Compare commits
13
Commits
| Author | SHA1 | Date | |
|---|---|---|---|
|
|
b4c3027e6d | ||
|
|
503fe84074 | ||
|
|
3b551b2918 | ||
|
|
51427a246e | ||
|
|
9b5ce2a850 | ||
|
|
8f3920519b | ||
|
|
2277f8a699 | ||
|
|
205d8f6206 | ||
|
|
ae06d33ccc | ||
|
|
c7b57daaf8 | ||
|
|
8aa0123047 | ||
|
|
4a3d348e39 | ||
|
|
4086b75224 |
@@ -18,6 +18,7 @@ type fleetEndpoint struct {
|
||||
Tier capability.FleetTier
|
||||
URL string
|
||||
Headers map[string]string
|
||||
Client *http.Client
|
||||
}
|
||||
|
||||
type fleetStatusResponse struct {
|
||||
@@ -41,7 +42,11 @@ func fleetReadinessHandler(client *http.Client, preferred capability.FleetTier,
|
||||
for name, value := range endpoint.Headers {
|
||||
request.Header.Set(name, value)
|
||||
}
|
||||
response, requestErr := client.Do(request)
|
||||
probeClient := endpoint.Client
|
||||
if probeClient == nil {
|
||||
probeClient = client
|
||||
}
|
||||
response, requestErr := probeClient.Do(request)
|
||||
if requestErr == nil {
|
||||
response.Body.Close()
|
||||
if response.StatusCode >= 200 && response.StatusCode < 300 {
|
||||
|
||||
@@ -45,7 +45,7 @@ func main() {
|
||||
discoveries := map[string]providerRegistration{}
|
||||
var builtinLocal *providerRegistration
|
||||
externalRecovery := []providerRegistration{}
|
||||
cpuAPIKey, err := loadLocalCPUProviderAPIKey()
|
||||
cpuAPIKey, err := loadProviderAPIKey("LOCAL_CPU_PROVIDER_API_KEY", "LOCAL_CPU_PROVIDER_API_KEY_FILE", "LOCAL_CPU_PROVIDER_CREDENTIAL_REF", "credential://domovoy/local-cpu/provider-api-key/v1")
|
||||
if err != nil {
|
||||
log.Fatal(err)
|
||||
}
|
||||
@@ -62,6 +62,27 @@ func main() {
|
||||
cpu := providerRegistration{ProviderInstanceID: "local-cpu-primary", Kind: "openai-compatible-chat-v1", EndpointRef: "local-cpu-primary-v1", CredentialRef: cpuCredentialRef, AuthMode: cpuAuthMode, BillingMode: capability.BillingFree, DataPolicy: capability.DataPolicyLocalOnly, Provider: cpuAdapter, Discovery: cpuAdapter}
|
||||
providers[cpu.Kind] = append(providers[cpu.Kind], cpu)
|
||||
discoveries["cpu"] = cpu
|
||||
mediumURL := strings.TrimSpace(os.Getenv("LOCAL_MEDIUM_PROVIDER_URL"))
|
||||
mediumAPIKey := ""
|
||||
if mediumURL != "" {
|
||||
mediumAPIKey, err = loadProviderAPIKey("LOCAL_MEDIUM_PROVIDER_API_KEY", "LOCAL_MEDIUM_PROVIDER_API_KEY_FILE", "LOCAL_MEDIUM_PROVIDER_CREDENTIAL_REF", "credential://domovoy/local-medium/provider-api-key/v1")
|
||||
if err != nil {
|
||||
log.Fatal(err)
|
||||
}
|
||||
mediumAdapter, adapterErr := models.NewOpenAICompatibleLocalAdapterWithAPIKey(strings.TrimRight(mediumURL, "/"), mediumAPIKey, nil)
|
||||
if adapterErr != nil {
|
||||
log.Fatal(adapterErr)
|
||||
}
|
||||
mediumAuthMode := capability.AuthNone
|
||||
mediumCredentialRef := capability.CredentialRef("")
|
||||
if mediumAPIKey != "" {
|
||||
mediumAuthMode = capability.AuthKey
|
||||
mediumCredentialRef = capability.CredentialRef(envOr("LOCAL_MEDIUM_PROVIDER_CREDENTIAL_REF", "credential://domovoy/local-medium/provider-api-key/v1"))
|
||||
}
|
||||
medium := providerRegistration{ProviderInstanceID: "local-medium-primary", Kind: "openai-compatible-chat-v1", EndpointRef: "local-medium-primary-v1", CredentialRef: mediumCredentialRef, AuthMode: mediumAuthMode, BillingMode: capability.BillingFree, DataPolicy: capability.DataPolicyLocalOnly, Provider: mediumAdapter, Discovery: mediumAdapter}
|
||||
