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.

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 : `${base}/api`
-            const data = await getJson(`${api}/tags`, bearer, fetchFn)
+            const data = await getJson(
+                `${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",