[feat] add local code

src/lib/ChatScreens/DefaultChatScreen.svelte

@@ -404,7 +404,7 @@
 
             {#if $DataBase.useAutoSuggestions}
                 <Suggestion messageInput={(msg)=>messageInput=(
-                    ($DataBase.subModel === "textgen_webui" || $DataBase.subModel === "mancer") && $DataBase.autoSuggestClean
+                    ($DataBase.subModel === "textgen_webui" || $DataBase.subModel === "mancer" || $DataBase.subModel.startsWith('local_')) && $DataBase.autoSuggestClean
                     ? msg.replace(/ +\(.+?\) *$| - [^"'*]*?$/, '')
                     : msg
                 )} {send}/>

src/lib/ChatScreens/Suggestion.svelte

@@ -75,7 +75,7 @@
             }
             ]
 
-            if($DataBase.subModel === "textgen_webui" || $DataBase.subModel === 'mancer'){
+            if($DataBase.subModel === "textgen_webui" || $DataBase.subModel === 'mancer' || $DataBase.subModel.startsWith('local_')){
                 promptbody = [
                     {
                         role: 'system',

src/lib/Setting/Pages/BotSettings.svelte

@@ -265,7 +265,7 @@
 {/if}
 <span class="text-textcolor2 mb-6 text-sm">{($DataBase.temperature / 100).toFixed(2)}</span>
 
-{#if $DataBase.aiModel === 'textgen_webui' || $DataBase.subModel === 'mancer'}
+{#if $DataBase.aiModel === 'textgen_webui' || $DataBase.subModel === 'mancer' || $DataBase.subModel.startsWith('local_')}
     <span class="text-textcolor">Repetition Penalty</span>
     <SliderInput min={1} max={1.5} step={0.01} bind:value={$DataBase.ooba.repetition_penalty}/>
     <span class="text-textcolor2 mb-6 text-sm">{($DataBase.ooba.repetition_penalty).toFixed(2)}</span>

src/lib/UI/ModelList.svelte

@@ -4,6 +4,8 @@
     import Arcodion from "./Arcodion.svelte";
     import { language } from "src/lang";
     import { isNodeServer, isTauri } from "src/ts/storage/globalApi";
+  import { checkLocalModel } from "src/ts/process/models/local";
+  import { alertError } from "src/ts/alert";
     let openAdv = true
 
     export let value = ""
@@ -32,6 +34,8 @@
                 return "GPT-4 0613"
             case "gpt4_32k_0613":
                 return "GPT-4 32k 0613"
+            case 'local_gptq':
+                return 'Local Model GPTQ'
             case "palm2":
                 return "PaLM2"
             case "textgen_webui":
@@ -111,10 +115,20 @@
                 {/if}
             </Arcodion>
             <button class="hover:bg-selected px-6 py-2 text-lg" on:click={() => {changeModel('reverse_proxy')}}>Reverse Proxy</button>
-            <button class="hover:bg-selected px-6 py-2 text-lg" on:click={() => {changeModel('openrouter')}}>OpenRouter</button>
+            <!-- <button class="hover:bg-selected px-6 py-2 text-lg" on:click={async () => {
+                const res = (await checkLocalModel())
+                if(res === 'success'){
+                    changeModel('local_gptq')
+                }
+                else{
+                    alertError("python 3.10, cuda 11.7 and git must be installed to run it. " + res)
+                }
+                // changeModel('local_gptq')
+            }}>Local Model GPTQ</button> -->
             <button class="hover:bg-selected px-6 py-2 text-lg" on:click={() => {changeModel('textgen_webui')}}>Oobabooga WebUI</button>
             <button class="hover:bg-selected px-6 py-2 text-lg" on:click={() => {changeModel('mancer')}}>Mancer</button>
             <button class="hover:bg-selected px-6 py-2 text-lg" on:click={() => {changeModel('palm2')}}>Google PaLM2</button>
+            <button class="hover:bg-selected px-6 py-2 text-lg" on:click={() => {changeModel('openrouter')}}>OpenRouter</button>
             <button class="hover:bg-selected px-6 py-2 text-lg" on:click={() => {changeModel('kobold')}}>Kobold</button>
             <Arcodion name="Novellist">
                 <button class="p-2 hover:text-green-500" on:click={() => {changeModel('novellist')}}>SuperTrin</button>