providers[medium.Kind] = append(providers[medium.Kind], medium)
|
||||
discoveries["medium"] = medium
|
||||
}
|
||||
if lifeboat, enabled, lifeboatErr := newLifeboatProvider(os.Getenv("LOCAL_LIFEBOAT_PROVIDER_URL")); lifeboatErr != nil {
|
||||
log.Fatal(lifeboatErr)
|
||||
} else if enabled {
|
||||
@@ -85,6 +106,8 @@ func main() {
|
||||
gateway := cognitiveGateway{resolver: resolver, providers: providers, builtinLocal: builtinLocal, externalRecovery: externalRecovery, now: time.Now, allowCanary: os.Getenv("COGNITIVE_ALLOW_CANARY") == "true", canaryMaxOutputBytes: 4096, admission: newHTTPAdmissionClient(envOr("CORE_RUNTIME_URL", "http://core-runtime:8080"), &http.Client{Timeout: 5 * time.Second})}
|
||||
mux.Handle("/healthz", health.Handler{Runtime: "cognitive-worker", SemanticGeneration: 1, ArchitectureGeneration: 1})
|
||||
readinessClient := &http.Client{Timeout: 2 * time.Second}
|
||||
cpuReadinessClient := &http.Client{Timeout: 10 * time.Second}
|
||||
mediumReadinessClient := &http.Client{Timeout: 10 * time.Second}
|
||||
lifeboatURL := strings.TrimSuffix(strings.TrimRight(envOr("LOCAL_LIFEBOAT_PROVIDER_URL", "http://cognitive-lifeboat:8086/v1"), "/"), "/v1") + "/health"
|
||||
fleetClient := &http.Client{Timeout: 2 * time.Second}
|
||||
preferredFleet := capability.FleetTier(envOr("COGNITIVE_PREFERRED_FLEET_TIER", string(capability.FleetTierHeavy)))
|
||||
@@ -92,17 +115,35 @@ func main() {
|
||||
if cpuAPIKey != "" {
|
||||
cpuFleetHeaders["Authorization"] = "Bearer " + cpuAPIKey
|
||||
}
|
||||
cpuHealthProbe := health.OptionalHTTPProbeWithHeaders(readinessClient, strings.TrimRight(envOr("LOCAL_CPU_PROVIDER_URL", "http://docker.cin.su:18082/v1"), "/")+"/models", cpuFleetHeaders)
|
||||
cpuEndpoint := strings.TrimRight(envOr("LOCAL_CPU_PROVIDER_URL", "http://docker.cin.su:18082/v1"), "/") + "/models"
|
||||
cpuHealthProbe := health.OptionalHTTPProbeWithHeaders(cpuReadinessClient, cpuEndpoint, cpuFleetHeaders)
|
||||
cpuHealthProbe.Timeout = 10 * time.Second
|
||||
mediumEndpoint := ""
|
||||
mediumHeaders := map[string]string{}
|
||||
mediumHealthProbe := health.Probe{}
|
||||
if mediumURL != "" {
|
||||
mediumEndpoint = strings.TrimRight(mediumURL, "/") + "/models"
|
||||
if mediumAPIKey != "" {
|
||||
mediumHeaders["Authorization"] = "Bearer " + mediumAPIKey
|
||||
}
|
||||
mediumHealthProbe = health.OptionalHTTPProbeWithHeaders(mediumReadinessClient, mediumEndpoint, mediumHeaders)
|
||||
mediumHealthProbe.Timeout = 10 * time.Second
|
||||
}
|
||||
readinessProbes := []health.Probe{
|
||||
health.HTTPProbe(readinessClient, strings.TrimRight(envOr("CORE_RUNTIME_URL", "http://core-runtime:8080"), "/")+"/healthz"),
|
||||
cpuHealthProbe,
|
||||
}
|
||||
if mediumURL != "" {
|
||||
readinessProbes = append(readinessProbes, mediumHealthProbe)
|
||||
}
|
||||
readinessProbes = append(readinessProbes,
|
||||
health.HTTPProbe(readinessClient, lifeboatURL),
|
||||
health.OptionalHTTPProbe(readinessClient, strings.TrimRight(envOr("LOCAL_GPU_PROVIDER_URL", "http://192.168.220.91:11434"), "/")+"/api/tags"),
|
||||
}
|
||||
)
|
||||
mux.Handle("/readyz", health.ReadinessHandler{Runtime: "cognitive-worker", Probes: readinessProbes})
|
||||
mux.Handle("/v1/fleet/readiness", fleetReadinessHandler(fleetClient, preferredFleet, []fleetEndpoint{
|
||||
