feat: litellm proxy frontend integration

This commit is contained in:
Timothy J. Baek 2024-02-24 17:56:53 -08:00
parent a453be07e1
commit ec6f53e298
6 changed files with 823 additions and 360 deletions

View file

@ -109,6 +109,7 @@ def get_models(user=Depends(get_current_user)):
models = r.json()
return models
except Exception as e:
app.state.ENABLED = False
raise HTTPException(status_code=400, detail=ERROR_MESSAGES.DEFAULT(e))
@ -120,6 +121,7 @@ async def get_default_model(user=Depends(get_admin_user)):
return {"model": options["sd_model_checkpoint"]}
except Exception as e:
app.state.ENABLED = False
raise HTTPException(status_code=400, detail=ERROR_MESSAGES.DEFAULT(e))

View file

@ -1,6 +1,6 @@
{
"name": "open-webui",
"version": "0.1.102",
"version": "0.1.103",
"private": true,
"scripts": {
"dev": "vite dev --host",

View file

@ -17,7 +17,7 @@ export const getLiteLLMModels = async (token: string = '') => {
})
.catch((err) => {
console.log(err);
error = `OpenAI: ${err?.error?.message ?? 'Network Problem'}`;
error = `LiteLLM: ${err?.error?.message ?? 'Network Problem'}`;
return [];
});
@ -40,3 +40,109 @@ export const getLiteLLMModels = async (token: string = '') => {
})
: models;
};
export const getLiteLLMModelInfo = async (token: string = '') => {
let error = null;
const res = await fetch(`${LITELLM_API_BASE_URL}/model/info`, {
method: 'GET',
headers: {
Accept: 'application/json',
'Content-Type': 'application/json',
...(token && { authorization: `Bearer ${token}` })
}
})
.then(async (res) => {
if (!res.ok) throw await res.json();
return res.json();
})
.catch((err) => {
console.log(err);
error = `LiteLLM: ${err?.error?.message ?? 'Network Problem'}`;
return [];
});
if (error) {
throw error;
}
const models = Array.isArray(res) ? res : res?.data ?? null;
return models;
};
type AddLiteLLMModelForm = {
name: string;
model: string;
api_base: string;
api_key: string;
rpm: string;
};
export const addLiteLLMModel = async (token: string = '', payload: AddLiteLLMModelForm) => {
let error = null;
const res = await fetch(`${LITELLM_API_BASE_URL}/model/new`, {
method: 'POST',
headers: {
Accept: 'application/json',
'Content-Type': 'application/json',
...(token && { authorization: `Bearer ${token}` })
},
body: JSON.stringify({
model_name: payload.name,
litellm_params: {
model: payload.model,
...(payload.api_base === '' ? {} : { api_base: payload.api_base }),
...(payload.api_key === '' ? {} : { api_key: payload.api_key }),
...(isNaN(parseInt(payload.rpm)) ? {} : { rpm: parseInt(payload.rpm) })
}
})
})
.then(async (res) => {
if (!res.ok) throw await res.json();
return res.json();
})
.catch((err) => {
console.log(err);
error = `LiteLLM: ${err?.error?.message ?? 'Network Problem'}`;
return [];
});
if (error) {
throw error;
}
return res;
};
export const deleteLiteLLMModel = async (token: string = '', id: string) => {
let error = null;
const res = await fetch(`${LITELLM_API_BASE_URL}/model/delete`, {
method: 'POST',
headers: {
Accept: 'application/json',
'Content-Type': 'application/json',
...(token && { authorization: `Bearer ${token}` })
},
body: JSON.stringify({
id: id
})
})
.then(async (res) => {
if (!res.ok) throw await res.json();
return res.json();
})
.catch((err) => {
console.log(err);
error = `LiteLLM: ${err?.error?.message ?? 'Network Problem'}`;
return [];
});
if (error) {
throw error;
}
return res;
};

