Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

More intuitive search settings interfaces #2899

Merged
merged 2 commits into from
Oct 24, 2024
Merged

More intuitive search settings interfaces #2899

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
fd099c2
clearer search settings interfaces
pablonyx Oct 24, 2024
17c345c
nits
pablonyx Oct 24, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 4 additions & 3 deletions web/src/app/admin/embeddings/RerankingFormPage.tsx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -90,9 +90,10 @@ const RerankingDetailsForm = forwardRef<

return (
<div className="p-2 rounded-lg max-w-4xl mx-auto">
<h2 className="text-2xl font-bold mb-4 text-text-800">
Post-processing
</h2>
<p className="mb-4">
Select from cloud, self-hosted models, or use no reranking
model.
</p>
<div className="text-sm mr-auto mb-6 divide-x-2 flex">
{originalRerankingDetails.rerank_model_name && (
<button
Expand Down
5 changes: 0 additions & 5 deletions web/src/app/admin/embeddings/interfaces.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,11 +21,6 @@ export enum RerankerProvider {
}

export interface AdvancedSearchConfiguration {
model_name: string;
model_dim: number;
normalize: boolean;
query_prefix: string;
passage_prefix: string;
index_name: string | null;
multipass_indexing: boolean;
multilingual_expansion: string[];
Expand Down
12 changes: 8 additions & 4 deletions web/src/app/admin/embeddings/pages/AdvancedEmbeddingFormPage.tsx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,11 @@ import * as Yup from "yup";
import { TrashIcon } from "@/components/icons/icons";
import { FaPlus } from "react-icons/fa";
import { AdvancedSearchConfiguration } from "../interfaces";
import { BooleanFormField } from "@/components/admin/connectors/Field";
import {
BooleanFormField,
Label,
SubLabel,
} from "@/components/admin/connectors/Field";
import NumberInput from "../../connectors/[connector]/pages/ConnectorInput/NumberInput";

interface AdvancedEmbeddingFormPageProps {
Expand All @@ -21,9 +25,6 @@ const AdvancedEmbeddingFormPage = forwardRef<
>(({ updateAdvancedEmbeddingDetails, advancedEmbeddingDetails }, ref) => {
return (
<div className="py-4 rounded-lg max-w-4xl px-4 mx-auto">
<h2 className="text-2xl font-bold mb-4 text-text-800">
Advanced Configuration
</h2>
<Formik
innerRef={ref}
initialValues={advancedEmbeddingDetails}
Expand Down Expand Up @@ -52,6 +53,9 @@ const AdvancedEmbeddingFormPage = forwardRef<
<FieldArray name="multilingual_expansion">
{({ push, remove }) => (
<div className="w-full">
<Label>Multi-lingual Expansion</Label>

<SubLabel>Add additional languages to the search.</SubLabel>
{values.multilingual_expansion.map(
(_: any, index: number) => (
<div key={index} className="w-full flex mb-4">
Expand Down
32 changes: 20 additions & 12 deletions web/src/app/admin/embeddings/pages/EmbeddingFormPage.tsx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,11 +33,6 @@ export default function EmbeddingForm() {

const [advancedEmbeddingDetails, setAdvancedEmbeddingDetails] =
useState<AdvancedSearchConfiguration>({
model_name: "",
model_dim: 0,
normalize: false,
query_prefix: "",
passage_prefix: "",
index_name: "",
multipass_indexing: true,
multilingual_expansion: [],
Expand Down Expand Up @@ -109,11 +104,6 @@ export default function EmbeddingForm() {
useEffect(() => {
if (searchSettings) {
setAdvancedEmbeddingDetails({
model_name: searchSettings.model_name,
model_dim: searchSettings.model_dim,
normalize: searchSettings.normalize,
query_prefix: searchSettings.query_prefix,
passage_prefix: searchSettings.passage_prefix,
index_name: searchSettings.index_name,
multipass_indexing: searchSettings.multipass_indexing,
multilingual_expansion: searchSettings.multilingual_expansion,
Expand Down Expand Up @@ -263,8 +253,8 @@ export default function EmbeddingForm() {
// This is a cloud model
newModel = {
...selectedProvider,
...rerankingDetails,
...advancedEmbeddingDetails,
...rerankingDetails,
provider_type:
(selectedProvider.provider_type
?.toLowerCase()
Expand All @@ -274,8 +264,8 @@ export default function EmbeddingForm() {
// This is a locally hosted model
newModel = {
...selectedProvider,
...rerankingDetails,
...advancedEmbeddingDetails,
...rerankingDetails,
provider_type: null,
};
}
Expand Down Expand Up @@ -388,6 +378,16 @@ export default function EmbeddingForm() {

{formStep == 1 && (
<>
<h2 className="text-2xl font-bold mb-4 text-text-800">
Select a Reranking Model
</h2>
<Text className="mb-4">
Updating the reranking model does not require re-indexing
documents. The reranker helps improve search quality by reordering
results after the initial embedding search. Changes will take
effect immediately for all new searches.
</Text>

<Card>
<RerankingDetailsForm
setModelTab={setModelTab}
Expand Down Expand Up @@ -429,6 +429,14 @@ export default function EmbeddingForm() {
)}
{formStep == 2 && (
<>
<h2 className="text-2xl font-bold mb-4 text-text-800">
Advanced Search Configuration
</h2>
<Text className="mb-4">
Configure advanced embedding and search settings. Changes will
require re-indexing documents.
</Text>