src/ts/alert.ts

@@ -87,6 +87,13 @@ export function alertWait(msg:string){
 
 }
 
+export function alertClear(){
+    alertStore.set({
+        'type': 'none',
+        'msg': ''
+    })
+}
+
 export async function alertSelectChar(){
     alertStore.set({
         'type': 'selectChar',

src/ts/process/models/local.ts

@@ -0,0 +1,134 @@
+import { invoke } from "@tauri-apps/api/tauri";
+import { globalFetch } from "src/ts/storage/globalApi";
+import { sleep } from "src/ts/util";
+import path from "@tauri-apps/api/path";
+import { exists } from "@tauri-apps/api/fs";
+import { alertClear, alertError, alertMd, alertWait } from "src/ts/alert";
+import { get } from "svelte/store";
+import { DataBase } from "src/ts/storage/database";
+let serverRunning = false;
+
+export function checkLocalModel():Promise<string>{
+    return invoke("check_requirements_local")
+}
+
+export async function startLocalModelServer(){
+    if(!serverRunning){
+        serverRunning = true
+        await invoke("run_server_local")
+    }
+    return
+}
+
+export async function checkLocalServerInstalled() {
+    const p = await path.join(await path.appDataDir(), 'local_server')
+    return await exists(p)
+}
+
+export interface LocalLoaderItem {
+    dir: string;
+    max_seq_len?: number ;
+    max_input_len?: number ;
+    max_attention_size?: number ;
+    compress_pos_emb?: number ;
+    alpha_value?: number ;
+    gpu_peer_fixed?: boolean ;
+    auto_map?: boolean ;
+    use_flash_attn_2?: boolean ;
+    matmul_recons_thd?: number ;
+    fused_mlp_thd?: number ;
+    sdp_thd?: number ;
+    fused_attn?: boolean ;
+    matmul_fused_remap?: boolean ;
+    rmsnorm_no_half2?: boolean ;
+    rope_no_half2?: boolean ;
+    matmul_no_half2?: boolean ;
+    silu_no_half2?: boolean ;
+    concurrent_streams?: boolean ;
+}
+// class GeneratorItem(BaseModel):
+//     temperature: Union[float, None]
+//     top_k: Union[int, None]
+//     top_p: Union[float, None]
+//     min_p: Union[float, None]
+//     typical: Union[float, None]
+//     token_repetition_penalty_max: Union[float, None]
+//     token_repetition_penalty_sustain: Union[int, None]
+//     token_repetition_penalty_decay: Union[int, None]
+//     beams: Union[int, None]
+//     beam_length: Union[int, None]
+//     disallowed_tokens: Union[list[int], None]
+//     prompt: str
+//     max_new_tokens: Union[int, None]
+interface LocalGeneratorItem {
+    temperature?: number;
+    top_k?: number;
+    top_p?: number;
+    min_p?: number;
+    typical?: number;
+    token_repetition_penalty_max?: number;
+    token_repetition_penalty_sustain?: number;
+    token_repetition_penalty_decay?: number;
+    beams?: number;
+    beam_length?: number;
+    disallowed_tokens?: number[];
+    prompt: string;
+    max_new_tokens?: number;
+}
+
+
+export async function loadExllamaFull(){
+
+    try {
+        await startLocalModelServer()
+        if(await checkLocalServerInstalled()){
+            alertWait("Loading exllama")
+        }
+        else{
+            alertWait("Installing & Loading exllama, this would take a while for the first time")
+        }
+        while(true){
+            //check is server is running by fetching the status
+            try {
+                const res = await globalFetch("http://localhost:7239/")
+                if(res.ok){
+                    break
+                }
+            } catch (error) {}
+            await sleep(1000)
+        }
+
+        const body:LocalLoaderItem = {
+            dir: "exllama",
+        }
+    
+        alertWait("Loading Local Model")
+        const res = await globalFetch("http://localhost:7239/load/", {
+            body: body
+        })
+        alertClear()
+    } catch (error) {
+        alertError("Error when loading Exllama: " + error)     
+    }
+
+}
+
+
+export async function runLocalModel(prompt:string){
+    const db = get(DataBase)
+
+    const body:LocalGeneratorItem = {
+        prompt: prompt,
+        temperature: db.temperature,
+        top_k: db.ooba.top_k,
+        top_p: db.ooba.top_p,
+        typical: db.ooba.typical_p,
+        max_new_tokens: db.maxResponse
+    }
+
+    const gen = await globalFetch("http://localhost:7239/generate/", {
+        body: body
+    })
+
+    console.log(gen)
+}