{ProviderID: "local-cpu-primary", ModelID: "local-cpu-model", Tier: capability.FleetTierFast, URL: strings.TrimRight(envOr("LOCAL_CPU_PROVIDER_URL", "http://docker.cin.su:18082/v1"), "/") + "/models", Headers: cpuFleetHeaders},
|
||||
{ProviderID: "local-medium-unconfigured", ModelID: "medium-unconfigured", Tier: capability.FleetTierMedium},
|
||||
{ProviderID: "local-cpu-primary", ModelID: "local-cpu-model", Tier: capability.FleetTierFast, URL: cpuEndpoint, Headers: cpuFleetHeaders, Client: &http.Client{Timeout: 10 * time.Second}},
|
||||
{ProviderID: "local-medium-primary", ModelID: "local-medium-model", Tier: capability.FleetTierMedium, URL: mediumEndpoint, Headers: mediumHeaders, Client: mediumReadinessClient},
|
||||
{ProviderID: "local-gpu-opportunistic", ModelID: "gpu-opportunistic", Tier: capability.FleetTierHeavy, URL: strings.TrimRight(envOr("LOCAL_GPU_PROVIDER_URL", "http://192.168.220.91:11434"), "/") + "/api/tags"},
|
||||
{ProviderID: "cognitive-lifeboat", ModelID: "qwen3-8b-lifeboat", Tier: capability.FleetTierLifeboat, URL: lifeboatURL},
|
||||
}))
|
||||
@@ -163,17 +204,20 @@ func refreshProviderDiscoveries(resolver bindingResolver, service discoveryServi
|
||||
}
|
||||
}
|
||||
|
||||
func loadLocalCPUProviderAPIKey() (string, error) {
|
||||
if _, set := os.LookupEnv("LOCAL_CPU_PROVIDER_API_KEY"); set {
|
||||
return "", fmt.Errorf("LOCAL_CPU_PROVIDER_API_KEY direct environment secret is forbidden; use LOCAL_CPU_PROVIDER_API_KEY_FILE")
|
||||
func loadProviderAPIKey(directEnvName, fileEnvName, refEnvName, defaultRef string) (string, error) {
|
||||
if _, set := os.LookupEnv(directEnvName); set {
|
||||
return "", fmt.Errorf("%s direct environment secret is forbidden; use %s", directEnvName, fileEnvName)
|
||||
}
|
||||
fileName := strings.TrimSpace(os.Getenv("LOCAL_CPU_PROVIDER_API_KEY_FILE"))
|
||||
fileName := strings.TrimSpace(os.Getenv(fileEnvName))
|
||||
if fileName == "" {
|
||||
return "", nil
|
||||
}
|
||||
ref := strings.TrimSpace(os.Getenv("LOCAL_CPU_PROVIDER_CREDENTIAL_REF"))
|
||||
ref := strings.TrimSpace(os.Getenv(refEnvName))
|
||||
if ref == "" {
|
||||
return "", fmt.Errorf("LOCAL_CPU_PROVIDER_CREDENTIAL_REF is required when LOCAL_CPU_PROVIDER_API_KEY_FILE is configured")
|
||||
ref = defaultRef
|
||||
}
|
||||
if ref == "" {
|
||||
return "", fmt.Errorf("%s is required when %s is configured", refEnvName, fileEnvName)
|
||||
}
|
||||
resolver, err := platformsecrets.NewFileResolver(envOr("SECRET_BACKEND_ROOT", "/run/secrets"))
|
||||
if err != nil {
|
||||
@@ -182,6 +226,10 @@ func loadLocalCPUProviderAPIKey() (string, error) {
|
||||
return resolver.ResolveFile(secretcontracts.SecretFileRef{CredentialRef: secretcontracts.CredentialRef(ref), Purpose: secretcontracts.SecretPurposeProviderAPIKey, Consumer: secretcontracts.RuntimeConsumerCognitive, Path: fileName})
|
||||
}
|
||||
|
||||
func loadLocalCPUProviderAPIKey() (string, error) {
|
||||
return loadProviderAPIKey("LOCAL_CPU_PROVIDER_API_KEY", "LOCAL_CPU_PROVIDER_API_KEY_FILE", "LOCAL_CPU_PROVIDER_CREDENTIAL_REF", "credential://domovoy/local-cpu/provider-api-key/v1")
|
||||
}
|
||||
|
||||
func deterministicAdvisory(w http.ResponseWriter, r *http.Request) {
|
||||
request, ok := decodeAdvisoryRequest(w, r)
|
||||
if !ok {
|
||||
|
||||
@@ -43,3 +43,28 @@ func TestLoadLocalCPUProviderAPIKeyRejectsDirectEnvironmentSecret(t *testing.T)
|
||||
t.Fatal("expected direct environment secret to be rejected")
|
||||
}
|
||||
}
|
||||
|
||||
