Compare commits

..
2 Commits
8 changed files with 351 additions and 104 deletions
+1 -1
View File
@@ -4,7 +4,7 @@
## 当前功能
- **模型**:启动时自动扫描模型目录(可手动重新扫描),支持导入本地 GGUF、添加 OpenAI 兼容在线 API 模型(聊天时可直接选用
- **模型管理**:启动时自动扫描模型目录(可手动重新扫描),支持导入本地 GGUF、添加 OpenAI 兼容在线 API 模型;本地模型可一键「部署」后台加载,部署成功后对话页可直接选用
- **模型广场**:搜索 Hugging Face / ModelScope 上的 GGUF 模型,按量化版本一键下载(ModelScope 文件自动附带 SHA256 校验),下载进度实时显示
- **聊天**:流式输出、Markdown/代码高亮、采样参数调节、多会话管理;本地模型与在线 API 模型统一入口
- **本地 API 服务**OpenAI 兼容端点(/v1/models、/v1/chat/completions、/v1/embeddings),仅本机监听,可选 API Key
+125 -2
View File
@@ -62,6 +62,14 @@ pub struct ChatErrorEvent {
pub message: String,
}
#[derive(Serialize, Clone)]
pub struct EngineDeployEvent {
pub model_id: String,
pub file_name: String,
pub state: String,
pub message: Option<String>,
}
#[derive(Deserialize, Clone)]
pub struct DownloadPayload {
pub url: String,
@@ -196,6 +204,7 @@ pub fn get_messages(state: State<'_, App>, conversation_id: String) -> Result<Ve
#[tauri::command]
pub async fn engine_start(
app: AppHandle,
state: State<'_, App>,
model_id: String,
params: ChatParams,
@@ -237,13 +246,16 @@ pub async fn engine_start(
};
engine.start(cfg).await.map_err(|e| e.to_string())?;
*base.write().await = engine.base_url().await;
Ok(engine.status().await)
let status = engine.status().await;
let _ = app.emit("engine://status", status.clone());
Ok(status)
}
#[tauri::command]
pub async fn engine_stop(state: State<'_, App>) -> Result<(), String> {
pub async fn engine_stop(app: AppHandle, state: State<'_, App>) -> Result<(), String> {
state.engine.stop().await.map_err(|e| e.to_string())?;
*state.engine_base.write().await = None;
let _ = app.emit("engine://status", state.engine.status().await);
Ok(())
}
@@ -254,6 +266,117 @@ pub async fn engine_status(
Ok(state.engine.status().await)
}
/// 后台部署(加载)本地模型:启动 llama-server 并等待就绪,通过事件广播进度。
#[tauri::command]
pub async fn deploy_model(
app: AppHandle,
state: State<'_, App>,
model_id: String,
params: ChatParams,
) -> Result<(), String> {
let core = state.core.clone();
let engine = state.engine.clone();
let base = state.engine_base.clone();
let (model, engine_bin) = {
let db = core.db.lock().unwrap();
let model = models_db::get(&db, &model_id)
.map_err(|e| e.to_string())?
.ok_or_else(|| "model not found".to_string())?;
if model.kind != "local" {
return Err("在线 API 模型无需部署".into());
}
let bin = settings_db::get(&db, "engine_bin")
.map_err(|e| e.to_string())?