View file

@ -38,6 +38,7 @@
</div>
</div>
{#if ollamaVersion}
<hr class=" dark:border-gray-700" />
<div>
@ -48,6 +49,7 @@
</div>
</div>
</div>
{/if}
<hr class=" dark:border-gray-700" />

View file

@ -15,6 +15,9 @@
let OPENAI_API_KEY = '';
let OPENAI_API_BASE_URL = '';
let showOpenAI = false;
let showLiteLLM = false;
const updateOpenAIHandler = async () => {
OPENAI_API_BASE_URL = await updateOpenAIUrl(localStorage.token, OPENAI_API_BASE_URL);
OPENAI_API_KEY = await updateOpenAIKey(localStorage.token, OPENAI_API_KEY);
@ -45,7 +48,7 @@
</script>
<form
class="flex flex-col h-full space-y-3 text-sm"
class="flex flex-col h-full justify-between text-sm"
on:submit|preventDefault={() => {
updateOpenAIHandler();
dispatch('save');
@ -56,6 +59,59 @@
// });
}}
>
<div class=" pr-1.5 overflow-y-scroll max-h-[21rem] space-y-3">
<div class=" space-y-3">
<div class="mt-2 space-y-2 pr-1.5">
<div class="flex justify-between items-center text-sm">
<div class=" font-medium">OpenAI API</div>
<button
class=" text-xs font-medium text-gray-500"
type="button"
on:click={() => {
showOpenAI = !showOpenAI;
}}>{showOpenAI ? 'Hide' : 'Show'}</button
>
</div>
{#if showOpenAI}
<div>
<div class=" mb-2.5 text-sm font-medium">API Key</div>
<div class="flex w-full">
<div class="flex-1">
<input
class="w-full rounded py-2 px-4 text-sm dark:text-gray-300 dark:bg-gray-800 outline-none"
placeholder="Enter OpenAI API Key"
bind:value={OPENAI_API_KEY}
autocomplete="off"
/>
</div>
</div>
</div>
<div>
<div class=" mb-2.5 text-sm font-medium">API Base URL</div>
<div class="flex w-full">
<div class="flex-1">
<input
class="w-full rounded py-2 px-4 text-sm dark:text-gray-300 dark:bg-gray-800 outline-none"
placeholder="Enter OpenAI API Base URL"
bind:value={OPENAI_API_BASE_URL}
autocomplete="off"
/>
</div>
</div>
<div class="mt-2 text-xs text-gray-400 dark:text-gray-500">
WebUI will make requests to <span class=" text-gray-200"
>'{OPENAI_API_BASE_URL}/chat'</span
>
</div>
</div>
{/if}
</div>
</div>
<hr class=" dark:border-gray-700" />
<div>
<div class=" mb-2.5 text-sm font-medium">Ollama API URL</div>
<div class="flex w-full">
@ -99,40 +155,6 @@
</a>
</div>
</div>
<hr class=" dark:border-gray-700" />
<div class=" space-y-3">
<div>
<div class=" mb-2.5 text-sm font-medium">OpenAI API Key</div>
<div class="flex w-full">
<div class="flex-1">
<input
class="w-full rounded py-2 px-4 text-sm dark:text-gray-300 dark:bg-gray-800 outline-none"
placeholder="Enter OpenAI API Key"
bind:value={OPENAI_API_KEY}
autocomplete="off"
/>
</div>
</div>
</div>
<div>
<div class=" mb-2.5 text-sm font-medium">OpenAI API Base URL</div>
<div class="flex w-full">
<div class="flex-1">
<input
class="w-full rounded py-2 px-4 text-sm dark:text-gray-300 dark:bg-gray-800 outline-none"
placeholder="Enter OpenAI API Base URL"
bind:value={OPENAI_API_BASE_URL}
autocomplete="off"
/>
</div>
</div>
<div class="mt-2 text-xs text-gray-400 dark:text-gray-500">
WebUI will make requests to <span class=" text-gray-200">'{OPENAI_API_BASE_URL}/chat'</span>
</div>
</div>
</div>
<div class="flex justify-end pt-3 text-sm font-medium">

