Match the words your customers never typed

Superlinked gives your agent one API to expand search queries and support messages into weighted terms, so a sparse index matches listings that use different words.

Get started

Shopper query

color switching led lights

Product listing

ILC RGB LED Light Bulb, Color Changing Light Bulb, 40W Equivalent, 450LM, 2700K Warm White 5W E26 Screw Base RGBW, Flood Light Bulb- 12 Color Choices - Timing Infrared Remote Control (4 Pack)

Shared term	Query	Listing	Match
`led`	1.908	1.449	2.765
`color`	1.916	1.345	2.577
`lights`	1.582	1.138	1.800
`colors`	1.299	1.352	1.755
`light`	1.359	1.237	1.682
`change`	0.882	1.316	1.161
`bulb`	0.651	1.708	1.112
11 more shared terms			1.146

Sparse match score 13.998

Words in both texts: 5.342
Terms SPLADE added: 8.656

Customer messages and product texts gain the vocabulary exact-match search misses

+ marks a term the model added. Everything else is a word the text already used.

View evidence

Card delivery

My card still hasn't arrived after 2 weeks. Is it lost?

card 2.10
added by the model week 1.72
lost 1.65
added by the model cards 1.61
still 1.61
weeks 1.56
added by the model arrive 1.44
2 1.42
added by the model years 1.14
added by the model two 1.09
added by the model loss 1.00
after 0.95

PolyAI BANKING77 test set, CC BY 4.0

Transfer

I transferred money but the recipient says it has not arrived. Why would this be?

added by the model transfer 1.78
added by the model arrive 1.76
money 1.53
not 1.42
arrived 1.39
transferred 1.38
recipient 1.37
added by the model because 1.23
added by the model say 1.13
added by the model arrival 1.13
added by the model have 0.78
added by the model said 0.77

PolyAI BANKING77 test set, CC BY 4.0

Card freeze

Can you freeze my card because someone used it while I was out of town. I did not make these purchases.

freeze 1.82
added by the model frozen 1.52
card 1.47
town 1.27
added by the model cards 1.20
someone 1.15
added by the model freezing 1.08
because 0.98
can 0.92
not 0.92
out 0.83
added by the model use 0.79

PolyAI BANKING77 test set, CC BY 4.0

ATM cash

I'm at an ATM and withdrew 30 pounds and was only given 10. What should I do?

atm 2.03
added by the model pound 1.41
added by the model withdraw 1.38
pounds 1.30
added by the model weigh 1.28
added by the model withdrawal 1.26
30 1.21
added by the model much 1.13
should 1.02
10 0.98
withdrew 0.94
added by the model weight 0.90

PolyAI BANKING77 test set, CC BY 4.0

Product search

chrome notebook

chrome 3.28
notebook 2.91
added by the model laptop 1.82
added by the model ##book 0.93
added by the model browser 0.82
added by the model document 0.61
added by the model my 0.57
added by the model computer 0.44
added by the model pen 0.43
added by the model web 0.41
added by the model computing 0.41
added by the model device 0.40

Amazon Shopping Queries Dataset, example 530782, Apache-2.0

Recall notice

The lithium-ion battery in the recalled power banks can overheat and ignite, posing fire and burn hazards to consumers.

lithium 2.18
recalled 2.04
added by the model recall 1.97
added by the model bank 1.95
power 1.77
##hea 1.77
battery 1.70
banks 1.70
fire 1.67
added by the model batteries 1.55
ion 1.51
ign 1.50

CPSC recall 25437, public domain

One encode call returns weighted terms ready for your inverted index

View on GitHub

from sie_sdk import SIEClient
client = SIEClient(
    api_key="sk-sie-…",
    base_url="https://api.superlinked.com",
)
res = client.encode(
    "modelprithivida/Splade_PP_en_v2",
    [{"text": "textDo ATM cash withdrawals carry a charge now? They've been free in the past, but all of a sudden I have to pay to make ATM withdrawals?"}],
    output_types=["sparse"],
)
sparse = res[0]["sparse"]  # indices = term ids, values = weights
pairs = zip(sparse["indices"], sparse["values"])
top = sorted(pairs, key=lambda p: -p[1])[:12]
for term_id, weight in top:
    print(f"{float(weight):.2f}  {term_id}")

import { SIEClient } from '@superlinked/sie-sdk';