.unwrap_or_default();
(model, bin)
};
let file_name = model.file_name.clone();
let emit_deploy = |state_name: &str, message: Option<String>| {
let _ = app.emit(
"engine://deploy",
EngineDeployEvent {
model_id: model_id.clone(),
file_name: file_name.clone(),
state: state_name.into(),
message,
},
);
};
let current = engine.status().await;
if current.running {
if current.model.as_deref() == Some(file_name.as_str()) {
emit_deploy("ready", None);
return Ok(());
}
// 部署另一个模型前,先停掉当前引擎
let _ = engine.stop().await;
*base.write().await = None;
let _ = app.emit("engine://status", engine.status().await);
}
let bin = PathBuf::from(engine_bin);
if !bin.exists() {
emit_deploy(
"error",
Some(format!(
"引擎文件不存在:{},请到设置页配置 llama-server 路径",
bin.display()
)),
);
return Ok(());
}
emit_deploy("loading", None);
let app2 = app.clone();
let core2 = core.clone();
let engine2 = engine.clone();
let base2 = base.clone();
let model_id2 = model_id.clone();
let file_name2 = file_name.clone();
tokio::spawn(async move {
let cfg = EngineConfig {
binary_path: bin,
model_path: PathBuf::from(model.file_path),
host: "127.0.0.1".into(),
ctx_size: params.ctx_size,
ngl: params.ngl,
threads: None,
log_file: core2.logs_dir.join(format!("engine-{}.log", model.file_name)),
};
match engine2.start(cfg).await {
Ok(()) => {
*base2.write().await = engine2.base_url().await;
let _ = app2.emit(
"engine://deploy",
EngineDeployEvent {
model_id: model_id2,
file_name: file_name2,
state: "ready".into(),
message: None,
},
);
}
Err(e) => {
let _ = app2.emit(
"engine://deploy",
EngineDeployEvent {
model_id: model_id2,
file_name: file_name2,
state: "error".into(),
message: Some(e.to_string()),
},
);
}
}
let _ = app2.emit("engine://status", engine2.status().await);
});
Ok(())
}
#[tauri::command]
pub async fn chat_send(
app: AppHandle,
+1
View File
@@ -81,6 +81,7 @@ pub fn run() {
commands::engine_start,
commands::engine_stop,
commands::engine_status,
commands::deploy_model,
commands::chat_send,
commands::download_enqueue,
commands::server_start,
+16 -4
View File
@@ -1,6 +1,6 @@
import { useEffect } from "react";
import { NavLink, Route, Routes } from "react-router-dom";
import { onEvent } from "./api";
import { EngineDeployEvent, onEvent } from "./api";
import { useStore } from "./store";
import ModelsPage from "./pages/ModelsPage";
import ChatPage from "./pages/ChatPage";
@@ -9,8 +9,8 @@ import ServerPage from "./pages/ServerPage";
import SettingsPage from "./pages/SettingsPage";
const navItems = [
{ to: "/", label: "模型库", end: true },
{ to: "/chat", label: "聊天" },
{ to: "/chat", label: "对话" },
{ to: "/", label: "模型管理", end: true },
{ to: "/plaza", label: "模型广场" },
{ to: "/server", label: "本地服务" },
{ to: "/settings", label: "设置" },
@@ -21,6 +21,7 @@ export default function App() {
const refreshEngine = useStore((s) => s.refreshEngine);
const refreshServer = useStore((s) => s.refreshServer);
const refreshConversations = useStore((s) => s.refreshConversations);
const setDeployState = useStore((s) => s.setDeployState);
useEffect(() => {
refreshModels();
@@ -31,13 +32,24 @@ export default function App() {
const un2 = onEvent("server://status", () => refreshServer());
const un3 = onEvent("download://done", () => refreshModels());
const un4 = onEvent("models://updated", () => refreshModels());
const un5 = onEvent<EngineDeployEvent>("engine://deploy", (e) => {
setDeployState(e.model_id, e.state);
refreshEngine();
});
return () => {
un1.then((f) => f());
un2.then((f) => f());