View file

@ -2,14 +2,33 @@
import queue from 'async/queue';
import toast from 'svelte-french-toast';
import { createModel, deleteModel, pullModel } from '$lib/apis/ollama';
import { createModel, deleteModel, getOllamaVersion, pullModel } from '$lib/apis/ollama';
import { WEBUI_API_BASE_URL, WEBUI_BASE_URL } from '$lib/constants';
import { WEBUI_NAME, models, user } from '$lib/stores';
import { splitStream } from '$lib/utils';
import { onMount } from 'svelte';
import { addLiteLLMModel, deleteLiteLLMModel, getLiteLLMModelInfo } from '$lib/apis/litellm';
export let getModels: Function;
let showLiteLLM = false;
let showLiteLLMParams = false;
let liteLLMModelInfo = [];
let liteLLMModel = '';
let liteLLMModelName = '';
let liteLLMAPIBase = '';
let liteLLMAPIKey = '';
let liteLLMRPM = '';
let deleteLiteLLMModelId = '';
$: liteLLMModelName = liteLLMModel;
// Models
let showExperimentalOllama = false;
let ollamaVersion = '';
const MAX_PARALLEL_DOWNLOADS = 3;
const modelDownloadQueue = queue(
(task: { modelName: string }, cb) =>
@ -286,22 +305,82 @@
opts.callback({ success: true, modelName: opts.modelName });
}
};
const addLiteLLMModelHandler = async () => {
if (!liteLLMModelInfo.find((info) => info.model_name === liteLLMModelName)) {
const res = await addLiteLLMModel(localStorage.token, {
name: liteLLMModelName,
model: liteLLMModel,
api_base: liteLLMAPIBase,
api_key: liteLLMAPIKey,
rpm: liteLLMRPM
}).catch((error) => {
toast.error(error);
return null;
});
if (res) {
if (res.message) {
toast.success(res.message);
}
}
} else {
toast.error(`Model ${liteLLMModelName} already exists.`);
}
liteLLMModelName = '';
liteLLMModel = '';
liteLLMAPIBase = '';
liteLLMAPIKey = '';
liteLLMRPM = '';
liteLLMModelInfo = await getLiteLLMModelInfo(localStorage.token);
models.set(await getModels());
};
const deleteLiteLLMModelHandler = async () => {
const res = await deleteLiteLLMModel(localStorage.token, deleteLiteLLMModelId).catch(
(error) => {
toast.error(error);
return null;
}
);
if (res) {
if (res.message) {
toast.success(res.message);
}
}
deleteLiteLLMModelId = '';
liteLLMModelInfo = await getLiteLLMModelInfo(localStorage.token);
models.set(await getModels());
};
onMount(async () => {
ollamaVersion = await getOllamaVersion(localStorage.token).catch((error) => false);
liteLLMModelInfo = await getLiteLLMModelInfo(localStorage.token);
});
</script>
<div class="flex flex-col h-full justify-between text-sm">
<div class=" space-y-3 pr-1.5 overflow-y-scroll h-80">
<div class=" space-y-3 pr-1.5 overflow-y-scroll h-[23rem]">
{#if ollamaVersion}
<div class="space-y-2 pr-1.5">
<div>
<div class=" mb-2.5 text-sm font-medium">Pull a model from Ollama.com</div>
<div class=" mb-2 text-sm font-medium">Manage Ollama Models</div>
<div class=" mb-2 text-sm font-medium">Pull a model from Ollama.com</div>
<div class="flex w-full">
<div class="flex-1 mr-2">
<input
class="w-full rounded py-2 px-4 text-sm dark:text-gray-300 dark:bg-gray-800 outline-none"
