dark-req

A requests-compatible Python HTTP client that fixes everything requests gets wrong — in the same lightweight footprint (only urllib3 + certifi).

import dark_req

# Works exactly like requests...
r = dark_req.get("https://api.example.com/users", timeout=10)
print(r.status_code, r.json())

# ...but with superpowers baked in
s = dark_req.Session(
    base_url="https://api.example.com",
    max_retries=3,          # auto-retry on 429/5xx + connection errors
    rate_limit=5,           # max 5 req/sec per host (token bucket)
    hooks={"on_error": [lambda e: print("failed:", e)]},
)
r = s.get("/users", params={"page": 2})   # -> https://api.example.com/users?page=2
r.raise_for_status()

# Downloads with progress, one line
s.download("https://example.com/big.zip", "big.zip",
           progress=lambda done, total: print(f"{done}/{total}"))

# Async, same API
import asyncio
async def main():
    async with dark_req.AsyncSession(base_url="https://api.example.com") as s:
        users, repos = await asyncio.gather(s.get("/users"), s.get("/repos"))
asyncio.run(main())

Install

pip install dark-req

Requires Python 3.9+.

Why not just requests?

Everything else stays familiar: get/post/put/patch/delete/head/options, Session, Request/PreparedRequest/Response, HTTPBasicAuth, and an exception hierarchy mirroring requests (so existing except blocks keep working).

Sane defaults

dark_req.get("https://slow.example.com")   # gives up after 5s connect / 30s read
dark_req.get("https://flaky.example.com")  # 3 automatic retries with backoff

Override per call with timeout=... (seconds, (connect, read) tuple, or timeout=None to wait indefinitely), max_retries=0 to disable retries, verify=False to skip TLS verification.

Roadmap

• files= multipart uploads
• Proxy support from environment variables
• HTTP/2 transport option
• Connection-level metrics / tracing hooks

License

MIT