<Card>
<AdvancedEmbeddingFormPage
advancedEmbeddingDetails={advancedEmbeddingDetails}
Expand Down
1 change: 0 additions & 1 deletion web/src/app/chat/ChatPage.tsx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -809,7 +809,6 @@ export function ChatPage({

// Check if all messages are currently rendered
if (currentVisibleRange.end < messageHistory.length) {
console.log("Updating visible range");
// Update visible range to include the last messages
updateCurrentVisibleRange({
start: Math.max(
Expand Down
9 changes: 0 additions & 9 deletions web/src/components/embedding/CustomEmbeddingModelForm.tsx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,10 +33,6 @@ export function CustomEmbeddingModelForm({
api_url: provider.api_url,
description: "",
index_name: "",
pricePerMillion: 0,
mtebScore: 0,
maxContext: 4096,
max_tokens: 1024,
}
}
validationSchema={Yup.object().shape({
Expand All @@ -51,13 +47,8 @@ export function CustomEmbeddingModelForm({
api_url: Yup.string().required("API base URL is required"),
description: Yup.string(),
index_name: Yup.string().nullable(),
pricePerMillion: Yup.number(),
mtebScore: Yup.number(),
maxContext: Yup.number(),
max_tokens: Yup.number(),
})}
onSubmit={async (values) => {
console.log(values);
setShowTentativeModel(values as CloudEmbeddingModel);
}}
>
Expand Down
31 changes: 0 additions & 31 deletions web/src/components/embedding/interfaces.tsx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -58,17 +58,10 @@ export interface EmbeddingModelDescriptor {

export interface CloudEmbeddingModel extends EmbeddingModelDescriptor {
pricePerMillion: number;
enabled?: boolean;
mtebScore: number;
maxContext: number;
}

export interface HostedEmbeddingModel extends EmbeddingModelDescriptor {
link?: string;
model_dim: number;
normalize: boolean;
query_prefix: string;
passage_prefix: string;
isDefault?: boolean;
}

Expand Down Expand Up @@ -197,9 +190,6 @@ export const AVAILABLE_CLOUD_PROVIDERS: CloudEmbeddingProvider[] = [
description:
"Cohere's English embedding model. Good performance for English-language tasks.",
pricePerMillion: 0.1,
mtebScore: 64.5,
maxContext: 512,
enabled: false,
model_dim: 1024,
normalize: false,
query_prefix: "",
Expand All @@ -214,9 +204,6 @@ export const AVAILABLE_CLOUD_PROVIDERS: CloudEmbeddingProvider[] = [
description:
"Cohere's lightweight English embedding model. Faster and more efficient for simpler tasks.",
pricePerMillion: 0.1,
mtebScore: 62,
maxContext: 512,
enabled: false,
model_dim: 384,
normalize: false,
query_prefix: "",
Expand Down Expand Up @@ -247,9 +234,6 @@ export const AVAILABLE_CLOUD_PROVIDERS: CloudEmbeddingProvider[] = [
normalize: false,
query_prefix: "",
passage_prefix: "",
mtebScore: 64.6,
maxContext: 8191,
enabled: false,
index_name: "",
api_key: null,
api_url: null,
Expand All @@ -264,9 +248,6 @@ export const AVAILABLE_CLOUD_PROVIDERS: CloudEmbeddingProvider[] = [
description:
"OpenAI's newer, more efficient embedding model. Good balance of performance and cost.",
pricePerMillion: 0.02,
enabled: false,
mtebScore: 62.3,
maxContext: 8191,
index_name: "",
api_key: null,
api_url: null,
Expand All @@ -290,9 +271,6 @@ export const AVAILABLE_CLOUD_PROVIDERS: CloudEmbeddingProvider[] = [
model_name: "text-embedding-004",
description: "Google's most recent text embedding model.",
pricePerMillion: 0.025,
mtebScore: 66.31,
maxContext: 2048,
enabled: false,
model_dim: 768,
normalize: false,
query_prefix: "",
Expand All @@ -306,9 +284,6 @@ export const AVAILABLE_CLOUD_PROVIDERS: CloudEmbeddingProvider[] = [
model_name: "textembedding-gecko@003",
description: "Google's Gecko embedding model. Powerful and efficient.",
pricePerMillion: 0.025,
mtebScore: 66.31,
maxContext: 2048,
enabled: false,
model_dim: 768,
normalize: false,
query_prefix: "",
Expand All @@ -335,9 +310,6 @@ export const AVAILABLE_CLOUD_PROVIDERS: CloudEmbeddingProvider[] = [
description:
"Voyage's large embedding model. High performance with instruction fine-tuning.",
pricePerMillion: 0.12,
mtebScore: 68.28,
maxContext: 4000,
enabled: false,
model_dim: 1024,
normalize: false,
query_prefix: "",
Expand All @@ -352,9 +324,6 @@ export const AVAILABLE_CLOUD_PROVIDERS: CloudEmbeddingProvider[] = [
description:
"Voyage's lightweight embedding model. Good balance of performance and efficiency.",
pricePerMillion: 0.12,
mtebScore: 67.13,
maxContext: 16000,
enabled: false,
model_dim: 1024,
normalize: false,
query_prefix: "",
Expand Down