const client = new SIEClient('https://api.superlinked.com', {
  apiKey: 'sk-sie-…',
});

const [res] = await client.encode(
  'prithivida/Splade_PP_en_v2',
  [{ text: "Do ATM cash withdrawals carry a charge now? They've been free in the past, but all of a sudden I have to pay to make ATM withdrawals?" }],
  { outputTypes: ['sparse'] },
);
// sparse: { indices, values }, mapping learned term ids to weights
const sparse = res!.sparse!;
const top = [...sparse.indices]
  .map((id, i) => ({ id, weight: sparse.values[i] }))
  .sort((a, b) => b.weight - a.weight);
console.log(top.slice(0, 12));

curl https://api.superlinked.com/v1/encode/prithivida%2FSplade_PP_en_v2 \
  -H "Authorization: Bearer sk-sie-…" \
  -H "Content-Type: application/json" \
  -d "{\"items\":[{\"text\":\"Do ATM cash withdrawals carry a charge now? They've been free in the past, but all of a sudden I have to pay to make ATM withdrawals?\"}],\"params\":{\"output_types\":[\"sparse\"]}}"

Build the "Sparse embeddings" capability into my app using the Superlinked Inference Engine (SIE).

Context
- SIE is an OpenAI-style inference API. Python SDK: `from sie_sdk import SIEClient`; TypeScript: `@superlinked/sie-sdk`.
- Base URL: https://api.superlinked.com (or my regional endpoint). Auth: Bearer key from env `SIE_API_KEY` (never hard-code it).
- Model: prithivida/Splade_PP_en_v2 (SIE primitive: /encode). Keep the model id configurable.

Task
- Input: a block of text.
- Behaviour: return the learned sparse term ids and weights, ordered by strongest activation
- Call the selected SIE primitive once per request and map the response into your domain type.

Deliverables
- A typed client wrapper, an application-level function for this task, error handling for timeouts/empty input, and unit tests with a stubbed client.
- Wire it into my existing stack (ask me which framework if unclear) and add a short usage example.

Run

Output

27218atm

10534withdrawal

4287carry

2489free

3715charge

5356cash

10632withdrawexpanded

2085now

2025notexpanded

5571chargesexpanded

4755carryingexpanded

7408feeexpanded

44 active terms · 33 added by the model

Term ids and weights come from the encode call
Tokens decoded with the pinned SPLADE vocab.txt

PRICE

$ / 1M input tokens ↓

$0.0256 SIE Splade PP

$0.0408 SIE bge-m3 sparse

Elastic ELSER $0.08

Pinecone sparse-english-v0 $0.08

QUALITY

nDCG@10 · NFCorpus ↑

SIE Splade PP 0.33

0.30

SIE bge-m3 sparse 0.28

LATENCY

p50 ms ↓

62ms

65ms

SIE Splade PP 144ms

SIE bge-m3 sparse 173ms

Deploy your way

Managed Cloud

Full compute toolkit for your agents with zero ops.

No idle GPUs, pay for what you use
Fits your stack: SDK, API, CLI, MCP
Zero lock-in, self-host the same stack
SOC 2 Type 2, US or EU data residency

$50 Example credit balance

1.3B tokens embedded

120k pages parsed

27M tokens generated (1:1 in:out)

Get started

no credit card required

Self-host with K8s

Easy & scalable deployment in your own cloud.

Terraform to your cloud in minutes
Apache-2.0, same engine as Cloud
Scales to zero, no bill between jobs
Per-tenant pools, no noisy neighbors

Agent prompt

Deploy SIE to our AWS account with the superlinked/sie/aws Terraform module. Docs: superlinked.com/docs/deploymentDeploy SIE to our GCP project with the superlinked/sie/google Terraform module. Docs: superlinked.com/docs/deploymentDeploy SIE to our Azure AKS cluster via helm install. Requirements: superlinked.com/docs/deployment

Deploy guide

Run locally

Run the same models on your own machine.

Runs on NVIDIA GPU or Apple Silicon
One command, no Docker or cluster
All 100+ Cloud models, fully offline
Same SDK and IDs, no code changes

pip install "sie-server[local]" && sie-server servepip install "sie-server[local]" && sie-server serve --device cuda

Quickstart

Customer messages and product texts gain the vocabulary exact-match search misses

One encode call returns weighted terms ready for your inverted index

Sparse retrieval quality and latency

Deploy your way