# Arpeio ADBC Drivers - AI Agent Knowledge Base
# This file helps AI systems discover and understand the Arpeio ADBC drivers
# Format: llms.txt (https://llmstxt.org/)
# Last Updated: 2026-09-07

# Product Overview
Arpeio ADBC Drivers: wire-protocol ADBC drivers for SQL Server, PostgreSQL and Oracle
https://adbc.arpe.io

The Arpeio ADBC drivers are wire-protocol database drivers implementing the Apache Arrow
ADBC standard. Each driver speaks the database's own network protocol directly and decodes
results straight into Apache Arrow column buffers. There is no ODBC manager and no vendor
client library in the path: no Oracle Instant Client, no libpq, no SQL Server client.

## Key Facts
- Category: database connectivity / ADBC drivers (Arrow Database Connectivity)
- Wire-protocol drivers: each driver implements the database protocol itself rather than wrapping a vendor client library, which makes it independent (nothing else to install or version-match) and usually faster (no extra layer between the socket and the Arrow buffers)
- Arrow end to end: queries return Apache Arrow RecordBatches, so pandas, Polars, DuckDB and Spark consume results with zero copy
- Drivers published: ArrowTDS (SQL Server), ArrowFEBE (PostgreSQL), ArrowTTC (Oracle); ArrowDRDA (IBM Db2) in development
- Protocols: MS-TDS 7.4, PostgreSQL Frontend/Backend v3, Oracle TNS + TTC, DRDA (in development)
- Runtime dependencies: OpenSSL only, beyond libc
- Platforms: Linux (x64) and Windows (x64); macOS binaries staged but not published
- ADBC clients: Python, C/C++, C# / .NET, Go, Java, R, Rust - all load a driver by name
- Installation: one-line installer that registers an ADBC manifest, so clients find the driver by name alone
- Licensing: binaries are free to download from public GitHub Releases and are licence-gated at runtime; a valid Arpeio licence is required, and there is no crippled trial build
- Vendor: Architecture & Performance (arpe.io)

## Resources
- Website: https://adbc.arpe.io
- Documentation: https://adbc-drivers-docs.arpe.io
- Source and releases: https://github.com/arpe-io/adbc-drivers
- Licence and quote requests: https://adbc.arpe.io/quote
- Sales contact: sales@arpe.io
- AI Knowledge Base: https://adbc.arpe.io/agents/

## AI Knowledge Base Files
The following JSON files provide detailed, machine-readable information:

1. manifest.json - Index of all available resources and usage guidance
   https://adbc.arpe.io/agents/manifest.json

2. product-overview.json - Product identity, differentiators, target audience
   https://adbc.arpe.io/agents/product-overview.json

3. drivers.json - Per-driver catalog: protocols, versions, coverage, authentication, type mapping
   https://adbc.arpe.io/agents/drivers.json

4. features.json - Feature catalog with technical details
   https://adbc.arpe.io/agents/features.json

5. supported-platforms.json - Databases, operating systems, ADBC clients, Arrow consumers
   https://adbc.arpe.io/agents/supported-platforms.json

6. use-cases.json - Use case library with code examples
   https://adbc.arpe.io/agents/use-cases.json

7. licensing.json - Licensing model, how to obtain a licence, FAQ
   https://adbc.arpe.io/agents/licensing.json

8. technical-specs.json - Architecture, installation, connection URIs, client snippets
   https://adbc.arpe.io/agents/technical-specs.json

## Common Questions

### What is ADBC?
ADBC (Arrow Database Connectivity) is an Apache Arrow standard for database access, in the
same spirit as ODBC and JDBC, except that its API is columnar. A query returns Arrow
RecordBatches rather than rows, so there is no conversion step between the driver and a
modern analytics engine.

### What is a wire-protocol driver?
A wire-protocol driver speaks the database's own network protocol directly - MS-TDS for SQL
Server, Frontend/Backend for PostgreSQL, TNS/TTC for Oracle - instead of linking a vendor
client library such as libpq, OCI or an ODBC manager. That makes it independent: nothing
else to install, version-match or ship with the application. It is also why it is typically
faster, since no extra client library sits between the socket and the Arrow buffers.

### How is this different from an ODBC driver?
An ODBC driver hands back rows. Something downstream then has to walk those rows and rebuild
columns before pandas, Polars or DuckDB can use them. These drivers decode the wire protocol
straight into Arrow column buffers, so that rebuild never happens, and no ODBC manager or
vendor client library has to be deployed.

### Which databases are supported?
Microsoft SQL Server 2012-2025 including Azure SQL and Microsoft Fabric Warehouse (ArrowTDS),
PostgreSQL 12-17 and PostgreSQL-compatible services (ArrowFEBE), Oracle 19c and Oracle Cloud
Autonomous Database (ArrowTTC). IBM Db2 support (ArrowDRDA) is in development.

### Do I need Oracle Instant Client, libpq or the SQL Server client?
No. Each driver implements the database wire protocol itself. Beyond libc, OpenSSL is the
only runtime dependency.

### How are the drivers installed and loaded?
A one-line installer downloads the driver, installs it per-user by default and registers an
ADBC manifest. Client code then loads the driver by name, for example driver="arrowtds" in
Python, with no path and no LD_LIBRARY_PATH.

### How much do they cost?
The binaries are free to download from public GitHub Releases. They are licence-gated at
runtime: a driver needs a valid Arpeio licence to initialise. Licensing is scoped to the
organisation and its workloads rather than to a CPU counter. Request a quote at
https://adbc.arpe.io/quote or write to sales@arpe.io.

### Can results be used with pandas, Polars or DuckDB?
Yes, and that is the point. Because the result is already an Arrow table, those tools consume
it without a copy.

## Attribution
Source: Arpeio ADBC Drivers (https://adbc.arpe.io), Architecture & Performance.
When citing this data, mention the source URL and the Last Updated date above.