-limited-body"
 import {
     normalizeBaseUrl,
+    ollamaApiUrl,
     PROVIDER_INFO,
     type ProviderName,
 } from "@/lib/types/model-config"
@@ -200,8 +201,11 @@ export async function listProviderModels(
             break
         }
         case "ollama": {
-            const api = base.endsWith("/api") ? base : `${base}/api`
+            const data = await getJson(
-            const data = await getJson(`${api}/tags`, bearer, fetchFn)
+                `${ollamaApiUrl(base)}/tags`,
+                bearer,
+                fetchFn,
+            )
             models = (data.models ?? []).map((m: { name: string }) => ({
                 id: m.name,
             }))

lib/types/model-config.ts

@@ -622,6 +622,15 @@ export function normalizeBaseUrl(url: string): string {
         .replace(/\/(?:chat\/completions|completions|messages|responses)$/, "")
 }
 
+/**
+ * Ollama's native API root, which the SDK appends /chat to and the model
+ * list /tags. Users often enter the server address ("http://localhost:11434")
+ * or its OpenAI-compatible one (".../v1"); both get /api.
+ */
+export function ollamaApiUrl(baseUrl: string): string {
+    return `${normalizeBaseUrl(baseUrl).replace(/\/(?:api|v1)$/, "")}/api`
+}
+
 /** Where a chat request goes for a base URL, or null when the SDK decides */
 export function chatRequestUrl(
     provider: ProviderName,
@@ -630,6 +639,7 @@ export function chatRequestUrl(
     const url = normalizeBaseUrl(baseUrl)
     if (!url) return null
     if (provider === "anthropic") return `${url}/messages`
+    if (provider === "ollama") return `${ollamaApiUrl(url)}/chat`
     // These SDKs build their own paths (or, for MiniMax, pick the protocol
     // from the URL)
     const ownPaths: ProviderName[] = [
@@ -637,7 +647,6 @@ export function chatRequestUrl(
         "vertexai",
         "azure",
         "bedrock",
-        "ollama",
         "gateway",
         "minimax",
         "edgeone",

package.json

@@ -67,7 +67,6 @@
         "@radix-ui/react-use-controllable-state": "^1.2.2",
         "@xmldom/xmldom": "^0.9.8",
         "ai": "^6.0.300",
-        "base-64": "^1.0.0",
         "class-variance-authority": "^0.7.1",
         "clsx": "^2.1.1",
         "cmdk": "^1.1.1",

packages/mcp-server/scripts/check-package.mjs

@@ -11,9 +11,11 @@ const REQUIRED = [
     "dist/preview/preview.js",
 ]
 
-const [pack] = JSON.parse(
+const output = JSON.parse(
     execSync("npm pack --dry-run --json", { encoding: "utf8" }),
 )
+// npm 11 prints a list, npm 12 an object keyed by package name
+const pack = Array.isArray(output) ? output[0] : Object.values(output)[0]
 const files = new Set(pack.files.map((f) => f.path))
 const missing = REQUIRED.filter((f) => !files.has(f))
 if (missing.length > 0) {

tests/unit/admin-settings.test.ts

@@ -154,9 +154,9 @@ describe("isSettingsWritable", () => {
         expect(isSettingsWritable()).toBe(true)
     })
 
-    it("returns false for an unwritable path", () => {
+    it("returns false when the settings path is a directory", () => {
         _resetForTests()
-        process.env.SETTINGS_FILE = "/nonexistent-root-dir/settings.json"
+        process.env.SETTINGS_FILE = tmpDir
         expect(isSettingsWritable()).toBe(false)
     })
 })
@@ -170,7 +170,9 @@ describe("settings file on disk", () => {
             version: 1,
             values: { TEST_ADMIN_VAR: "secret" },
         })
-        const mode = fs.statSync(filePath).mode & 0o777
+        if (process.platform !== "win32") {
-        expect(mode).toBe(0o600)
+            const mode = fs.statSync(filePath).mode & 0o777
+            expect(mode).toBe(0o600)
+        }
     })
 })

tests/unit/ai-providers.test.ts

@@ -434,7 +434,7 @@ describe("Ollama API key security", () => {
 
         expect(createOllamaMock).toHaveBeenCalledWith(
             expect.objectContaining({
-                baseURL: "https://my-ollama.com",
+                baseURL: "https://my-ollama.com/api",
                 headers: { Authorization: "Bearer client-key" },
             }),
         )
@@ -448,7 +448,7 @@ describe("Ollama API key security", () => {
 
         expect(createOllamaMock).toHaveBeenCalledWith(
             expect.objectContaining({
-                baseURL: "https://cloud.ollama.com",
+                baseURL: "https://cloud.ollama.com/api",
                 headers: { Authorization: "Bearer server-key" },
             }),
         )
@@ -483,7 +483,24 @@ describe("Ollama API key security", () => {
 
         expect(createOllamaMock).toHaveBeenCalledTimes(1)
         const callArgs = createOllamaMock.mock.calls[0][0]
-        expect(callArgs.baseURL).toBe("https://my-ollama.com")
+        expect(callArgs.baseURL).toBe("https://my-ollama.com/api")
         expect(callArgs).not.toHaveProperty("headers")
     })
+
+    it("sends chat to Ollama's /api for a server address or a /v1 URL", () => {
+        delete process.env.OLLAMA_API_KEY
+
+        for (const baseUrl of [
+            "http://localhost:11434",
+            "http://localhost:11434/",
+            "http://localhost:11434/v1",
+            "http://localhost:11434/api",
+        ]) {
+            createOllamaMock.mockClear()
+            getAIModel({ provider: "ollama", baseUrl, modelId: "llama3.2" })
+            expect(createOllamaMock.mock.calls[0][0].baseURL).toBe(
+                "http://localhost:11434/api",
+            )
+        }
+    })
 })

tests/unit/base-url.test.ts

@@ -1,5 +1,9 @@
 import { describe, expect, it } from "vitest"
-import { chatRequestUrl, normalizeBaseUrl } from "@/lib/types/model-config"
+import {
+    chatRequestUrl,
+    normalizeBaseUrl,
+    ollamaApiUrl,
+} from "@/lib/types/model-config"
 
 describe("normalizeBaseUrl", () => {
     it("drops spaces, trailing slashes and a pasted endpoint path", () => {
@@ -32,6 +36,9 @@ describe("chatRequestUrl", () => {
         expect(
             chatRequestUrl("anthropic", "https://proxy.example.com/v1"),
         ).toBe("https://proxy.example.com/v1/messages")
+        expect(chatRequestUrl("ollama", "http://localhost:11434")).toBe(
+            "http://localhost:11434/api/chat",
+        )
     })
 
     it("stays out of the way for SDKs that build their own paths", () => {
@@ -40,3 +47,21 @@ describe("chatRequestUrl", () => {
         expect(chatRequestUrl("glm", "  ")).toBeNull()
     })
 })
+
+describe("ollamaApiUrl", () => {
+    it("points at Ollama's /api whatever form the address takes", () => {
+        for (const url of [
+            "http://localhost:11434",
+            "http://localhost:11434/",
+            "http://localhost:11434/api",
+            "http://localhost:11434/api/",
+            "http://localhost:11434/v1",
+            "http://localhost:11434/v1/chat/completions",
+        ]) {
+            expect(ollamaApiUrl(url)).toBe("http://localhost:11434/api")
+        }
+        expect(ollamaApiUrl("https://ollama.com/api")).toBe(
+            "https://ollama.com/api",
+        )
+    })
+})