Reuse HTTPS connections, negotiate HTTP/2, add new response fields

.github/workflows/python-publish.yml

@@ -6,11 +6,36 @@ on:
       - master
     tags:
       - '*'
+  pull_request:
 
 jobs:
+  test:
+    name: Test (Python ${{ matrix.python-version }})
+    # ubuntu-latest now resolves to Ubuntu 26.04, for which actions/setup-python has no Python 3.9 builds
+    runs-on: ubuntu-24.04
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ["3.9", "3.10", "3.11", "3.12", "3.13"]
+    steps:
+    - uses: actions/checkout@v4
+    - name: Set up Python ${{ matrix.python-version }}
+      uses: actions/setup-python@v5
+      with:
+        python-version: ${{ matrix.python-version }}
+    - name: Install dependencies
+      run: |
+        python -m pip install --upgrade pip
+        python -m pip install -r requirements.txt
+    - name: Test with pytest (offline tests only, no API key needed)
+      run: |
+        python -m pytest -m "not api"
+
   build-n-publish:
     name: Build and publish to PyPI
-    runs-on: ubuntu-latest
+    if: github.event_name == 'push'
+    # the pinned Python 3.10.4 below has no build for Ubuntu 26.04 either
+    runs-on: ubuntu-24.04
     steps:
     - uses: actions/checkout@v4
     - name: Set up Python 3.10.4

README.md

@@ -21,6 +21,7 @@ Visit our [Documentation](https://docs.ipdata.co/) for more examples and tutoria
   - [Getting only one field](#getting-only-one-field)
   - [Getting a number of specific fields](#getting-a-number-of-specific-fields)
   - [Bulk Lookups](#bulk-lookups)
+  - [Connection reuse and HTTP/2](#connection-reuse-and-http2)
 - [Using the ipdata CLI](#using-the-ipdata-cli)
   - [Windows Installation Notes](#windows-installation-notes)
   - [Available commands](#available-commands)
@@ -367,6 +368,24 @@ NOTE: Alternatively it is much simpler to process bulk lookups using the `ipdata
 ```
 </details>
 
+### Connection reuse and HTTP/2
+
+An `IPData` client keeps a persistent HTTPS connection pool for as long as it lives. Create one client per process (or one per API key or endpoint), keep it, and reuse it for every lookup: each new client pays a fresh TLS handshake. Requests to api.ipdata.co are multiplexed over HTTP/2 when the platform negotiates it, and fall back to HTTP/1.1 keep-alive otherwise.
+
+The client is built on [httpx](https://www.python-httpx.org/) and is thread-safe: one instance can serve every thread of a process at once (the CLI's `batch` command shares a single client across its thread pool). Idle connections are kept open for 60 seconds and up to 20 connections are pooled. Call `close()` at shutdown, or use the client as a context manager, to release the pool. Pass `http2=False` to stay on HTTP/1.1 keep-alive, or `transport=` to inject your own `httpx.BaseTransport` (for example an `httpx.MockTransport` in tests).
+
+```python
+from ipdata import IPData
+
+with IPData("<YOUR API KEY>") as client:  # one client, many lookups
+    for ip in ["8.8.8.8", "1.1.1.1"]:
+        print(client.lookup(ip).country_name)
+```
+
+The module-level helpers `ipdata.lookup()` and `ipdata.bulk()` share one process-wide client, `ipdata.default_client`, which is created on first use and lives until the process exits. Connections must not be shared across `os.fork()` or `multiprocessing` boundaries: create your `IPData` instance (or make the first module-level call) in the child process, after forking.
+
+Failed requests are retried up to `retry_limit` times (default 7). GET and POST requests that fail with a transport error (connection reset, timeout) or with HTTP 500, 502, 503 or 504 are retried with exponential backoff: 0.5, 1, 2, 4, 8, 16 and 32 seconds for the default `retry_backoff_factor=1`. A 429 response is retried only when it carries a `Retry-After` header (waiting at most 60 seconds); without one it is returned immediately, so an exhausted quota fails fast. The `HTTPS_PROXY` and `NO_PROXY` environment variables are honoured; a custom CA bundle is read from `SSL_CERT_FILE` (httpx does not read `REQUESTS_CA_BUNDLE`).
+
 ## Using the ipdata CLI
 
 
@@ -546,6 +565,18 @@ The `--fields` option is required to use CSV output.
 
 A list of all the fields returned by the API is maintained at [Response Fields](https://docs.ipdata.co/api-reference/response-fields)
 
+The following fields were added to the API in November 2026. The library passes them through unchanged, and both top-level flags are accepted by the `fields` and `select_field` parameters of `lookup()` and `bulk()` and by the CLI's `-f` and `-e` options.
+
+| Field | Description |
+|-------|-------------|
+| `is_satellite` | `true` when the address belongs to a satellite internet provider (for example Starlink). Present only when true, omitted otherwise. |
+| `is_anycast` | `true` when the address is an anycast address announced from several locations at once (for example `8.8.8.8`). Present only when true, omitted otherwise. |
+| `threat.blocklists[].first_seen` | The UTC calendar date (a `YYYY-MM-DD` string) on which that blocklist first listed the address. Optional: some feeds, for example ipdata's own vpn feed, do not provide it. |
+| `threat.blocklists[].last_seen` | The UTC calendar date (a `YYYY-MM-DD` string) on which that blocklist last listed the address. Optional, as above. |
+| `threat.blocklists[].days_seen` | The number of days the address has been listed by that blocklist (an integer). Optional, as above. |
+
+`is_satellite` and `is_anycast` are included in a response only when they are `true`. When absent they are missing from the dictionary and read as `None` through dot notation (`response.is_satellite`), so treat anything falsy as `false`. The dates are kept as strings and are not parsed. Fields the library does not know about yet are passed through as well.
+
 ## Geofeed tools
 
 Geofeed publishers can use the `ipdata validate` command to validate their geofeeds before submission to ipdata. This will catch most but not all issues that might cause processing your geofeed to fail.
@@ -754,3 +785,9 @@ To run all tests
 ```shell
 pytest
 ```
+
+The tests marked `api` call the live API and need an `IPDATA_API_KEY` environment variable (or a `src/ipdata/.env` file). To run only the offline tests, as CI does for pull requests:
+
+```shell
+pytest -m "not api"
+```

requirements.txt

@@ -1,4 +1,4 @@
-requests
+httpx[http2]>=0.28,<1
 rich
 click
 click_default_group

setup.cfg

@@ -1,6 +1,6 @@
 [metadata]
 name = ipdata
-version = 4.0.9
+version = 4.1.0
 author = Jonathan Kosgei
 author_email = jonathan@ipdata.co
 description = This is the official IPData client library for Python
@@ -24,7 +24,7 @@ package_dir =
 packages = find:
 python_requires = >=3.9
 install_requires =
-    requests
+    httpx[http2]>=0.28,<1
     rich
     click
     click_default_group

src/ipdata/geofeeds.py

@@ -8,7 +8,7 @@
 import random
 from pathlib import Path
 
-import requests
+import httpx
 
 from .codes import COUNTRIES, REGION_CODES
 
@@ -49,7 +49,7 @@ def _download(self):
         cache_path = f"{self.dir}/{random_name}.csv"
 
         try:
-            response = requests.get(self.source, timeout=60)
+            response = httpx.get(self.source, timeout=60, follow_redirects=True)
             text = response.text
             with open(cache_path, "w") as f:
                 f.write(