Copilot commented on code in PR #13316:
URL: https://github.com/apache/apisix/pull/13316#discussion_r3159049326
##########
t/plugin/ai-rate-limiting.t:
##########
@@ -1470,3 +1470,71 @@ X-AI-Fixture: openai/chat-model-echo.json
X-AI-Project-RateLimit-Limit: 30
X-AI-Project-RateLimit-Remaining: 30
X-AI-Project-RateLimit-Reset: 10
+
+
+
+=== TEST 33: setup route with instance name containing dots
+--- config
+ location /t {
+ content_by_lua_block {
+ local t = require("lib.test_admin").test
+ local code, body = t('/apisix/admin/routes/1',
+ ngx.HTTP_PUT,
+ [[{
+ "uri": "/ai",
+ "plugins": {
+ "ai-proxy-multi": {
+ "fallback_strategy":
"instance_health_and_rate_limiting",
+ "instances": [
+ {
+ "name": "Qwen3.5-397B-10.249.238.157",
+ "provider": "openai",
+ "weight": 1,
+ "priority": 1,
+ "auth": {"header": {"Authorization":
"Bearer token"}},
+ "options": {"model": "gpt-4"},
+ "override": {"endpoint":
"http://127.0.0.1:1980"}
+ }
+ ],
+ "ssl_verify": false
+ },
+ "ai-rate-limiting": {
+ "instances": [
+ {
+ "name": "Qwen3.5-397B-10.249.238.157",
+ "limit": 100,
+ "time_window": 60
+ }
+ ]
+ }
+ },
+ "upstream": {
+ "type": "roundrobin",
+ "nodes": {
+ "canbeanything.com": 1
+ }
+ }
+ }]]
+ )
+
+ if code >= 300 then
+ ngx.status = code
+ end
+ ngx.say(body)
+ }
+ }
+--- response_body
+passed
+
+
+
+=== TEST 34: instance name with dots should not trigger ctx.var resolution
+--- request
+POST /ai
+{"messages":[{"role":"system","content":"You are a
mathematician"},{"role":"user","content":"What is 1+1?"}]}
+--- more_headers
+Authorization: Bearer token
+X-AI-Fixture: openai/chat-model-echo.json
+--- error_code: 200
+--- no_error_log
+[error]
Review Comment:
TEST 34 only asserts HTTP 200 and absence of `[error]` logs. A regression
where the instance name is treated as a `ctx.var[...]` key (instead of a
constant) wouldn’t necessarily log an error: `limit-count` would resolve
`ctx.var["Qwen3.5-..."]` to nil and silently fall back to `remote_addr`,
changing semantics to per-client rate limiting. To actually cover the reported
bug, add an assertion that the rate-limit key is independent of client IP
(e.g., make two requests with different `X-Real-IP` values after setting a low
enough `limit` so the first request exhausts the quota; the second request
should still be rejected if the key is truly constant).
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
