neudata-api-client

A Python client for the Neudata Developer API. It lets you search and retrieve dataset metadata, additional datasets, filters, and DDQs — and ships with resilient defaults: it retries dropped connections, backs off and honours Retry-After on rate limits (429), and auto-paginates list endpoints.

Installation

pip install neudata-api-client

For development (using uv):

git clone https://github.com/neudata/neudata-api-client
cd neudata-api-client
uv sync

Authentication

The API authenticates with two headers, Email and Api-Token. Your token is available at https://www.neudata.co/users/edit.

Provide them via environment variables (recommended):

export NEUDATA_EMAIL="you@example.com"
export NEUDATA_API_TOKEN="your-api-token"

…or pass them directly to the client:

from neudata import NeudataClient

client = NeudataClient(email="you@example.com", api_token="your-api-token")

A .env.example is included as a template.

Quickstart

from neudata import NeudataClient

with NeudataClient() as client:
    # Search a single page of datasets
    results = client.search_datasets(q="satellite", per_page=10)

    # Or iterate over *every* matching dataset (pagination handled for you)
    for dataset in client.iter_datasets(equity_region=["uk", "europe"]):
        print(dataset["id"], dataset["title"])

    # Fetch the full detail record for one dataset
    detail = client.get_dataset(1234)

API reference

Method	Endpoint	Description
search_datasets(**filters)	GET /datasets	One page of dataset search results
iter_datasets(per_page=99, **filters)	GET /datasets	Iterate all datasets across pages
get_dataset(id)	GET /datasets/{id}	Full dataset detail
list_filters()	GET /filters	Available filter parameter names
get_filter_options(name)	GET /filters/{name}	Allowable options for a filter
search_additional_datasets(**filters)	GET /additional-datasets	One page of additional datasets
iter_additional_datasets(per_page=99, **filters)	GET /additional-datasets	Iterate all additional datasets
get_additional_dataset(id)	GET /additional-datasets/{id}	Full additional-dataset detail
list_ddqs(page, per_page)	GET /ddqs	List DDQs (Sentry subscribers)
iter_ddqs(per_page=99)	GET /ddqs	Iterate all DDQs
get_ddq(id)	GET /ddqs/{id}	DDQ detail + 10-minute download URL

Filters

Search methods accept the documented query parameters as keyword arguments, e.g. q, page, per_page, sort, sort_by, source, from_date, data_provider_id, and the array filters equity_region, asset_class, combined_dataset_types, frequency, history, subscription_price, report_type, feed_type, output_format, esg_topics, filter_labels, ticker_indices, and more.

Array filters are passed as Python lists and serialized into the repeated name[] form the API expects:

client.search_datasets(asset_class=["public-equity", "fixed-income"])
# -> GET /datasets?asset_class[]=public-equity&asset_class[]=fixed-income

Use client.list_filters() and client.get_filter_options(name) to discover the valid values for each filter at runtime.

Resilience: retries, backoff & rate limits

The client is configured on construction:

NeudataClient(
    timeout=30,                  # per-request timeout (seconds)
    max_retries=5,               # retries for dropped connections / 5xx
    backoff_factor=0.5,          # exponential backoff base
    max_backoff=60,              # cap on any single backoff sleep
    max_rate_limit_retries=10,   # retries specifically for 429 responses
    requests_per_minute=90,      # client-side throttle (0 disables it)
)

• Dropped/closed connections (and timeouts and 5xx server errors) are retried up to max_retries times with jittered exponential backoff.
• 429 rate limits are retried up to max_rate_limit_retries times. The Retry-After header is honoured when present; otherwise exponential backoff is used. When retries are exhausted a RateLimitError is raised.
• A built-in client-side throttle keeps the sustained request rate below requests_per_minute (the API guideline is ~100/min) so bulk jobs avoid tripping the limiter in the first place.

Enable logging to see retry/backoff activity:

import logging
logging.getLogger("neudata").setLevel(logging.INFO)

Exceptions

All exceptions derive from neudata.NeudataError:

• AuthenticationError — 401/403 (bad credentials; not retried)
• NotFoundError — 404
• RateLimitError — 429 retries exhausted (carries retry_after)
• NeudataAPIError — other unexpected API responses (carries status_code, response_body)

Example: download the whole catalogue

A console script is installed with the package:

# Dump all datasets + additional datasets to neudata_export.json
neudata-download

# Custom output, only datasets, with full per-item detail records
neudata-download --output datasets.json --datasets-only --details

# Filter by date or free text
neudata-download --from-date 2026-01-01 --query "credit card"

Run neudata-download --help for all options. A minimal, readable version using the library directly is in examples/download_all.py:

python examples/download_all.py out.json

The output JSON has the shape:

{
  "datasets": [ ... ],
  "additional_datasets": [ ... ]
}

Development

uv sync             # create the venv and install deps + dev tools
uv run pytest       # run the test suite (no network calls)

Tests use responses to mock the API and a monkeypatched sleep, so they run instantly and offline.

There is also an opt-in live smoke test that hits every endpoint against the real API. It is excluded from the default run and requires credentials:

export NEUDATA_EMAIL="you@example.com"
export NEUDATA_API_TOKEN="your-api-token"
uv run pytest -m smoke -v      # hits the real Neudata API

(The DDQ checks skip automatically if your account isn't a Sentry subscriber.)

Releasing to PyPI

uv build                          # build sdist + wheel into dist/
uvx twine check dist/*            # validate metadata / long description

# Test upload first
uv publish --publish-url https://test.pypi.org/legacy/ --token "$TEST_PYPI_TOKEN"

# Then the real thing
uv publish --token "$PYPI_TOKEN"

License

MIT