un3.then((f) => f());
un4.then((f) => f());
un5.then((f) => f());
};
}, [refreshModels, refreshEngine, refreshServer, refreshConversations]);
}, [
refreshModels,
refreshEngine,
refreshServer,
refreshConversations,
setDeployState,
]);
return (
<div className="flex h-screen overflow-hidden">
+9
View File
@@ -108,6 +108,13 @@ export interface ChatErrorEvent {
message: string;
}
export interface EngineDeployEvent {
model_id: string;
file_name: string;
state: string;
message: string | null;
}
export interface DownloadProgressEvent {
id: string;
url: string;
@@ -148,6 +155,8 @@ export const api = {
invoke<Message[]>("get_messages", { conversationId }),
engineStart: (modelId: string, params: ChatParams) =>
invoke<EngineStatus>("engine_start", { modelId, params }),
deployModel: (modelId: string, params: ChatParams) =>
invoke<void>("deploy_model", { modelId, params }),
engineStop: () => invoke<void>("engine_stop"),
engineStatus: () => invoke<EngineStatus>("engine_status"),
chatSend: (payload: ChatSendPayload) => invoke<string>("chat_send", { payload }),
+98 -86
View File
@@ -168,6 +168,7 @@ export default function ChatPage() {
}
const engineRunning = useStore((s) => s.engine?.running ?? false);
const engineModel = useStore((s) => s.engine?.model ?? null);
const selectedIsRemote = useMemo(
() => models.find((m) => m.id === selectedModelId)?.kind === "remote",
[models, selectedModelId],
@@ -206,45 +207,6 @@ export default function ChatPage() {
</div>
))}
</div>
<div className="mt-3 border-t border-border pt-3">
<label className="mb-1 block text-xs text-slate-400"></label>
<select
className="input w-full"
value={selectedModelId}
onChange={(e) => setSelectedModelId(e.target.value)}
>
{models.map((m) => (
<option key={m.id} value={m.id}>
{m.file_name}
{m.kind === "remote" ? "API" : ""}
</option>
))}
</select>
{selectedIsRemote ? (
<div className="mt-2 rounded-lg border border-sky-500/30 bg-sky-500/10 px-3 py-1.5 text-xs text-sky-300">
线 API
</div>
) : (
<div className="mt-2 flex gap-2">
<button
className="flex-1 rounded-lg border border-border bg-panel-2 py-1.5 text-xs hover:bg-panel-2/60"
onClick={() =>
api.engineStart(selectedModelId, params).catch((e) => alert(String(e)))
}
disabled={engineRunning}
>
</button>
<button
className="flex-1 rounded-lg border border-border bg-panel-2 py-1.5 text-xs hover:bg-panel-2/60"
onClick={() => api.engineStop().catch((e) => alert(String(e)))}
disabled={!engineRunning}
>
</button>
</div>
)}
</div>
</div>
<div className="flex flex-1 flex-col">
@@ -289,49 +251,6 @@ export default function ChatPage() {
</div>
<div className="border-t border-border bg-panel p-4">
<div className="mb-2 flex flex-wrap items-center gap-x-4 gap-y-1 text-xs text-slate-400">
<ParamSlider
label="temperature"
value={params.temperature}
min={0}
max={2}
step={0.05}
onChange={(v) => setParams({ ...params, temperature: v })}
/>
<ParamSlider
label="top_p"
value={params.top_p}
min={0}
max={1}
step={0.05}
onChange={(v) => setParams({ ...params, top_p: v })}
/>
<ParamSlider
label="max_tokens"
value={params.max_tokens}
min={64}
max={8192}
step={64}
onChange={(v) => setParams({ ...params, max_tokens: v })}
/>
<ParamSlider
label="ctx_size"
value={params.ctx_size}
min={512}
max={32768}
step={512}
onChange={(v) => setParams({ ...params, ctx_size: v })}
/>
<ParamSlider
label="ngl(全部=99)"
value={params.ngl}
min={0}
max={99}
step={1}
onChange={(v) => setParams({ ...params, ngl: v })}
/>
<span className="ml-auto text-slate-500">{engineLabel}</span>
