fastreq 3.2.0

High-performance async HTTP client built on niquests (default) and httpx (optional).

Features

• Two transports: niquests (default, required) and httpx (optional extra)
• HTTP/2 support via niquests/qh3 (always) and httpx+hyper (when h2 installed)
• Explicit proxy pools with health tracking, cooldown, round-robin rotation, and Webshare import — no free-proxy discovery
• Token-bucket rate limiting — rate token acquired before the concurrency slot
• Typed retry — only transient transport errors and configured retryable status codes (429, 500, 502, 503, 504); honors Retry-After
• User-agent rotation from a built-in list or custom list
• Cookie management with session-scoped, proxy-isolated storage
• Chunked streaming via callbacks in both transports
• Flexible response parsing — JSON, text, content, response object, or stream
• Keyed responses for dict-style result mapping
• Optional progress reporting via Rich, tqdm, or a headless callback

Installation

# Core library (niquests is the default and required transport)
pip install fastreq

# With HTTPX as an alternative transport
pip install fastreq[httpx]

# With HTTP/2 support for httpx (niquests supports HTTP/2 natively)
pip install fastreq[httpx,h2]

# Optional terminal progress bars
pip install 'fastreq[progress-rich]'
pip install 'fastreq[progress-tqdm]'

Quick Start

Sync Usage

from fastreq import fastreq

result = fastreq("https://api.github.com/repos/python/cpython")
print(result)

Async Usage

import asyncio
from fastreq import fastreq_async

async def main():
    results = await fastreq_async([
        "https://api.github.com/repos/python/cpython",
        "https://api.github.com/repos/python/cpython/issues",
    ], concurrency=5)
    return results

results = asyncio.run(main())

Context Manager

from fastreq import FastRequests

async def main():
    async with FastRequests(concurrency=5) as client:
        result = await client.request("https://api.github.com/repos/python/cpython")
    return result

Optional Progress

Progress reporting is disabled by default and adds no core dependency. Choose Rich or tqdm explicitly, or use True to select an installed backend:

async with FastRequests(concurrency=5) as client:
    results = await client.request(urls, progress="rich")

# Headless/CI-friendly progress callback
results = await fastreq_async(
    urls,
    progress_callback=lambda completed, total: print(f"{completed}/{total}"),
)

Streaming

from fastreq import FastRequests, ReturnType

async def main():
    async with FastRequests() as client:
        await client.request(
            "https://example.com/large-file.zip",
            return_type=ReturnType.STREAM,
            stream_callback=lambda chunk: print(f"Got {len(chunk)} bytes"),
        )

Explicit Proxy Rotation

from fastreq import FastRequests

# Provide a proxy list directly
proxies = [
    "http://user:pass@proxy1:8080",
    "http://user:pass@proxy2:8080",
]

async with FastRequests(
    proxies=proxies,
    random_proxy=True,
    proxy_selection="round_robin",
    proxy_cooldown=60.0,
) as client:
    result = await client.request("https://api.example.com/data")

You can also set proxies via the FASTREQ_PROXIES environment variable (comma-separated) or import from a Webshare text file via webshare_file=.

Backend Selection

backend="auto" (the default) always selects niquests. To use another backend explicitly:

FastRequests(backend="httpx")

Browser impersonation (curl_cffi backend)

The curl_cffi backend can replicate a real browser's TLS fingerprint (JA3/JA4), HTTP/2 settings, and default headers via the impersonate parameter. This defeats bot detection that blocks plain HTTP clients:

async with FastRequests(backend="curl_cffi", impersonate="chrome") as client:
    result = await client.request("https://example.com/bot-protected")

# or a random recent browser target per client
FastRequests(backend="curl_cffi", impersonate="random")

When impersonate is set, user-agent rotation is disabled automatically — curl_cffi supplies the browser-matching User-Agent itself. One session is kept per proxy route, so cookies stay isolated between proxies.

Migration from 2.x to 3.0

Breaking changes

• Removed backends: aiohttp and requests transports are gone. Use niquests (default) or httpx.
• Removed free proxies: The library no longer discovers or fetches free proxies. Provide explicit proxies via proxies=, FASTREQ_PROXIES, or webshare_file=.
• Canonical name: The project is now fastreq (previously parallel-requests). ParallelRequests remains as a Python alias for backward compatibility.

What to do

1. If you used backend="aiohttp" or backend="requests", switch to the default niquests transport or install fastreq[httpx] and use backend="httpx".
2. If you used free_proxies=True, remove it and provide an explicit proxy list or Webshare source instead.
3. If you imported ParallelRequests, it still works as an alias — but the canonical name is FastRequests.

Proxy Policy

fastreq 3.0 does not fetch free proxies. Proxy sources in priority order:

1. proxies= constructor argument (list of URLs)
2. FASTREQ_PROXIES environment variable (comma-separated)
3. webshare_file= path to a Webshare plain-text proxy file (ip:port:user:password per line)

Failed proxies enter a configurable cooldown. Successful requests clear the cooldown. Selection strategies: round_robin (default) or random.

Documentation

• Full Documentation
• Examples

Development

This project uses uv for dependency management and packaging.

# Install all dependencies
uv sync --all-groups

# Run quality gates
uv run ruff format --check .
uv run ruff check .
uv run ty check src
uv run pytest -q

# Build
uv build

License

MIT License — see LICENSE.