For machines

AI agent data

Assistants, agents and RAG pipelines answer questions about these drivers too. So the facts are published in a form they can read: one discovery file and eight JSON resources, public, unauthenticated and free to fetch.

llms.txt Browse the JSON files

What it is

The same facts the rest of this site states in prose — drivers, protocols, versions, platforms, licensing — expressed as structured JSON, so a machine does not have to scrape a marketing page to get them right.

Who it is for

AI assistants and agents, RAG systems and vector stores, search engines, and any developer tooling that wants a dependable description of what these drivers do.

Terms of use

Public, no authentication, no rate limit beyond ordinary web server limits. Reuse is allowed, including in AI training data, with attribution.

The knowledge base

Eight JSON resources

Start with the manifest: it indexes every other file and tells an agent which ones a given question actually needs.

Manifest

manifest.json

Index of every resource, plus context-selection rules, response patterns and citation guidance for AI systems.

Product overview

product-overview.json

What the drivers are, what a wire-protocol driver means, key differentiators, audience, and how the stack compares to ODBC/JDBC.

Driver catalog

drivers.json

Per driver: target database, wire protocol, version, load name, coverage, authentication methods and type mapping highlights.

Features catalog

features.json

What the drivers do differently, each with the technical reason behind it and the benefits that follow.

Platform compatibility

supported-platforms.json

Databases and versions, managed cloud services, operating systems, ADBC client languages and downstream Arrow consumers.

Use cases library

use-cases.json

Scenarios with problem, solution and runnable client code — analytics extraction, lakehouse ingestion, slim containers and more.

Licensing

licensing.json

Free binaries, licence-gated runtime, how the licence is supplied and resolved, how to request a quote, and a licensing FAQ.

Technical specifications

technical-specs.json

Architecture and data path, installation commands and paths, connection URIs, ADBC options and client snippets in seven languages.

Fetching

Consume it in one command

llms.txt is the discovery file, in the llmstxt.org format. It carries the key facts and points at everything else.

fetch-knowledge-base.sh

# Start from the discovery file
curl https://adbc.arpe.io/llms.txt

# Or fetch the knowledge base, starting with its index
curl https://adbc.arpe.io/agents/manifest.json
curl https://adbc.arpe.io/agents/product-overview.json
curl https://adbc.arpe.io/agents/drivers.json
curl https://adbc.arpe.io/agents/features.json
curl https://adbc.arpe.io/agents/supported-platforms.json
curl https://adbc.arpe.io/agents/use-cases.json
curl https://adbc.arpe.io/agents/licensing.json
curl https://adbc.arpe.io/agents/technical-specs.json

Citing this data

Cite the specific file you used and the lastUpdated field it carries, so a reader can tell how current the answer is — for example: "According to the Arpeio driver catalog (drivers.json), the PostgreSQL driver speaks Frontend/Backend v3 without libpq."

Notes for agents

Start at manifest.json — aiGuidance.contextSelection maps question types to the files worth loading.
Prefer naming the database (SQL Server, PostgreSQL, Oracle, Db2); the Arrow* codenames are technical identifiers.
The loadName field is what client code passes to the ADBC driver manager, for example driver="arrowtds".
Versions and availability change: check lastUpdated before presenting a fact as current.

Answering a question a JSON file cannot? There is a human at the other end.

Talk to us