</div>
<div className="flex items-end gap-3">
<textarea
className="input min-h-[72px] flex-1 resize-none"
@@ -355,6 +274,97 @@ export default function ChatPage() {
</div>
</div>
</div>
{/* 右侧边栏:当前模型 + 部署/请求参数 */}
<div className="flex w-72 flex-col overflow-y-auto border-l border-border bg-panel p-3">
<div className="mb-1 text-xs text-slate-400"></div>
<select
className="input w-full"
value={selectedModelId}
onChange={(e) => setSelectedModelId(e.target.value)}
>
{models.map((m) => (
<option key={m.id} value={m.id}>
{m.file_name}
{m.kind === "remote" ? "API" : ""}
{engineRunning && engineModel === m.file_name ? " ●已部署" : ""}
</option>
))}
</select>
{selectedIsRemote ? (
<div className="mt-2 rounded-lg border border-sky-500/30 bg-sky-500/10 px-3 py-1.5 text-xs text-sky-300">
线 API
</div>
) : (
<div className="mt-2 flex gap-2">
<button
className="flex-1 rounded-lg border border-border bg-panel-2 py-1.5 text-xs hover:bg-panel-2/60"
onClick={() =>
api.engineStart(selectedModelId, params).catch((e) => alert(String(e)))
}
disabled={engineRunning}
>
</button>
<button
className="flex-1 rounded-lg border border-border bg-panel-2 py-1.5 text-xs hover:bg-panel-2/60"
onClick={() => api.engineStop().catch((e) => alert(String(e)))}
disabled={!engineRunning}
>
</button>
</div>
)}
<div className="mt-1 truncate text-xs text-slate-500">{engineLabel}</div>
<div className="mt-4 border-t border-border pt-3">
<div className="mb-3 text-xs uppercase text-slate-400"></div>
<ParamSlider
label="上下文长度 ctx_size"
value={params.ctx_size}
min={512}
max={32768}
step={512}
onChange={(v) => setParams({ ...params, ctx_size: v })}
/>
<ParamSlider
label="GPU 层数 ngl99=全部)"
value={params.ngl}
min={0}
max={99}
step={1}
onChange={(v) => setParams({ ...params, ngl: v })}
/>
</div>
<div className="mt-4 border-t border-border pt-3">
<div className="mb-3 text-xs uppercase text-slate-400"></div>
<ParamSlider
label="temperature"
value={params.temperature}
min={0}
max={2}
step={0.05}
onChange={(v) => setParams({ ...params, temperature: v })}
/>
<ParamSlider
label="top_p"
value={params.top_p}
min={0}
max={1}
step={0.05}
onChange={(v) => setParams({ ...params, top_p: v })}
/>
<ParamSlider
label="max_tokens"
value={params.max_tokens}
min={64}
max={8192}
step={64}
onChange={(v) => setParams({ ...params, max_tokens: v })}
/>
</div>
</div>
</div>
);
}
@@ -375,18 +385,20 @@ function ParamSlider({
onChange: (v: number) => void;
}) {
return (
<label className="flex items-center gap-2">
<span className="w-24">{label}</span>
<label className="mb-3 block">
<div className="mb-1 flex items-center justify-between text-xs text-slate-400">
<span>{label}</span>
<span className="font-mono text-slate-300">{value}</span>
</div>
<input
type="range"
className="w-28 accent-indigo-500"
className="w-full accent-indigo-500"
min={min}
max={max}
step={step}
value={value}
onChange={(e) => onChange(Number(e.target.value))}
/>
<span className="w-12 font-mono text-slate-300">{value}</span>
</label>
);
}
+96 -10
View File
@@ -11,6 +11,8 @@ function fmtSize(bytes: number) {
export default function ModelsPage() {
const models = useStore((s) => s.models);
const refreshModels = useStore((s) => s.refreshModels);
const engine = useStore((s) => s.engine);
const deployStates = useStore((s) => s.deployStates);
const [showRemoteForm, setShowRemoteForm] = useState(false);
const [scanning, setScanning] = useState(false);
const [remote, setRemote] = useState({
@@ -82,13 +84,36 @@ export default function ModelsPage() {