func TestLoadProviderAPIKeySupportsMediumCredentialRef(t *testing.T) {
|
||||
t.Setenv("LOCAL_MEDIUM_PROVIDER_API_KEY", "")
|
||||
_ = os.Unsetenv("LOCAL_MEDIUM_PROVIDER_API_KEY")
|
||||
root := t.TempDir()
|
||||
path := filepath.Join(root, "medium_provider_api_key")
|
||||
if err := os.WriteFile(path, []byte("medium-key\n"), 0o600); err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
t.Setenv("LOCAL_MEDIUM_PROVIDER_API_KEY_FILE", path)
|
||||
t.Setenv("LOCAL_MEDIUM_PROVIDER_CREDENTIAL_REF", "credential://domovoy/local-medium/provider-api-key/v1")
|
||||
t.Setenv("SECRET_BACKEND_ROOT", root)
|
||||
key, err := loadProviderAPIKey("LOCAL_MEDIUM_PROVIDER_API_KEY", "LOCAL_MEDIUM_PROVIDER_API_KEY_FILE", "LOCAL_MEDIUM_PROVIDER_CREDENTIAL_REF", "credential://domovoy/local-medium/provider-api-key/v1")
|
||||
if err != nil || key != "medium-key" {
|
||||
t.Fatalf("unexpected medium key resolution result: err=%v", err)
|
||||
}
|
||||
}
|
||||
|
||||
func TestLoadProviderAPIKeyRejectsMediumDirectEnvironmentSecret(t *testing.T) {
|
||||
t.Setenv("LOCAL_MEDIUM_PROVIDER_API_KEY", "must-not-be-used")
|
||||
t.Setenv("LOCAL_MEDIUM_PROVIDER_API_KEY_FILE", "")
|
||||
if _, err := loadProviderAPIKey("LOCAL_MEDIUM_PROVIDER_API_KEY", "LOCAL_MEDIUM_PROVIDER_API_KEY_FILE", "LOCAL_MEDIUM_PROVIDER_CREDENTIAL_REF", "credential://domovoy/local-medium/provider-api-key/v1"); err == nil {
|
||||
t.Fatal("expected direct medium environment secret to be rejected")
|
||||
}
|
||||
}
|
||||
|
||||
@@ -214,6 +214,9 @@ services:
|
||||
LOCAL_CPU_PROVIDER_URL: http://docker.cin.su:18082/v1
|
||||
LOCAL_CPU_PROVIDER_API_KEY_FILE: ${LOCAL_CPU_PROVIDER_API_KEY_FILE:-}
|
||||
LOCAL_CPU_PROVIDER_CREDENTIAL_REF: ${LOCAL_CPU_PROVIDER_CREDENTIAL_REF:-credential://domovoy/local-cpu/provider-api-key/v1}
|
||||
LOCAL_MEDIUM_PROVIDER_URL: ${LOCAL_MEDIUM_PROVIDER_URL:-http://llm-local-api:8080/v1}
|
||||
LOCAL_MEDIUM_PROVIDER_API_KEY_FILE: ${LOCAL_MEDIUM_PROVIDER_API_KEY_FILE:-}
|
||||
LOCAL_MEDIUM_PROVIDER_CREDENTIAL_REF: ${LOCAL_MEDIUM_PROVIDER_CREDENTIAL_REF:-credential://domovoy/local-medium/provider-api-key/v1}
|
||||
LOCAL_LIFEBOAT_PROVIDER_URL: http://cognitive-lifeboat:8086/v1
|
||||
LOCAL_GPU_PROVIDER_URL: http://192.168.220.91:11434
|
||||
COGNITIVE_OWNER_PRINCIPAL_ID: domovoy-creator-principal
|
||||
@@ -221,6 +224,7 @@ services:
|
||||
depends_on: [core-runtime, cognitive-lifeboat]
|
||||
volumes:
|
||||
- assistant_core_secrets:/run/secrets:ro
|
||||
networks: [default, llm-local]
|
||||
|
||||
cognitive-lifeboat:
|
||||
image: ghcr.io/ggml-org/llama.cpp:server
|
||||
@@ -287,3 +291,8 @@ volumes:
|
||||
external: true
|
||||
assistant_research_sources:
|
||||
name: assistant_research_sources
|
||||
|
||||
networks:
|
||||
llm-local:
|
||||
external: true
|
||||
name: llm-local_default
|
||||
|
||||
@@ -1,5 +1,5 @@
|
||||
{
|
||||
"task_file": "dev/tasks/DEV-000165-authorized-cpu-health-probe.json",
|
||||
"task_id": "DEV-000165",
|
||||
"task_file": "dev/tasks/DEV-000171-connect-existing-medium-runtime.json",
|
||||
"task_id": "DEV-000171",
|
||||
"task_revision": 1
|
||||
}
|
||||
|
||||
@@ -0,0 +1,23 @@
|
||||
{
|
||||
"task_id": "DEV-000166",
|
||||
"task_revision": 1,
|
||||
"architecture_revision": "ADR-0018",
|
||||
"base_repo_revision": "4086b7522460a78afc341d996ca7874505018871",
|
||||
"risk_tier": "E",
|
||||
