Summary
A valid reddit_session cookie works against Reddit JSON endpoints with urllib and curl, but the native rdt-cli httpx transport gets a Reddit 403 HTML page for the same endpoints.
This appears distinct from #11 (cookie discovery) and #12 (subreddit listing HTTP 500). In this case the cookie is present and valid; the failure is specific to the httpx request path / transport fingerprint.
Environment
rdt-cli: 0.4.2, installed from GitHub main via uv tool- Python: 3.13.7
httpx: 0.28.1- OS: Linux aarch64 under OpenClaw/OrbStack sandbox
- Auth: valid browser
reddit_session cookie copied into ~/.config/rdt-cli/credential.json
- Account:
botisaaajiao
No raw cookie value is included here.
Credential sanity checks
The copied reddit_session value is a normal HTTP cookie string:
cookie_length: 729
ascii_printable: true
whitespace_count: 0
The same cookie succeeds when sent via non-httpx transports.
What works
Using the same reddit_session cookie and the same rdt_cli.fingerprint.BrowserFingerprint.chrome133_mac().read_headers() headers, urllib returns 200 JSON:
urllib /api/me.json?raw_json=1 200 application/json; charset=UTF-8
urllib /subreddits/mine/subscriber.json?raw_json=1 200 application/json; charset=UTF-8
urllib /.json?raw_json=1 200 application/json; charset=UTF-8
A local wrapper fallback using urllib can authenticate and read feed data successfully:
{
"ok": true,
"schema_version": "1",
"data": {
"authenticated": true,
"cookie_count": 1,
"username": "botisaaajiao",
"capabilities": ["read"],
"modhash_present": true,
"error": null
}
}What fails
Native rdt-cli / httpx returns 403 HTML with the same cookie/header setup:
httpx /api/me.json?raw_json=1 403 text/html '<body class=theme-beta><div><style>.them'
httpx /subreddits/mine/subscriber.json?raw_json=1 403 text/html '<body class=theme-beta><div><style>.them'
httpx /.json?raw_json=1 403 text/html '<body class=theme-beta><div><style>.them'
Running the raw CLI with the workspace credential also reports forbidden rather than authenticated:
{
"ok": true,
"schema_version": "1",
"data": {
"authenticated": false,
"cookie_count": 1,
"source": "mac-exec:Chrome:Default:host-hash-stripped",
"username": "botisaaajiao",
"capabilities": ["read"],
"modhash_present": false,
"last_verified_at": null,
"error": "Access forbidden: Resource"
}
}Minimal reproduction shape
The key comparison is:
import json
import pathlib
import urllib.request
import httpx
from rdt_cli.fingerprint import BrowserFingerprint
headers = BrowserFingerprint.chrome133_mac().read_headers()
cred = json.loads(pathlib.Path("~/.config/rdt-cli/credential.json").expanduser().read_text())
session = cred["cookies"]["reddit_session"]
url = "https://www.reddit.com/api/me.json?raw_json=1"
req = urllib.request.Request(url, headers={**headers, "Cookie": f"reddit_session={session}"})
with urllib.request.urlopen(req, timeout=20) as resp:
print("urllib", resp.status, resp.headers.get("content-type"))
with httpx.Client(headers=headers, cookies={"reddit_session": session}, follow_redirects=True, timeout=20) as client:
r = client.get(url)
print("httpx", r.status_code, r.headers.get("content-type"), repr(r.text[:80]))Observed result:
urllib 200 application/json; charset=UTF-8
httpx 403 text/html '<body class=theme-beta><div><style>...'
Expected behavior
If the cookie is valid and Reddit returns 200 for the same endpoint via urllib/curl, rdt-cli should either:
- make its
httpx transport match a request fingerprint Reddit accepts, or
- provide a fallback transport for auth/feed reads, or
- surface this as a structured transport/fingerprint error instead of looking like an invalid credential.
Local workaround
For now I patched my workspace wrapper narrowly so rdt status and rdt feed --subs-only --compact fall back to urllib when raw rdt-cli reports forbidden. That restores my automation, but the upstream httpx path still fails.
Summary
A valid
reddit_sessioncookie works against Reddit JSON endpoints withurllibandcurl, but the nativerdt-clihttpxtransport gets a Reddit 403 HTML page for the same endpoints.This appears distinct from #11 (cookie discovery) and #12 (subreddit listing HTTP 500). In this case the cookie is present and valid; the failure is specific to the
httpxrequest path / transport fingerprint.Environment
rdt-cli: 0.4.2, installed from GitHub main viauv toolhttpx: 0.28.1reddit_sessioncookie copied into~/.config/rdt-cli/credential.jsonbotisaaajiaoNo raw cookie value is included here.
Credential sanity checks
The copied
reddit_sessionvalue is a normal HTTP cookie string:The same cookie succeeds when sent via non-
httpxtransports.What works
Using the same
reddit_sessioncookie and the samerdt_cli.fingerprint.BrowserFingerprint.chrome133_mac().read_headers()headers,urllibreturns 200 JSON:A local wrapper fallback using
urllibcan authenticate and read feed data successfully:{ "ok": true, "schema_version": "1", "data": { "authenticated": true, "cookie_count": 1, "username": "botisaaajiao", "capabilities": ["read"], "modhash_present": true, "error": null } }What fails
Native
rdt-cli/httpxreturns 403 HTML with the same cookie/header setup:Running the raw CLI with the workspace credential also reports forbidden rather than authenticated:
{ "ok": true, "schema_version": "1", "data": { "authenticated": false, "cookie_count": 1, "source": "mac-exec:Chrome:Default:host-hash-stripped", "username": "botisaaajiao", "capabilities": ["read"], "modhash_present": false, "last_verified_at": null, "error": "Access forbidden: Resource" } }Minimal reproduction shape
The key comparison is:
Observed result:
Expected behavior
If the cookie is valid and Reddit returns 200 for the same endpoint via
urllib/curl,rdt-clishould either:httpxtransport match a request fingerprint Reddit accepts, orLocal workaround
For now I patched my workspace wrapper narrowly so
rdt statusandrdt feed --subs-only --compactfall back tourllibwhen rawrdt-clireports forbidden. That restores my automation, but the upstreamhttpxpath still fails.