Skip to content

OpenRAG is a comprehensive, single package Retrieval-Augmented Generation platform built on Langflow, Docling, and Opensearch.

Python

refactor: update design for vlm advanced settings (#2159)

Mike Fortmanaa473078bf4f
diff --git a/frontend/app/settings/_components/ingest-settings-section.tsx b/frontend/app/settings/_components/ingest-settings-section.tsx
index 97ec84fe8..32a91b404 100644
--- a/frontend/app/settings/_components/ingest-settings-section.tsx
+++ b/frontend/app/settings/_components/ingest-settings-section.tsx
@@ -1,6 +1,6 @@
"use client";
-import { ArrowUpRight, Loader2, Minus, Plus } from "lucide-react";
+import { ArrowUpRight, ChevronDown, Loader2, Minus, Plus } from "lucide-react";
import { useCallback, useEffect, useMemo, useRef, useState } from "react";
import { toast } from "sonner";
import {
@@ -13,12 +13,6 @@ import { useGetSettingsQuery } from "@/app/api/queries/useGetSettingsQuery";
import { ConfirmationDialog } from "@/components/confirmation-dialog";
import { LabelWrapper } from "@/components/label-wrapper";
import { RequirePermission } from "@/components/require-permission";
-import {
- Accordion,
- AccordionContent,
- AccordionItem,
- AccordionTrigger,
-} from "@/components/ui/accordion";
import { Button } from "@/components/ui/button";
import {
Card,
@@ -27,6 +21,11 @@ import {
CardHeader,
CardTitle,
} from "@/components/ui/card";
+import {
+ Collapsible,
+ CollapsibleContent,
+ CollapsibleTrigger,
+} from "@/components/ui/collapsible";
import { Input } from "@/components/ui/input";
import { NumberInput } from "@/components/ui/inputs/number-input";
import { Label } from "@/components/ui/label";
@@ -338,7 +337,7 @@ export function IngestSettingsSection() {
setVlmWatsonxApiVersion(k.vlm_watsonx_api_version);
}, [settings.knowledge]);
- const [vlmAccordionValue, setVlmAccordionValue] = useState<string>("");
+ const [vlmOpen, setVlmOpen] = useState(false);
const autoSelectedVlm = useRef(false);
useEffect(() => {
@@ -825,161 +824,164 @@ export function IngestSettingsSection() {
/>
</div>
{showVlmSettings && (
- <div
- className={cn(
- "mt-4 border border-border rounded-lg bg-muted/5 overflow-hidden transition-all duration-200",
- !pictureDescriptions &&
- "opacity-50 cursor-not-allowed select-none",
- )}
- >
- <Accordion
- type="single"
- collapsible
- disabled={!pictureDescriptions}
- value={pictureDescriptions ? vlmAccordionValue : ""}
- onValueChange={setVlmAccordionValue}
+ <>
+ <hr className="mt-4 border-border" />
+ <Collapsible
+ open={vlmOpen}
+ onOpenChange={setVlmOpen}
+ className={cn(
+ "mt-4 px-4 transition-all duration-200",
+ !pictureDescriptions && "opacity-50",
+ )}
>
- <AccordionItem value="vlm-settings" className="border-none">
- <AccordionTrigger
+ <CollapsibleTrigger className="flex w-full items-center justify-between py-2 text-sm font-medium text-foreground hover:text-foreground/80">
+ Advanced Vision Model (VLM) Settings
+ <ChevronDown
className={cn(
- "hover:no-underline font-medium text-foreground px-4 py-3 bg-muted/10 border-b border-border",
- !pictureDescriptions && "pointer-events-none",
+ "h-4 w-4 text-muted-foreground transition-transform duration-200",
+ vlmOpen && "rotate-180",
)}
- >
- Advanced Vision Model (VLM) Settings
- </AccordionTrigger>
- <AccordionContent className="p-4 space-y-6">
+ />
+ </CollapsibleTrigger>
+ <CollapsibleContent className="pt-4 space-y-6">
+ <div className="space-y-2">
+ <LabelWrapper
+ id="vlm-model"
+ label="Vision model"
+ helperText="Pick a vision-capable model; the provider is set from your selection"
+ required={pictureDescriptions}
+ >
+ <ModelSelector
+ groupedOptions={groupedVlmModels}
+ noOptionsPlaceholder={
+ isLoadingAnyVlmModels
+ ? "Loading models..."
+ : "No models detected. Configure OpenAI, Anthropic, Ollama, or IBM watsonx.ai first."
+ }
+ value={vlmModel}
+ onValueChange={handleVlmModelChange}
+ hasError={!!validationError}
+ disabled={!pictureDescriptions}
+ />
+ </LabelWrapper>
+ {providerWarning && (
+ <p className="text-sm text-destructive" role="alert">
+ Configure a provider with vision-capable models in
+ Settings &gt; Providers first.