"intent": "Give the authenticated CPU readiness probe the provider adapter timeout while retaining short opportunistic GPU probing.",
|
||||
"requirement_ids": ["WP-601", "Model fleet/failover"],
|
||||
"allowed_change_scope": {"modules": ["cognitive-worker", "platform-health"], "paths": ["runtime/health/server.go", "runtime/health/server_test.go", "cmd/cognitive-worker/main.go", "cmd/cognitive-worker/fleet_status.go", "dev/ACTIVE_TASK.json", "dev/tasks/DEV-000166-per-provider-readiness-timeouts.json"]},
|
||||
"forbidden_change_scope": {"paths": ["architecture/**", "security/**", "database/**", "deploy/**", "contracts/**", "adapters/**", ".gitea/**"]},
|
||||
"allowed_architecture_delta": {"new_cross_module_dependencies": 0, "new_canonical_owners": 0, "new_privileges": 0, "new_third_party_dependencies": 0, "new_database_owner_roles": 0},
|
||||
"contracts": {"consume": ["existing health Probe contract", "existing fleet readiness probe"], "may_change": ["per-probe timeout metadata"]},
|
||||
"semantic_owner": "cognitive-worker/platform health",
|
||||
"canonical_state": [],
|
||||
"state_transitions": ["provider-specific probe deadline -> truthful readiness result"],
|
||||
"side_effects_and_privileges": ["no new privilege; existing provider auth only"],
|
||||
"persistence_and_migrations": ["none"],
|
||||
"failure_retry_restart": ["CPU gets bounded 10-second probe; GPU retains 2-second opportunistic timeout"],
|
||||
"assumptions": ["CPU provider catalog may take longer than opportunistic probes while remaining within adapter bounds"],
|
||||
"acceptance_scenarios": ["per-probe health timeout is honored", "CPU fleet client uses 10 seconds", "GPU remains bounded", "Lifeboat fallback remains unchanged"],
|
||||
"required_tests": ["go test ./...", "go vet ./...", "preflight", "check-diff", "check-architecture", "independent", "git diff --check", "CI", "production /readyz and fleet readiness"],
|
||||
"terminal_gap_states": ["BLOCKED_BY_MISSING_CAPABILITY", "BLOCKED_BY_CONFLICT"]
|
||||
}
|
||||
@@ -0,0 +1,23 @@
|
||||
{
|
||||
"task_id": "DEV-000167",
|
||||
"task_revision": 1,
|
||||
"architecture_revision": "ADR-0018",
|
||||
"base_repo_revision": "c7b57daaf8233b0f0e79c77c02e92040378921d5",
|
||||
"risk_tier": "E",
|
||||
"intent": "Use the bounded CPU provider HTTP client for the authenticated readiness probe so /readyz reports the same truthful result as fleet readiness.",
|
||||
"requirement_ids": ["WP-601", "Model fleet/failover"],
|
||||
"allowed_change_scope": {"modules": ["cognitive-worker"], "paths": ["cmd/cognitive-worker/main.go", "dev/ACTIVE_TASK.json", "dev/tasks/DEV-000167-cpu-readiness-client-timeout.json"]},
|
||||
"forbidden_change_scope": {"paths": ["architecture/**", "security/**", "database/**", "deploy/**", "contracts/**", "adapters/**", ".gitea/**"]},
|
||||
"allowed_architecture_delta": {"new_cross_module_dependencies": 0, "new_canonical_owners": 0, "new_privileges": 0, "new_third_party_dependencies": 0, "new_database_owner_roles": 0},
|
||||
"contracts": {"consume": ["existing readiness Probe contract", "existing authenticated CPU fleet endpoint"], "may_change": []},
|
||||
"semantic_owner": "cognitive-worker",
|
||||
"canonical_state": [],
|
||||
"state_transitions": ["CPU provider response -> consistent /readyz and fleet readiness state"],
|
||||
"side_effects_and_privileges": ["no new privilege; existing provider auth only"],
|
||||
"persistence_and_migrations": ["none"],
