Autopoll Ollama models for inference endpoints
This commit is contained in:
@@ -0,0 +1,86 @@
|
||||
import { NextResponse } from "next/server"
|
||||
|
||||
const DEFAULT_OLLAMA_URL = "http://127.0.0.1:11434"
|
||||
const OLLAMA_BASE_URL = (process.env.OLLAMA_BASE_URL || DEFAULT_OLLAMA_URL).replace(/\/+$/, "")
|
||||
|
||||
type OllamaTag = {
|
||||
name?: string
|
||||
model?: string
|
||||
modified_at?: string
|
||||
size?: number
|
||||
digest?: string
|
||||
details?: {
|
||||
family?: string
|
||||
parameter_size?: string
|
||||
quantization_level?: string
|
||||
}
|
||||
}
|
||||
|
||||
function formatBytes(size: unknown) {
|
||||
if (typeof size !== "number" || !Number.isFinite(size) || size <= 0) return null
|
||||
const units = ["B", "KB", "MB", "GB", "TB"]
|
||||
let value = size
|
||||
let unit = 0
|
||||
while (value >= 1024 && unit < units.length - 1) {
|
||||
value /= 1024
|
||||
unit += 1
|
||||
}
|
||||
return `${value >= 10 || unit === 0 ? value.toFixed(0) : value.toFixed(1)} ${units[unit]}`
|
||||
}
|
||||
|
||||
export async function GET() {
|
||||
const startedAt = Date.now()
|
||||
const controller = new AbortController()
|
||||
const timeout = setTimeout(() => controller.abort(), 2500)
|
||||
|
||||
try {
|
||||
const response = await fetch(`${OLLAMA_BASE_URL}/api/tags`, {
|
||||
cache: "no-store",
|
||||
signal: controller.signal,
|
||||
})
|
||||
const data = await response.json()
|
||||
if (!response.ok) {
|
||||
throw new Error(data?.error || `Ollama returned HTTP ${response.status}`)
|
||||
}
|
||||
|
||||
const models = Array.isArray(data?.models)
|
||||
? data.models.map((item: OllamaTag) => ({
|
||||
name: item.model || item.name || "",
|
||||
model: item.model || item.name || "",
|
||||
modifiedAt: item.modified_at || null,
|
||||
size: typeof item.size === "number" ? item.size : null,
|
||||
sizeLabel: formatBytes(item.size),
|
||||
digest: item.digest || null,
|
||||
family: item.details?.family || null,
|
||||
parameterSize: item.details?.parameter_size || null,
|
||||
quantization: item.details?.quantization_level || null,
|
||||
})).filter((item: { name: string }) => item.name)
|
||||
: []
|
||||
|
||||
return NextResponse.json({
|
||||
ok: true,
|
||||
available: true,
|
||||
baseUrl: OLLAMA_BASE_URL,
|
||||
models,
|
||||
count: models.length,
|
||||
elapsedMs: Date.now() - startedAt,
|
||||
})
|
||||
} catch (error) {
|
||||
const message = error instanceof Error && error.name === "AbortError"
|
||||
? "Ollama probe timed out"
|
||||
: error instanceof Error
|
||||
? error.message
|
||||
: "Failed to fetch Ollama models"
|
||||
return NextResponse.json({
|
||||
ok: false,
|
||||
available: false,
|
||||
baseUrl: OLLAMA_BASE_URL,
|
||||
models: [],
|
||||
count: 0,
|
||||
elapsedMs: Date.now() - startedAt,
|
||||
error: message,
|
||||
}, { status: 200 })
|
||||
} finally {
|
||||
clearTimeout(timeout)
|
||||
}
|
||||
}
|
||||
@@ -1,6 +1,6 @@
|
||||
"use client"
|
||||
|
||||
import { useEffect, useMemo, useState } from "react"
|
||||
import { useCallback, useEffect, useMemo, useState } from "react"
|
||||
|
||||
type InferenceRoute = {
|
||||
configured: boolean
|
||||
@@ -24,6 +24,16 @@ type InferenceResponse = {
|
||||
providers: ProviderSummary[]
|
||||
}
|
||||
|
||||
type OllamaModel = {
|
||||