+ </p>
+ )}
+ </div>
+
+ {vlmProvider === "watsonx" && (
<div className="space-y-2">
<LabelWrapper
- id="vlm-model"
- label="Vision model"
- helperText="Pick a vision-capable model; the provider is set from your selection"
- required={pictureDescriptions}
+ id="vlm-watsonx-api-version"
+ label="watsonx API version"
+ helperText="API version date sent to watsonx.ai"
>
- <ModelSelector
- groupedOptions={groupedVlmModels}
- noOptionsPlaceholder={
- isLoadingAnyVlmModels
- ? "Loading models..."
- : "No models detected. Configure OpenAI, Anthropic, Ollama, or IBM watsonx.ai first."
+ <Input
+ id="vlm-watsonx-api-version"
+ type="text"
+ placeholder={DEFAULT_WATSONX_API_VERSION}
+ value={vlmWatsonxApiVersion}
+ onChange={(e) =>
+ setVlmWatsonxApiVersion(e.target.value)
}
- value={vlmModel}
- onValueChange={handleVlmModelChange}
- hasError={!!validationError}
+ disabled={!pictureDescriptions}
/>
</LabelWrapper>
- {providerWarning && (
- <p className="text-sm text-destructive" role="alert">
- {providerLabel} is not configured. Configure it in
- Settings &gt; Providers first.
- </p>
- )}
</div>
+ )}
- {vlmProvider === "watsonx" && (
- <div className="space-y-2">
- <LabelWrapper
- id="vlm-watsonx-api-version"
- label="watsonx API version"
- helperText="API version date sent to watsonx.ai"
- >
- <Input
- id="vlm-watsonx-api-version"
- type="text"
- placeholder={DEFAULT_WATSONX_API_VERSION}
- value={vlmWatsonxApiVersion}
- onChange={(e) =>
- setVlmWatsonxApiVersion(e.target.value)
- }
- />
- </LabelWrapper>
- </div>
- )}
-
- <div className="space-y-2">
- <LabelWrapper
+ <div className="space-y-2">
+ <LabelWrapper
+ id="vlm-prompt"
+ label="Prompt"
+ helperText="Sent to the VLM for every page"
+ >
+ <Textarea
id="vlm-prompt"
- label="Prompt"
- helperText="Sent to the VLM for every page"
- >
- <Textarea
- id="vlm-prompt"
- rows={3}
- value={vlmPrompt}
- onChange={(e) => setVlmPrompt(e.target.value)}
- />
- </LabelWrapper>
- </div>
+ rows={3}
+ value={vlmPrompt}
+ onChange={(e) => setVlmPrompt(e.target.value)}
+ disabled={!pictureDescriptions}
+ />
+ </LabelWrapper>
+ </div>
- <div className="space-y-2">
- <LabelWrapper
- id="vlm-response-format"
- label="Response format"
- helperText="Per-page VLM output. Markdown is compatible with the existing pipeline; the final document is always Docling JSON."
+ <div className="space-y-2">
+ <LabelWrapper
+ id="vlm-response-format"
+ label="Response format"
+ helperText="Per-page VLM output. Markdown is compatible with the existing pipeline; the final document is always Docling JSON."
+ >
+ <Select
+ value={vlmResponseFormat}
+ onValueChange={setVlmResponseFormat}
+ disabled={!pictureDescriptions}
>
- <Select
- value={vlmResponseFormat}
- onValueChange={setVlmResponseFormat}
+ <SelectTrigger
+ id="vlm-response-format"
+ disabled={!pictureDescriptions}
>
- <SelectTrigger id="vlm-response-format">
- <SelectValue placeholder="Select a format" />
- </SelectTrigger>
- <SelectContent>
- {RESPONSE_FORMATS.map((format) => (
- <SelectItem
- key={format.value}
- value={format.value}
- >
- {format.label}
- </SelectItem>
- ))}
- </SelectContent>
- </Select>
- </LabelWrapper>
- </div>
+ <SelectValue placeholder="Select a format" />
+ </SelectTrigger>
+ <SelectContent>
+ {RESPONSE_FORMATS.map((format) => (
+ <SelectItem
+ key={format.value}
+ value={format.value}
+ >
+ {format.label}
+ </SelectItem>
+ ))}
+ </SelectContent>
+ </Select>
+ </LabelWrapper>
+ </div>
- <div className="grid grid-cols-3 gap-4">
- <NumberInput
- id="vlm-max-tokens"
- label="Max tokens per page"
- value={vlmMaxTokens}
- onChange={(value) =>
- setVlmMaxTokens(Math.max(1, value))
- }
- unit="tokens"
- min={1}
- />
- <NumberInput
- id="vlm-concurrency"
- label="Concurrency"
- value={vlmConcurrency}
- onChange={(value) =>
- setVlmConcurrency(Math.max(1, value))
- }
- unit="requests"
- min={1}
- />
- <NumberInput
- id="vlm-timeout"
- label="API timeout"
- value={vlmTimeout}
- onChange={(value) =>
- setVlmTimeout(Math.max(1, value))
- }
- unit="seconds"
- min={1}
- />
- </div>