|
||||
"failure_retry_restart": ["CPU readiness remains bounded at 10 seconds; opportunistic GPU remains 2 seconds"],
|
||||
"assumptions": ["The CPU provider is authenticated with the existing file-backed API key"],
|
||||
"acceptance_scenarios": ["/readyz can report CPU AVAILABLE when fleet readiness reports CPU AVAILABLE", "GPU remains bounded", "Lifeboat fallback remains unchanged"],
|
||||
"required_tests": ["go test ./...", "go vet ./...", "preflight", "check-diff", "check-architecture", "independent", "git diff --check", "CI", "production /readyz and fleet readiness"],
|
||||
"terminal_gap_states": ["BLOCKED_BY_MISSING_CAPABILITY", "BLOCKED_BY_CONFLICT"]
|
||||
}
|
||||
@@ -0,0 +1,23 @@
|
||||
{
|
||||
"task_id": "DEV-000168",
|
||||
"task_revision": 1,
|
||||
"architecture_revision": "ADR-0018",
|
||||
"base_repo_revision": "205d8f620603ccf3e0138f6beb93156450664314",
|
||||
"risk_tier": "E",
|
||||
"intent": "Make the MEDIUM model-fleet slot a real optional OpenAI-compatible provider path with protected file credentials, discovery, readiness and gateway registration.",
|
||||
"requirement_ids": ["WP-200", "WP-301", "Model fleet/failover"],
|
||||
"allowed_change_scope": {"modules": ["cognitive-worker", "deployment-policy", "tests"], "paths": ["cmd/cognitive-worker/main.go", "cmd/cognitive-worker/main_test.go", "deploy/docker-compose.prod.yml", "tests/cognitive_advisory/cognitive_worker_env_policy.ps1", "tests/cognitive_advisory/cognitive_worker_env_policy_test.ps1", "dev/ACTIVE_TASK.json", "dev/tasks/DEV-000168-configurable-medium-fleet-slot.json"]},
|
||||
"forbidden_change_scope": {"paths": ["architecture/**", "security/**", "database/**", "contracts/**", "adapters/**", ".gitea/**"]},
|
||||
"allowed_architecture_delta": {"new_cross_module_dependencies": 0, "new_canonical_owners": 0, "new_privileges": 1, "new_third_party_dependencies": 0, "new_database_owner_roles": 0},
|
||||
"contracts": {"consume": ["existing provider adapter contract", "existing discovery registration", "existing fleet readiness contract", "existing CredentialRef file contract"], "may_change": []},
|
||||
"semantic_owner": "cognitive-worker/capability registration boundary",
|
||||
"canonical_state": [],
|
||||
"state_transitions": ["MEDIUM endpoint absent -> NOT_CONFIGURED", "MEDIUM endpoint configured -> discovery/readiness registration", "available provider -> eligible gateway binding subject to core registry policy"],
|
||||
"side_effects_and_privileges": ["declares one new deployment configuration surface for an existing read-only secret volume; provider key is resolved only through the existing protected file CredentialRef"],
|
||||
"persistence_and_migrations": ["none"],
|
||||
"failure_retry_restart": ["medium probe has a bounded 10-second client timeout; absent endpoint remains NOT_CONFIGURED; fallback selection remains Lifeboat when preferred tiers are unavailable"],
|
||||
"assumptions": ["A real medium OpenAI-compatible runtime may be supplied later through deployment environment injection", "No production medium runtime is claimed by this change"],
|
||||
"acceptance_scenarios": ["empty medium URL produces NOT_CONFIGURED without startup failure", "configured medium URL is registered for discovery and fleet readiness", "direct medium secret environment is rejected", "medium CredentialRef file is accepted", "CPU/GPU/Lifeboat behavior remains unchanged"],