name: string
|
||||
model: string
|
||||
modifiedAt: string | null
|
||||
sizeLabel: string | null
|
||||
family: string | null
|
||||
parameterSize: string | null
|
||||
quantization: string | null
|
||||
}
|
||||
|
||||
const providerTypeOptions = [
|
||||
{ value: "openai", label: "OpenAI compatible" },
|
||||
{ value: "nvidia", label: "NVIDIA" },
|
||||
@@ -55,11 +65,15 @@ export default function InferenceEndpointPanel() {
|
||||
const [route, setRoute] = useState<"gateway" | "system">("gateway")
|
||||
const [timeout, setTimeoutValue] = useState("0")
|
||||
const [noVerify, setNoVerify] = useState(true)
|
||||
const [ollamaModels, setOllamaModels] = useState<OllamaModel[]>([])
|
||||
const [ollamaLoading, setOllamaLoading] = useState(false)
|
||||
const [ollamaMessage, setOllamaMessage] = useState("")
|
||||
|
||||
const activeProvider = useMemo(
|
||||
() => providers.find((provider) => provider.name === gateway.provider) ?? null,
|
||||
[providers, gateway.provider]
|
||||
)
|
||||
const isOllamaEndpoint = name.toLowerCase().includes("ollama") || baseUrl.includes("11434")
|
||||
|
||||
async function load() {
|
||||
try {
|
||||
@@ -84,12 +98,45 @@ export default function InferenceEndpointPanel() {
|
||||
load()
|
||||
}, [])
|
||||
|
||||
const loadOllamaModels = useCallback(async ({ silent = false }: { silent?: boolean } = {}) => {
|
||||
try {
|
||||
setOllamaLoading(true)
|
||||
if (!silent) setOllamaMessage("")
|
||||
const response = await fetch("/api/ollama/models", { cache: "no-store" })
|
||||
const data = await response.json()
|
||||
if (!data.available) {
|
||||
setOllamaModels([])
|
||||
setOllamaMessage(data.error || "Ollama is not reachable")
|
||||
return
|
||||
}
|
||||
const models = Array.isArray(data.models) ? data.models : []
|
||||
setOllamaModels(models)
|
||||
setOllamaMessage(models.length > 0 ? `${models.length} Ollama model${models.length === 1 ? "" : "s"} available.` : "Ollama is reachable but has no models installed.")
|
||||
if (models[0]?.name) setModel((current) => current || models[0].name)
|
||||
} catch (error) {
|
||||
setOllamaModels([])
|
||||
setOllamaMessage(error instanceof Error ? error.message : "Failed to fetch Ollama models")
|
||||
} finally {
|
||||
setOllamaLoading(false)
|
||||
}
|
||||
}, [])
|
||||
|
||||
useEffect(() => {
|
||||
if (!isOllamaEndpoint) return
|
||||
loadOllamaModels({ silent: true })
|
||||
const interval = window.setInterval(() => loadOllamaModels({ silent: true }), 10000)
|
||||
return () => window.clearInterval(interval)
|
||||
}, [isOllamaEndpoint, loadOllamaModels])
|
||||
|
||||
function selectProvider(provider: ProviderSummary) {
|
||||
if (!provider.name) return
|
||||
setName(provider.name)
|
||||
setType(provider.type || "openai")
|
||||
setCredentialKey(provider.credentialKeys[0] || "OPENAI_API_KEY")
|
||||
setApiKey("")
|
||||
if (provider.name.toLowerCase().includes("ollama")) {
|
||||
setBaseUrl("http://host.docker.internal:11434/v1")
|
||||
setApiKey("")
|
||||
}
|
||||
setMessage(`Loaded ${provider.name}. Enter a model and save to make it active.`)
|
||||
}
|
||||
|
||||
@@ -181,7 +228,13 @@ export default function InferenceEndpointPanel() {
|
||||
</div>
|
||||
<div className="space-y-2 md:col-span-2">
|
||||
<FieldLabel>Model</FieldLabel>
|
||||