class="w-full rounded py-2 px-4 text-sm dark:text-gray-300 dark:bg-gray-850 outline-none"
placeholder="Enter model tag (e.g. mistral:7b)"
bind:value={modelTag}
/>
</div>
<button
class="px-3 text-gray-100 bg-emerald-600 hover:bg-emerald-700 disabled:bg-gray-700 disabled:cursor-not-allowed rounded transition"
class="px-3 bg-gray-100 hover:bg-gray-200 text-gray-800 dark:bg-gray-850 dark:hover:bg-gray-800 dark:text-gray-100 rounded transition"
on:click={() => {
pullModelHandler();
}}
@ -379,14 +458,12 @@
{/if}
</div>
<hr class=" dark:border-gray-700" />
<div>
<div class=" mb-2.5 text-sm font-medium">Delete a model</div>
<div class=" mb-2 text-sm font-medium">Delete a model</div>
<div class="flex w-full">
<div class="flex-1 mr-2">
<select
class="w-full rounded py-2 px-4 text-sm dark:text-gray-300 dark:bg-gray-800 outline-none"
class="w-full rounded py-2 px-4 text-sm dark:text-gray-300 dark:bg-gray-850 outline-none"
bind:value={deleteModelTag}
placeholder="Select a model"
>
@ -401,7 +478,7 @@
</select>
</div>
<button
class="px-3 bg-red-700 hover:bg-red-800 text-gray-100 rounded transition"
class="px-3 bg-gray-100 hover:bg-gray-200 text-gray-800 dark:bg-gray-850 dark:hover:bg-gray-800 dark:text-gray-100 rounded transition"
on:click={() => {
deleteModelHandler();
}}
@ -422,21 +499,27 @@
</div>
</div>
<hr class=" dark:border-gray-700" />
<div>
<div class="flex justify-between items-center text-xs">
<div class=" text-sm font-medium">Experimental</div>
<button
class=" text-xs font-medium text-gray-500"
type="button"
on:click={() => {
showExperimentalOllama = !showExperimentalOllama;
}}>{showExperimentalOllama ? 'Show' : 'Hide'}</button
>
</div>
</div>
{#if showExperimentalOllama}
<form
on:submit|preventDefault={() => {
uploadModelHandler();
}}
>
<div class=" mb-2 flex w-full justify-between">
<div class=" text-sm font-medium">
Upload a GGUF model <a
class=" text-xs font-medium text-gray-500 underline"
href="https://github.com/jmorganca/ollama/blob/main/README.md#import-from-gguf"
target="_blank">(Experimental)</a
>
</div>
<div class=" text-sm font-medium">Upload a GGUF model</div>
<button
class="p-1 px-3 text-xs flex rounded transition"
@ -475,7 +558,7 @@
<button
type="button"
class="w-full rounded text-left py-2 px-4 dark:text-gray-300 dark:bg-gray-800"
class="w-full rounded text-left py-2 px-4 dark:text-gray-300 dark:bg-gray-850"
on:click={() => {
document.getElementById('model-upload-input').click();
}}
@ -490,7 +573,7 @@
{:else}
<div class="flex-1 {modelFileUrl !== '' ? 'mr-2' : ''}">
<input
class="w-full rounded text-left py-2 px-4 dark:text-gray-300 dark:bg-gray-800 outline-none {modelFileUrl !==
class="w-full rounded text-left py-2 px-4 dark:text-gray-300 dark:bg-gray-850 outline-none {modelFileUrl !==
''
? 'mr-2'
: ''}"
@ -592,5 +675,253 @@
</div>
{/if}
</form>
{/if}
</div>
<hr class=" dark:border-gray-700 my-2" />
{/if}
<div class=" space-y-3">
<div class="mt-2 space-y-3 pr-1.5">
<div>
<div class=" mb-2 text-sm font-medium">Manage LiteLLM Models</div>
<div>