|
||||
"required_tests": ["go test ./...", "go vet ./...", "preflight", "check-diff", "check-architecture", "independent", "git diff --check", "environment policy tests", "CI", "production fleet readiness"],
|
||||
"terminal_gap_states": ["BLOCKED_BY_MISSING_CAPABILITY", "BLOCKED_BY_CONFLICT"]
|
||||
}
|
||||
@@ -0,0 +1,23 @@
|
||||
{
|
||||
"task_id": "DEV-000169",
|
||||
"task_revision": 1,
|
||||
"architecture_revision": "ADR-0018",
|
||||
"base_repo_revision": "8f3920519b6a228d78b6ad775a7c19838dfbb70b",
|
||||
"risk_tier": "E",
|
||||
"intent": "Keep an absent optional MEDIUM provider out of runtime readiness while preserving its explicit NOT_CONFIGURED fleet member.",
|
||||
"requirement_ids": ["WP-200", "WP-301", "Model fleet/failover"],
|
||||
"allowed_change_scope": {"modules": ["cognitive-worker"], "paths": ["cmd/cognitive-worker/main.go", "dev/ACTIVE_TASK.json", "dev/tasks/DEV-000169-optional-medium-readiness.json"]},
|
||||
"forbidden_change_scope": {"paths": ["architecture/**", "security/**", "database/**", "deploy/**", "contracts/**", "adapters/**", ".gitea/**"]},
|
||||
"allowed_architecture_delta": {"new_cross_module_dependencies": 0, "new_canonical_owners": 0, "new_privileges": 0, "new_third_party_dependencies": 0, "new_database_owner_roles": 0},
|
||||
"contracts": {"consume": ["existing readiness handler", "existing fleet readiness contract"], "may_change": []},
|
||||
"semantic_owner": "cognitive-worker",
|
||||
"canonical_state": [],
|
||||
"state_transitions": ["MEDIUM absent -> fleet NOT_CONFIGURED without runtime readiness degradation", "MEDIUM configured -> readiness probe included"],
|
||||
"side_effects_and_privileges": ["none"],
|
||||
"persistence_and_migrations": ["none"],
|
||||
"failure_retry_restart": ["configured MEDIUM remains bounded by its 10-second client; absent MEDIUM is not probed"],
|
||||
"assumptions": ["Fleet status remains the explicit inventory surface for unconfigured optional tiers"],
|
||||
"acceptance_scenarios": ["production /readyz stays ready when MEDIUM URL is empty", "fleet readiness retains MEDIUM NOT_CONFIGURED", "configured MEDIUM is included in /readyz"],
|
||||
"required_tests": ["go test ./...", "go vet ./...", "preflight", "check-diff", "check-architecture", "independent", "git diff --check", "CI", "production /readyz and fleet readiness"],
|
||||
"terminal_gap_states": ["BLOCKED_BY_MISSING_CAPABILITY", "BLOCKED_BY_CONFLICT"]
|
||||
}
|
||||
@@ -0,0 +1,23 @@
|
||||
{
|
||||
"task_id": "DEV-000170",
|
||||
"task_revision": 1,
|
||||
"architecture_revision": "ADR-0018",
|
||||
"base_repo_revision": "51427a246e1ccc115a6c2640790d9220beca95e9",
|
||||
"risk_tier": "E",
|
||||
"intent": "Refresh audit completion and production fleet evidence after the configurable MEDIUM slot and optional readiness fixes were deployed.",
|
||||
"requirement_ids": ["AUDIT-COMPLETION-TRUTH", "WP-200", "WP-301"],
|
||||
"allowed_change_scope": {"modules": [], "paths": ["docs/evidence/audit-completion-matrix-2026-08-22-v1.json", "docs/evidence/wp-200-wp-301-fleet-execution-status-2026-08-22-v2.json", "docs/PROJECT_STATUS.md", "dev/ACTIVE_TASK.json", "dev/tasks/DEV-000170-refresh-fleet-audit-evidence.json"]},
|
||||
"forbidden_change_scope": {"paths": ["architecture/**", "security/**", "cmd/**", "modules/**", "contracts/**", "adapters/**", "database/**", "deploy/**", ".gitea/**"]},
|
