Skip to contentК содержанию
LLMcode
ModelsМоделиPricingЦеныInstructionsИнструкцииDocsДокументацияNewsНовости
ENRU
EnglishРусский
Sign inВойтиCreate accountСоздать аккаунтAccountАккаунт
MenuМеню
ModelsМоделиPricingЦеныInstructionsИнструкцииDocsДокументацияNewsНовостиCreate accountСоздать аккаунт
LLMcode

Model names in the API are LLMcode profiles with their own routing. They do not mean direct access to the same-named models of OpenAI LLC or Anthropic PBC. Trademarks belong to their owners. The service is independent and is not affiliated with OpenAI or Anthropic.Названия моделей в API — профили LLMcode с собственной маршрутизацией. Они не означают прямой доступ к одноимённым моделям OpenAI LLC или Anthropic PBC. Товарные знаки принадлежат их правообладателям. Сервис независим и не аффилирован с OpenAI/Anthropic.

CatalogКаталог

ModelsМоделиPricingЦеныInstructionsИнструкцииDocsДокументацияKey checkПроверка ключа

LegalДокументы

OfferОфертаPrivacyКонфиденциальностьConsentСогласиеAcceptable useДопустимое использованиеRefundВозврат

LearnМатериалы

GuidesГайдыComparisonsСравненияAlternativesАльтернативыIntegrationsИнтеграцииUse casesСценарииGlossaryГлоссарийNewsНовостиRSSTelegramTelegram-канал
HomeГлавнаяModelsМоделиPricingЦеныSign inВойтиAccountАккаунт
LLMcode/Use cases/Retrieval-augmented generation over your own documents

Use case

Retrieval-augmented generation over your own documents

RAG answers questions from your data instead of from memory. LLMcode serves embeddings and chat on the same key, so the whole pipeline has one billing surface.

Updated 2026-11-16

Embed, store, retrieve

Turn documents into vectors with POST /v1/embeddings, store them in a vector index you operate, and retrieve the top chunks for each question.

The gateway does not host a vector store; the index lives on your side. This keeps the retrieval quality under your control.

Answer with grounding

Feed the retrieved chunks into the prompt and ask the chat profile to answer only from them, citing the source chunk ids.

A larger profile helps with synthesis and long contexts; a cheaper one handles short, well-grounded answers. Choose per request.

Measure what matters

Retrieval quality is a search problem: measure recall against a golden set before touching the model.

Billing is per token for embeddings and chat from the same balance. The journal separates models, so the two legs of the pipeline are easy to compare.

Published model prices

USD per 1M tokens. The numbers come from the public catalog on every page rebuild — this is the sale price, not an estimate.

ModelInputOutputCache
Claude Opus 5.5$2.70$8.00$0.14
DeepSeek Flash$0.50$2.00$0.01
Claude Sonnet 5.5$1.33$6.67$0.07
DeepSeek V4 Pro$2.20$6.60$0.07

Top-up from $10, balance in USD. 1 USD = 1,000,000 UT. No subscription and no auto-charge.

Questions

Which embeddings are available?

The published catalog lists embedding-enabled profiles. Check /models or GET /v1/models with your key.

Do I need two keys for embeddings and chat?

No. One key works across endpoints and models; the model is chosen per request.

Is the vector store hosted?

No. The index is operated by you; the API serves embeddings and completions only.

How is caching billed?

When the provider reports cached input tokens, the separate cache price applies. It is not guaranteed for every request.

See also

OpenAI-compatible APIModels and pricesPricing and top-upSetup instructionsAPI docs

One key for every model

OpenAI- and Anthropic-compatible endpoints, balance in USD, top-up from $10.

Create accountPricing
Base URL
https://llmcode.io/v1
Key format
sk-llmcode-…