<div class="flex justify-between items-center text-xs">
<div class=" text-sm font-medium">Add a model</div>
<button
class=" text-xs font-medium text-gray-500"
type="button"
on:click={() => {
showLiteLLMParams = !showLiteLLMParams;
}}>{showLiteLLMParams ? 'Advanced' : 'Default'}</button
>
</div>
</div>
<div class="my-2 space-y-2">
<div class="flex w-full mb-1.5">
<div class="flex-1 mr-2">
<input
class="w-full rounded py-2 px-4 text-sm dark:text-gray-300 dark:bg-gray-850 outline-none"
placeholder="Enter LiteLLM Model (litellm_params.model)"
bind:value={liteLLMModel}
autocomplete="off"
/>
</div>
<button
class="px-3 bg-gray-100 hover:bg-gray-200 text-gray-800 dark:bg-gray-850 dark:hover:bg-gray-800 dark:text-gray-100 rounded transition"
on:click={() => {
addLiteLLMModelHandler();
}}
>
<svg
xmlns="http://www.w3.org/2000/svg"
viewBox="0 0 16 16"
fill="currentColor"
class="w-4 h-4"
>
<path
d="M8.75 3.75a.75.75 0 0 0-1.5 0v3.5h-3.5a.75.75 0 0 0 0 1.5h3.5v3.5a.75.75 0 0 0 1.5 0v-3.5h3.5a.75.75 0 0 0 0-1.5h-3.5v-3.5Z"
/>
</svg>
</button>
</div>
{#if showLiteLLMParams}
<div>
<div class=" mb-1.5 text-sm font-medium">Model Name</div>
<div class="flex w-full">
<div class="flex-1">
<input
class="w-full rounded py-2 px-4 text-sm dark:text-gray-300 dark:bg-gray-850 outline-none"
placeholder="Enter Model Name (model_name)"
bind:value={liteLLMModelName}
autocomplete="off"
/>
</div>
</div>
</div>
<div>
<div class=" mb-1.5 text-sm font-medium">API Base URL</div>
<div class="flex w-full">
<div class="flex-1">
<input
class="w-full rounded py-2 px-4 text-sm dark:text-gray-300 dark:bg-gray-850 outline-none"
placeholder="Enter LiteLLM API Base URL (litellm_params.api_base)"
bind:value={liteLLMAPIBase}
autocomplete="off"
/>
</div>
</div>
</div>
<div>
<div class=" mb-1.5 text-sm font-medium">API Key</div>
<div class="flex w-full">
<div class="flex-1">
<input
class="w-full rounded py-2 px-4 text-sm dark:text-gray-300 dark:bg-gray-850 outline-none"
placeholder="Enter LiteLLM API Key (litellm_params.api_key)"
bind:value={liteLLMAPIKey}
autocomplete="off"
/>
</div>
</div>
</div>
<div>
<div class="mb-1.5 text-sm font-medium">API RPM</div>
<div class="flex w-full">
<div class="flex-1">
<input
class="w-full rounded py-2 px-4 text-sm dark:text-gray-300 dark:bg-gray-850 outline-none"
placeholder="Enter LiteLLM API RPM (litellm_params.rpm)"
bind:value={liteLLMRPM}
autocomplete="off"
/>
</div>
</div>
</div>
{/if}
</div>
<div class="mb-2 text-xs text-gray-400 dark:text-gray-500">
Not sure what to add?
<a
class=" text-gray-300 font-medium"
href="https://litellm.vercel.app/docs/proxy/configs#quick-start"
target="_blank"
>
Click here for help.
</a>
</div>
<div>
<div class=" mb-2.5 text-sm font-medium">Delete a model</div>
<div class="flex w-full">
<div class="flex-1 mr-2">
<select
class="w-full rounded py-2 px-4 text-sm dark:text-gray-300 dark:bg-gray-850 outline-none"
bind:value={deleteLiteLLMModelId}
placeholder="Select a model"
>
{#if !deleteLiteLLMModelId}
<option value="" disabled selected>Select a model</option>