+ <div className="grid grid-cols-3 gap-4">
+ <NumberInput
+ id="vlm-max-tokens"
+ label="Max tokens per page"
+ value={vlmMaxTokens}
+ onChange={(value) =>
+ setVlmMaxTokens(Math.max(1, value))
+ }
+ unit="tokens"
+ min={1}
+ disabled={!pictureDescriptions}
+ />
+ <NumberInput
+ id="vlm-concurrency"
+ label="Concurrency"
+ value={vlmConcurrency}
+ onChange={(value) =>
+ setVlmConcurrency(Math.max(1, value))
+ }
+ unit="requests"
+ min={1}
+ disabled={!pictureDescriptions}
+ />
+ <NumberInput
+ id="vlm-timeout"
+ label="API timeout"
+ value={vlmTimeout}
+ onChange={(value) => setVlmTimeout(Math.max(1, value))}
+ unit="seconds"
+ min={1}
+ disabled={!pictureDescriptions}
+ />
+ </div>
- {validationError && (
- <p className="text-sm text-destructive" role="alert">
- {validationError}
- </p>
- )}
- </AccordionContent>
- </AccordionItem>
- </Accordion>
- </div>
+ {validationError && (
+ <p className="text-sm text-destructive" role="alert">
+ {validationError}
+ </p>
+ )}
+ </CollapsibleContent>
+ </Collapsible>
+ </>
)}
</div>
<div className="flex justify-end pt-2">
diff --git a/frontend/components/ui/inputs/number-input.tsx b/frontend/components/ui/inputs/number-input.tsx
index 3a5969560..d81c3e860 100644
--- a/frontend/components/ui/inputs/number-input.tsx
+++ b/frontend/components/ui/inputs/number-input.tsx
@@ -35,39 +35,45 @@ export const NumberInput = ({
min={min}
value={value}
onChange={(e) => onChange(parseInt(e.target.value) || 0)}
- className="w-full pr-20 [appearance:textfield] [&::-webkit-outer-spin-button]:appearance-none [&::-webkit-inner-spin-button]:appearance-none"
+ className={
+ disabled
+ ? "w-full pr-8 [appearance:textfield] [&::-webkit-outer-spin-button]:appearance-none [&::-webkit-inner-spin-button]:appearance-none"
+ : "w-full pr-20 [appearance:textfield] [&::-webkit-outer-spin-button]:appearance-none [&::-webkit-inner-spin-button]:appearance-none"
+ }
/>
- <div className="absolute inset-y-0 right-0 top-0 p-[1px] py-[1.5px] flex items-center ">
- <span className="text-sm text-placeholder-foreground mr-4 pointer-events-none">
- {unit}
- </span>
- <div className="flex flex-col mt-[2px] mb-[2px]">
- <Button
- aria-label={`Increase ${label} value`}
- className="h-5 rounded-l-none rounded-br-none border-input border-t-transparent border-r-transparent border-b-[0.5px] hover:border-t-[.5px] hover:border-foreground"
- variant="outline"
- size="iconSm"
- onClick={() => onChange(value + 1)}
- >
- <Plus
- className="text-muted-foreground hover:text-foreground"
- size={8}
- />
- </Button>
- <Button
- aria-label={`Decrease ${label} value`}
- className="h-5 rounded-l-none rounded-tr-none border-input border-b-transparent border-r-transparent hover:border-b-1 hover:border-b-[.5px] hover:border-foreground"
- variant="outline"
- size="iconSm"
- onClick={() => onChange(value - 1)}
- >
- <Minus
- className="text-muted-foreground hover:text-foreground"
- size={8}
- />
- </Button>
+ {!disabled && (
+ <div className="absolute inset-y-0 right-0 top-0 p-[1px] py-[1.5px] flex items-center ">
+ <span className="text-sm text-placeholder-foreground mr-4 pointer-events-none">
+ {unit}
+ </span>
+ <div className="flex flex-col mt-[2px] mb-[2px]">
+ <Button
+ aria-label={`Increase ${label} value`}
+ className="h-5 rounded-l-none rounded-br-none border-input border-t-transparent border-r-transparent border-b-[0.5px] hover:border-t-[.5px] hover:border-foreground"
+ variant="outline"
+ size="iconSm"
+ onClick={() => onChange(value + 1)}
+ >
+ <Plus
+ className="text-muted-foreground hover:text-foreground"
+ size={8}
+ />
+ </Button>
+ <Button
+ aria-label={`Decrease ${label} value`}
+ className="h-5 rounded-l-none rounded-tr-none border-input border-b-transparent border-r-transparent hover:border-b-1 hover:border-b-[.5px] hover:border-foreground"
+ variant="outline"
+ size="iconSm"
+ onClick={() => onChange(value - 1)}
+ >
+ <Minus
+ className="text-muted-foreground hover:text-foreground"
+ size={8}
+ />
+ </Button>
+ </div>
</div>
- </div>
+ )}
</div>
</LabelWrapper>
);