FREE BROWSER-LOCAL AVRO TOOLS

Open Avro files. See the data inside.

Inspect Avro records and their embedded writer schema in one free, browser-local workspace. Choose a file to begin.

No account neededFiles stay in your browser

Download a sample Avro file
  1. Choose your .avro container file.
  2. Search records and inspect the embedded schema.
  3. Export records as JSON.

Inspect records, embedded schemas, blocks and codecs. Record tables also support CSV. Large files export the displayed page as JSON. JSON/CSV exports are limited to 10 MB. File contents are decoded locally; normal page requests reach the hosting provider.

Schema beside your records

Inspect the embedded writer schema, named types and logical-type annotations alongside your data.

Useful local exports

Search and filter records, then export JSON or CSV using the existing Avro workspace.

Verified container decoding

Uncompressed, deflate and Snappy blocks are supported. Sync markers and Snappy checksums are checked; unsupported files show an error.

How to inspect an Avro file

  1. Choose your Avro Object Container File or try the fictional sample.
  2. The shared application opens with the original file already loaded.
  3. Inspect records and schema, then export what you need.

Keep your workspace when returning to the page

Temporary handoffs use local browser storage and expire after 30 minutes. Refresh can reopen the original file while its handoff remains available; download edits before reloading. Back to website preserves the live workspace for explicit Resume.

Other ways to inspect an Avro file

Inspect Avro with Python

Read the embedded writer schema, codec and first ten records without collecting the complete file.

Before you start

  • Python 3 and the packages below.
  • Work in a local folder containing your Avro Object Container File as input.avro. These examples do not use a server or upload the file.
  1. Save the Python example as inspect_avro.py and run python inspect_avro.py. It prints schema JSON, the codec and at most ten records.
Install in a Python environment
# macOS/Linux (Windows: py -m venv .venv):
python3 -m venv .venv
# Windows PowerShell activation: .venv\Scripts\Activate.ps1
. .venv/bin/activate
python -m pip install fastavro python-snappy

Inspect schema and preview records
import json
from itertools import islice
from pprint import pprint
from fastavro import reader

with open("input.avro", "rb") as source:
    records = reader(source)
    print(json.dumps(records.writer_schema, indent=2))
    print("Codec:", records.codec)
    for record in islice(records, 10):
        pprint(record)

Limits and output

  • This previews ten records; it does not validate every record or edit the file. Decoding still needs memory for individual blocks and values. Install codec support required by your file.

Documentation: fastavro file reader

Inspect Avro with DuckDB

Use SQL to inspect column types and a ten-row preview.

Before you start

  • Install the DuckDB CLI from duckdb.org. The initial INSTALL downloads its Avro extension.
  • Work in a local folder containing your Avro Object Container File as input.avro. These examples do not use a server or upload the file.
  1. Run duckdb in this folder and paste the SQL below. DESCRIBE prints mapped column names/types; SELECT returns at most ten rows.
Schema and bounded preview
INSTALL avro;
LOAD avro;
DESCRIBE SELECT * FROM read_avro('input.avro');
SELECT * FROM read_avro('input.avro') LIMIT 10;

Limits and output

  • DuckDB maps the embedded Avro schema into SQL types, which is different from displaying the original schema. Recursive schemas and some platform builds are unsupported. Nullable unions are simplified; other unions can become SQL union types.

Documentation: DuckDB Avro extension

Inspect Avro with ClickHouse local

Run the ClickHouse binary locally to inspect inferred SQL types and ten records.

Before you start

  • Install the ClickHouse binary using the official local-tool instructions linked below and make clickhouse available on your PATH.
  • Work in a local folder containing your Avro Object Container File as input.avro. These examples do not use a server or upload the file.
Inspect columns and preview rows
clickhouse local --query "DESCRIBE TABLE file('input.avro', 'Avro')"
clickhouse local --query "SELECT * FROM file('input.avro', 'Avro') LIMIT 10 FORMAT PrettyCompact"

Limits and output

  • These commands run locally; using file() on a ClickHouse server would instead read that server’s filesystem. Avro types are mapped into supported ClickHouse types. A ten-row preview is not a complete file validation.

Documentation: ClickHouse local installation and usage; ClickHouse Avro format

Frequently asked questions

Does my Avro file get uploaded?

Local editing decodes files in your browser. Optional assistance sends only your goal, an explicitly included schema and relevant requested record count to OpenRouter and the configured model provider; OCF record values are not included automatically. Original-input handoffs use local browser storage and expire after 30 minutes. Normal page and sample requests reach the hosting provider.

Which Avro files and codecs can I open?

Choose an Avro Object Container File (.avro) containing its embedded writer schema. Uncompressed, deflate and Snappy blocks are supported. Sync markers and Snappy checksums are checked. Unsupported codecs or invalid containers show an error; standalone binary records without a container are not supported.

Can I export Avro records as JSON or CSV?

Yes. The decoded record table exports all matching applied records across pages as JSON or CSV, limited to 10 MB; unapplied drafts are excluded. Download a copy includes all applied records, including filtered-out records. Large-file page exports contain only the displayed saved-source records and exclude staged edits. Copies leave source edits unsaved. You can also export the embedded writer schema as an .avsc file below 1 MB.