{/if}
{#each liteLLMModelInfo as model}
<option value={model.model_info.id} class="bg-gray-100 dark:bg-gray-700"
>{model.model_name}</option
>
{/each}
</select>
</div>
<button
class="px-3 bg-gray-100 hover:bg-gray-200 text-gray-800 dark:bg-gray-850 dark:hover:bg-gray-800 dark:text-gray-100 rounded transition"
on:click={() => {
deleteLiteLLMModelHandler();
}}
>
<svg
xmlns="http://www.w3.org/2000/svg"
viewBox="0 0 16 16"
fill="currentColor"
class="w-4 h-4"
>
<path
fill-rule="evenodd"
d="M5 3.25V4H2.75a.75.75 0 0 0 0 1.5h.3l.815 8.15A1.5 1.5 0 0 0 5.357 15h5.285a1.5 1.5 0 0 0 1.493-1.35l.815-8.15h.3a.75.75 0 0 0 0-1.5H11v-.75A2.25 2.25 0 0 0 8.75 1h-1.5A2.25 2.25 0 0 0 5 3.25Zm2.25-.75a.75.75 0 0 0-.75.75V4h3v-.75a.75.75 0 0 0-.75-.75h-1.5ZM6.05 6a.75.75 0 0 1 .787.713l.275 5.5a.75.75 0 0 1-1.498.075l-.275-5.5A.75.75 0 0 1 6.05 6Zm3.9 0a.75.75 0 0 1 .712.787l-.275 5.5a.75.75 0 0 1-1.498-.075l.275-5.5a.75.75 0 0 1 .786-.711Z"
clip-rule="evenodd"
/>
</svg>
</button>
</div>
</div>
</div>
</div>
<!-- <div class="mt-2 space-y-3 pr-1.5">
<div>
<div class=" mb-2.5 text-sm font-medium">Add LiteLLM Model</div>
<div class="flex w-full mb-2">
<div class="flex-1">
<input
class="w-full rounded py-2 px-4 text-sm dark:text-gray-300 dark:bg-gray-800 outline-none"
placeholder="Enter LiteLLM Model (e.g. ollama/mistral)"
bind:value={liteLLMModel}
autocomplete="off"
/>
</div>
</div>
<div class="flex justify-between items-center text-sm">
<div class=" font-medium">Advanced Model Params</div>
<button
class=" text-xs font-medium text-gray-500"
type="button"
on:click={() => {
showLiteLLMParams = !showLiteLLMParams;
}}>{showLiteLLMParams ? 'Hide' : 'Show'}</button
>
</div>
{#if showLiteLLMParams}
<div>
<div class=" mb-2.5 text-sm font-medium">LiteLLM API Key</div>
<div class="flex w-full">
<div class="flex-1">
<input
class="w-full rounded py-2 px-4 text-sm dark:text-gray-300 dark:bg-gray-800 outline-none"
placeholder="Enter LiteLLM API Key (e.g. os.environ/AZURE_API_KEY_CA)"
bind:value={liteLLMAPIKey}
autocomplete="off"
/>
</div>
</div>
</div>
<div>
<div class=" mb-2.5 text-sm font-medium">LiteLLM API Base URL</div>
<div class="flex w-full">
<div class="flex-1">
<input
class="w-full rounded py-2 px-4 text-sm dark:text-gray-300 dark:bg-gray-800 outline-none"
placeholder="Enter LiteLLM API Base URL"
bind:value={liteLLMAPIBase}
autocomplete="off"
/>
</div>
</div>
</div>
<div>
<div class=" mb-2.5 text-sm font-medium">LiteLLM API RPM</div>
<div class="flex w-full">
<div class="flex-1">
<input
class="w-full rounded py-2 px-4 text-sm dark:text-gray-300 dark:bg-gray-800 outline-none"
placeholder="Enter LiteLLM API RPM"
bind:value={liteLLMRPM}
autocomplete="off"
/>
</div>
</div>
</div>
{/if}
<div class="mt-2 text-xs text-gray-400 dark:text-gray-500">
Not sure what to add?
<a
class=" text-gray-300 font-medium"
href="https://litellm.vercel.app/docs/proxy/configs#quick-start"
target="_blank"
>
Click here for help.
</a>
</div>
</div>
</div> -->
</div>
</div>
</div>