<input value={model} onChange={(event) => setModel(event.target.value)} placeholder="provider/model-name" className="w-full rounded-sm border border-[var(--border-subtle)] bg-[var(--background-tertiary)] px-3 py-2 text-sm font-mono text-[var(--foreground)] focus:outline-none focus:border-[var(--nvidia-green)]" />
|
||||
{isOllamaEndpoint && ollamaModels.length > 0 ? (
|
||||
<select value={model} onChange={(event) => setModel(event.target.value)} className="w-full rounded-sm border border-[var(--border-subtle)] bg-[var(--background-tertiary)] px-3 py-2 text-sm font-mono text-[var(--foreground)] focus:outline-none focus:border-[var(--nvidia-green)]">
|
||||
{ollamaModels.map((item) => <option key={item.name} value={item.name}>{item.name}</option>)}
|
||||
</select>
|
||||
) : (
|
||||
<input value={model} onChange={(event) => setModel(event.target.value)} placeholder="provider/model-name" className="w-full rounded-sm border border-[var(--border-subtle)] bg-[var(--background-tertiary)] px-3 py-2 text-sm font-mono text-[var(--foreground)] focus:outline-none focus:border-[var(--nvidia-green)]" />
|
||||
)}
|
||||
</div>
|
||||
<div className="space-y-2 md:col-span-2">
|
||||
<FieldLabel>Endpoint URL</FieldLabel>
|
||||
@@ -213,6 +266,32 @@ export default function InferenceEndpointPanel() {
|
||||
Skip provider verification while saving
|
||||
</label>
|
||||
|
||||
{isOllamaEndpoint && (
|
||||
<div className="rounded-sm border border-[var(--border-subtle)] bg-[var(--background-tertiary)] p-4">
|
||||
<div className="flex items-center justify-between gap-4">
|
||||
<div>
|
||||
<h3 className="text-xs uppercase tracking-wider text-[var(--foreground)]">Ollama Models</h3>
|
||||
<p className="mt-1 text-xs text-[var(--foreground-dim)]">{ollamaMessage || "Polling local Ollama every 10 seconds."}</p>
|
||||
</div>
|
||||
<button type="button" onClick={() => loadOllamaModels()} disabled={ollamaLoading} className="px-3 py-2 rounded-sm bg-[var(--background)] text-[var(--foreground)] text-xs font-mono uppercase tracking-wider hover:bg-[var(--background-panel)] disabled:opacity-50">
|
||||
{ollamaLoading ? "Polling..." : "Poll"}
|
||||
</button>
|
||||
</div>
|
||||
{ollamaModels.length > 0 && (
|
||||
<div className="mt-3 grid grid-cols-1 md:grid-cols-2 gap-2">
|
||||
{ollamaModels.map((item) => (
|
||||
<button key={item.name} type="button" onClick={() => setModel(item.name)} className={`rounded-sm border p-3 text-left ${model === item.name ? "border-[var(--nvidia-green)] bg-[rgba(118,185,0,0.08)]" : "border-[var(--border-subtle)] bg-[var(--background)]"}`}>
|
||||
<div className="text-xs font-mono text-[var(--foreground)]">{item.name}</div>
|
||||
<div className="mt-1 text-[11px] text-[var(--foreground-dim)]">
|
||||
{[item.parameterSize, item.quantization, item.sizeLabel].filter(Boolean).join(" · ") || item.family || "local model"}
|
||||
</div>
|
||||
</button>
|
||||
))}
|
||||
</div>
|
||||
)}
|
||||
</div>
|
||||
)}
|
||||
|
||||
<div className="flex items-center gap-3">
|
||||
<button onClick={saveEndpoint} disabled={saving} className="px-4 py-2 rounded-sm bg-[var(--nvidia-green)] text-white text-xs font-mono uppercase tracking-wider disabled:opacity-50">
|
||||
{saving ? "Saving..." : "Save Endpoint"}
|
||||
|
||||
Reference in New Issue
Block a user