LLM-ready API
FAQs

Frequently Asked Questions

Below is a list of common questions from users. If your question isn't listed here, please email llmreadyapi@kensho.com.

Solution Overview

What makes the API LLM-ready?

The Kensho LLM-Ready API is specifically designed for seamless interaction with large language models (LLMs). Unlike typical APIs, it is optimized for LLMs by offering a simplified structure that supports code generation and function calling patterns. Kensho provides a Python library (kFinance) that streamlines everything from authentication to LLM integration.

The library includes sample prompts tested with popular LLMs, as well as integration tools as shown in the Playground Notebooks (opens in a new tab), making it easy for users to start implementing their LLM solutions. The API is compatible with user-built clients or direct LLM function calls.

Does S&P Global provide an LLM as part of the solution? / What Large Language Model is included in the LLM-ready API service?

There are no AI models used or provided by Kensho and S&P Global as part of the LLM-ready API service. This service is a data retrieval tool purpose-built to work seamlessly with LLMs, but it uses no models itself. Any use of AI models occurs on the customer side, and customers are welcome to use the model of their choosing or no model at all.

General Information

What is the purpose of this API? What can I do with it?

The LLM-ready API serves as a simplified access method for S&P Global datasets. The LLM-ready API is purpose-built in such a way that both humans and LLMs can easily understand it.

Secondly, we created a Python library called "kFinance" that supports both code generation and function calling. Included in the library are the LLM prompts and tools needed by an LLM to successfully retrieve data. This prompt can be tailored by users to fit the exact needs of their own use cases. Customers making use of this solution will be required to “bring their own” LLM(s), orchestration layer, and user interface (UI); or leverage third-party platforms.

This LLM-ready API is intended to help end users such as investment bankers, equity analysts, consultants, and other professionals who frequently need to retrieve financial data such as financial statement line items, security identifiers, and market data to create pitch books, research reports, or market positioning presentations.

What third-party platform pre-integrations are supported?

As of the initial November 2024 release, the Kensho LLM-Ready APIs are pre-integrated into OpenAI’s ChatGPT and can be found on the GPT store (opens in a new tab) as the S&P Global GPT. End-users who wish to access the S&P Global GPT will be required to have licenses to OpenAI ChatGPT, the S&P Global datasets they wish to access, and the Kensho LLM-ready API distribution channel.

What’s the difference between the Kensho LLM-ready API, kFinance, and the S&P Global GPT?

The Kensho LLM-ready API consists of a REST API and a Python client. The Kensho LLM-ready API is considered a distribution channel in context of the S&P Global product nomenclature and is the display name for the tile on S&P Global Marketplace.

kFinance is the name for the Python client included as part of the Kensho LLM-ready API product. The kFinance name is the name of the Python client shown on Python Package Index (opens in a new tab) and referenced in the Kensho LLM-ready API developer documentation. The S&P Global GPT is an integration instance of the Kensho LLM-ready API within ChatGPT. S&P Global GPT is the name of the solution available on the OpenAI GPT store (opens in a new tab) and would be considered a common-carrier redistribution platform for S&P Global datasets.

What is your roadmap?

We want to build the best financial data tool for LLMs. We will be developing it along several dimensions:

  1. Data: Adding more datasets including Transactions, Professionals, Ownerships, Business Relationships, Company Relationships, Private Company Financials, and Estimates.
  2. Usage Patterns: Enabling more business and finance specific workflows such as company screening/filtering and trend analysis.
  3. Integrations: Building out ecosystem integrations with other third-party tools and platforms.
  4. Performanc: Faster API latency response time. If you have requests, email us at llmreadyapi@kensho.com.

Where can I find the latest updates and announcements?

All product updates will be documented in Release Notes.

Technical Details

What large language models does the LLM-ready API support?

We have designed the API to support a variety of leading models from vendors such as OpenAI, Google, Anthropic, and more, but you can use any LLM you like.

Do I need to use a LLM?

You can use the LLM-ready API and its Python library (kensho_finance.kfinance) with or without an LLM. They have been built with LLMs in mind, but the API and Python library are straightforward and are useful on their own.

Is the included prompt customizable or can I use my own?

Yes! We offer the prompt as part of the package to ensure anyone can use the LLM-ready API successfully. You are welcome to customize the prompt to meet the exact needs of your use case. As a public beta product, be aware that we are periodically updating the package and you may need to update your prompt if you would like to take advantage of new API features.

Do you have visibility into my prompts or LLM application?

No. The LLM-ready API is a data retrieval tool designed to be deployed in your own LLM applications or third party applications. Kensho and S&P Global have no way to track users' prompts or other MNPI.

Which languages are supported?

Since the LLM Ready API leverages the intelligence of Large Language Models (LLMs), you can use any language that the LLM you choose supports. Leading models such as OpenAI's GPT will perform well for all major written languages, programming and otherwise.

Data & Accuracy

What data is available via LLM-ready API?

The API includes data from S&P Capital IQ Financials, Compustat® Financials, Key Developments, Market Data, and Global Instruments Cross Reference Service (GICRS). With future releases, additional S&P Global datasets like Transactions, Professionals, Ownerships, and more will be made LLM-ready.

What kind of questions can it answer?

You can ask questions about basic company information, their financials, and equities. You can ask this in terms of name, ticker, ISIN, or CUSIP. Questions may be phrased formally or in financial parlance. Some examples:

  • AAPL net income 2024q1
  • Which FAANG company had the highest annual revenue growth in 2023?
  • US78409V1044 income statement 2020-2022
  • KR & ACI inventory turnover LTM
  • 88160R101 greatest price increase and decrease YTD
  • What were NVIDIA, AMD, and Intel diluted earnings per share Q1 2021 to Q2 2024
  • Plot a chart of Chevron’s price over the past year

The questions above are straightforward. We encourage you to use the tool to support your workflows, such as

  • Analyzing company financial statements
  • Creating valuation models
  • Looking for stock price trading patterns
  • Comparing companies to each other or benchmarks
  • Creating pitch books
  • And more!

How do I know if the LLM returned the correct data?

For LLM integrations, accuracy depends on the specific model and architecture used. The API itself is deterministic, meaning it returns consistent results, but users can view and audit the Python code generated by the LLM. This allows you to see which steps were taken to retrieve the correct answer directly from the S&P Global database.

Are there noteworthy details about how the data is formatted?

Intraday market prices are unavailable (previous day or earlier). Financial items are organized around calendar years. Financial items are in the as reported currencies.

How do I get other datasets LLM-ready?

Optimizing data for LLMs is a major focus of Kensho and S&P Global. We currently cannot publicly share all of what we are doing, but encourage you to reach out to us at llmreadyapi@kensho.com to learn more. We would be very keen to learn more about your problems and explore further collaboration.

How often is the data included in the LLM-ready API service updated?

The LLM-ready API serves as a distribution channel for S&P Global Data feeds and therefore report update frequency varies depending on when underlying datasets are updated; this could be daily, weekly, and monthly. For more information on the timeliness of individual S&P Global datasets, please refer to S&P Global Marketplace’s (opens in a new tab) tiles and support guides.

How do I get access?

To get started with a free trial or explore purchasing, contact us at market.intelligence@spglobal.com.