feat(cag): memory standing-context adapter, warmup semaphore, cache-aware routing

Issue #44: code-register memory adapter with standing-context resource; fix
dispatcher method-level Auth.Capability override (was always using verb-derived
capability, ignoring per-method Auth field).

Issue #45: per-backend bounded in-memory queue (VLLM_MAX_QUEUE_DEPTH=20) +
concurrency semaphore (VLLM_MAX_NUM_SEQS=4); warmup requests (X-Poimen-Warmup:
true) sleep WARMUP_DELAY_MS before queue entry; /metrics endpoint wired.

Issue #46: rendezvous-hash backend selection on X-Poimen-Cache-Key; round-robin
fallback; MarkFailed excludes unhealthy backends 30s; ModifyResponse captures
vllm-num-cached-tokens + emits structured log + increments counter.

Closes #44 Closes #45 Closes #46
This commit is contained in:
Admin Bot
2026-09-27 18:21:41 +08:00
parent ea7bd8336b
commit fc5c0e967d
16 changed files with 861 additions and 17 deletions
+137 -1
View File
@@ -161,7 +161,7 @@ models:
if err == nil {
t.Errorf("expected error for missing address, got nil")
}
if err.Error() != "model \"reasoning\": field 'address' is required" {
if err.Error() != "model \"reasoning\": 'address' or 'addresses' is required" {
t.Errorf("expected missing address error, got: %v", err)
}
}
@@ -283,3 +283,139 @@ models:
t.Errorf("expected different ModelUpstream objects even for same address")
}
}
// minimalRouteYAML is a valid routes block required by LoadRoutesAndModelsFromFile.
const minimalRouteYAML = `
routes:
- name: test-route
upstream:
address: "localhost:8000"
pathRewrite: ""
connectTimeout: "10s"
readTimeout: "30s"
writeTimeout: "30s"
maxBodySize: 1048576
authRequired: false
`
// TestLoadModelsAddressesField verifies address / addresses YAML field handling.
func TestLoadModelsAddressesField(t *testing.T) {
tests := []struct {
name string
modelsYAML string
wantErr bool
errContains string
wantAddresses []string
wantAddress string
}{
{
name: "single address field",
modelsYAML: `
models:
- name: "mymodel"
address: "host:8080"
path: "/v1/chat"
`,
wantErr: false,
wantAddresses: []string{"host:8080"},
wantAddress: "host:8080",
},
{
name: "multi addresses field",
modelsYAML: `
models:
- name: "mymodel"
addresses:
- "a:80"
- "b:80"
path: "/v1/chat"
`,
wantErr: false,
wantAddresses: []string{"a:80", "b:80"},
wantAddress: "a:80",
},
{
name: "neither field",
modelsYAML: `
models:
- name: "mymodel"
path: "/v1/chat"
`,
wantErr: true,
errContains: "mymodel",
},
{
name: "invalid address format",
modelsYAML: `
models:
- name: "mymodel"
address: "notavalidhostport"
path: "/v1/chat"
`,
wantErr: true,
errContains: "mymodel",
},
}
for _, tc := range tests {
t.Run(tc.name, func(t *testing.T) {
data := minimalRouteYAML + tc.modelsYAML
tmpFile, err := os.CreateTemp("", "config-*.yaml")
if err != nil {
t.Fatalf("CreateTemp: %v", err)
}
defer os.Remove(tmpFile.Name())
tmpFile.WriteString(data)
tmpFile.Close()
_, models, _, _, loadErr := LoadRoutesAndModelsFromFile(tmpFile.Name())
if tc.wantErr {
if loadErr == nil {
t.Fatalf("expected error, got nil")
}
if tc.errContains != "" && !containsStr(loadErr.Error(), tc.errContains) {
t.Errorf("error %q does not contain %q", loadErr.Error(), tc.errContains)
}
return
}
if loadErr != nil {
t.Fatalf("unexpected error: %v", loadErr)
}
m, ok := models["mymodel"]
if !ok {
t.Fatalf("model 'mymodel' not found in result")
}
if m.Address != tc.wantAddress {
t.Errorf("Address: want %q, got %q", tc.wantAddress, m.Address)
}
if len(m.Addresses) != len(tc.wantAddresses) {
t.Errorf("Addresses len: want %d, got %d (%v)", len(tc.wantAddresses), len(m.Addresses), m.Addresses)
} else {
for i, a := range tc.wantAddresses {
if m.Addresses[i] != a {
t.Errorf("Addresses[%d]: want %q, got %q", i, a, m.Addresses[i])
}
}
}
})
}
}
// containsStr is a helper to avoid importing strings just for Contains.
func containsStr(s, sub string) bool {
return len(s) >= len(sub) && (s == sub || len(sub) == 0 ||
func() bool {
for i := 0; i <= len(s)-len(sub); i++ {
if s[i:i+len(sub)] == sub {
return true
}
}
return false
}())
}