await api.openPath(info.models_dir);
}
async function handleDeploy(id: string, params?: Partial<import("../api").ChatParams>) {
try {
await api.deployModel(id, {
temperature: 0.7,
top_p: 0.9,
max_tokens: 2048,
ctx_size: 4096,
ngl: 99,
...params,
});
} catch (e) {
alert(String(e));
}
}
async function handleStopEngine() {
try {
await api.engineStop();
} catch (e) {
alert(String(e));
}
}
return (
<div className="h-full overflow-auto p-6">
<div className="mb-4 flex items-center justify-between">
<div>
<h1 className="text-xl font-semibold"></h1>
<h1 className="text-xl font-semibold"></h1>
<p className="mt-1 text-sm text-slate-400">
GGUF 线 API
</p>
</div>
<div className="flex gap-2">
@@ -208,19 +233,33 @@ export default function ModelsPage() {
{m.kind === "remote" ? "-" : fmtSize(m.file_size)}
</td>
<td className="px-4 py-3">
{m.status === "ready" ? (
{m.kind === "remote" ? (
<span className="text-slate-400">-</span>
) : m.status === "ready" ? (
<span className="text-emerald-400">{m.status}</span>
) : (
<span className="text-amber-400">{m.status}</span>
)}
</td>
<td className="px-4 py-3 text-right">
<button
className="text-xs text-slate-400 hover:text-red-400"
onClick={() => handleRemove(m.id, m.file_name)}
>
</button>
<td className="px-4 py-3">
<div className="flex items-center justify-end gap-2">
{m.kind === "local" ? (
<DeployButton
model={m}
engineRunning={engine?.running ?? false}
deployed={engine?.model === m.file_name}
deployState={deployStates[m.id]}
onDeploy={handleDeploy}
onStop={handleStopEngine}
/>
) : null}
<button
className="text-xs text-slate-400 hover:text-red-400"
onClick={() => handleRemove(m.id, m.file_name)}
>
</button>
</div>
</td>
</tr>
))}
@@ -232,3 +271,50 @@ export default function ModelsPage() {
);
}
function DeployButton({
model,
engineRunning,
deployed,
deployState,
onDeploy,
onStop,
}: {
model: { id: string; file_name: string };
engineRunning: boolean;
deployed: boolean;
deployState?: string;
onDeploy: (id: string) => void;
onStop: () => void;
}) {
if (deployed) {
return (
<div className="flex items-center gap-2">
<span className="rounded bg-emerald-500/20 px-2 py-1 text-xs text-emerald-300">
</span>
<button
className="text-xs text-slate-400 hover:text-red-400"
onClick={onStop}
disabled={!engineRunning}
>
</button>
</div>
);
}
if (deployState === "loading") {
return (
<button className="btn-secondary !px-3 !py-1 text-xs" disabled>
</button>
);
}
return (
<button
className="btn-secondary !px-3 !py-1 text-xs"
onClick={() => onDeploy(model.id)}
>
{deployState === "error" ? "重试部署" : "部署"}
</button>
);
}
+5 -1
View File
@@ -6,6 +6,8 @@ interface Store {
engine: EngineStatus | null;
server: ServerStatus | null;
conversations: Conversation[];
deployStates: Record<string, string>;
setDeployState: (modelId: string, state: string) => void;
refreshModels: () => Promise<void>;
refreshEngine: () => Promise<void>;
refreshServer: () => Promise<void>;
@@ -17,10 +19,12 @@ export const useStore = create<Store>((set) => ({
engine: null,
server: null,
conversations: [],
deployStates: {},
setDeployState: (modelId, state) =>
set((s) => ({ deployStates: { ...s.deployStates, [modelId]: state } })),
refreshModels: async () => set({ models: await api.listModels() }),
refreshEngine: async () => set({ engine: await api.engineStatus() }),
refreshServer: async () => set({ server: await api.serverStatus() }),
refreshConversations: async () =>
set({ conversations: await api.listConversations() }),
}));