||||
"allowed_architecture_delta": {"new_cross_module_dependencies": 0, "new_canonical_owners": 0, "new_privileges": 0, "new_third_party_dependencies": 0, "new_database_owner_roles": 0},
|
||||
"contracts": {"consume": ["live fleet readiness evidence", "audit completion matrix"], "may_change": []},
|
||||
"semantic_owner": "none; audit evidence projection",
|
||||
"canonical_state": [],
|
||||
"state_transitions": ["stale evidence -> current production evidence"],
|
||||
"side_effects_and_privileges": ["none"],
|
||||
"persistence_and_migrations": ["immutable JSON evidence artifacts only"],
|
||||
"failure_retry_restart": ["not applicable"],
|
||||
"assumptions": ["live probe output was captured after production deployment of 51427a24", "missing external capabilities remain explicitly incomplete"],
|
||||
"acceptance_scenarios": ["production revision is current", "MEDIUM configuration and readiness semantics are recorded", "agent bridge and authenticated UI gaps remain visible", "no secret value appears"],
|
||||
"required_tests": ["JSON validation", "go test ./...", "go vet ./...", "preflight", "check-diff", "check-architecture", "independent", "git diff --check", "CI"],
|
||||
"terminal_gap_states": ["BLOCKED_BY_MISSING_CAPABILITY", "BLOCKED_BY_CONFLICT"]
|
||||
}
|
||||
@@ -0,0 +1,23 @@
|
||||
{
|
||||
"task_id": "DEV-000171",
|
||||
"task_revision": 1,
|
||||
"architecture_revision": "ADR-0018",
|
||||
"base_repo_revision": "503fe840741ac99499e06bd93df977acf1baad41",
|
||||
"risk_tier": "E",
|
||||
"intent": "Connect cognitive-worker to the already-operated llm-local CPU model runtime as the configured MEDIUM fleet member through an explicit external Docker network.",
|
||||
"requirement_ids": ["WP-200", "WP-301", "Model fleet/failover"],
|
||||
"allowed_change_scope": {"modules": ["deployment-topology", "cognitive-worker"], "paths": ["deploy/docker-compose.prod.yml", "dev/ACTIVE_TASK.json", "dev/tasks/DEV-000171-connect-existing-medium-runtime.json"]},
|
||||
"forbidden_change_scope": {"paths": ["architecture/**", "security/**", "database/**", "contracts/**", "adapters/**", ".gitea/**"]},
|
||||
"allowed_architecture_delta": {"new_cross_module_dependencies": 0, "new_canonical_owners": 0, "new_privileges": 1, "new_third_party_dependencies": 0, "new_database_owner_roles": 0},
|
||||
"contracts": {"consume": ["existing OpenAI-compatible local provider adapter", "existing MEDIUM fleet readiness contract", "existing llm-local Docker network"], "may_change": []},
|
||||
"semantic_owner": "deployment topology/cognitive-worker registration",
|
||||
"canonical_state": [],
|
||||
"state_transitions": ["MEDIUM unconfigured -> live provider discovery/readiness", "preferred HEAVY unavailable -> highest available fallback selection"],
|
||||
"side_effects_and_privileges": ["one explicit read-only network attachment from cognitive-worker to existing llm-local runtime; no secret or database privilege added"],
|
||||
"persistence_and_migrations": ["none"],
|
||||
"failure_retry_restart": ["if external llm-local network is absent deployment fails closed; provider timeout remains bounded; Lifeboat remains fallback"],
|
||||
"assumptions": ["production llm-local_default network and llm-local-api are independently operated and present", "llm-local-api exposes OpenAI-compatible /v1/models and chat routes on port 8080"],
|
||||
"acceptance_scenarios": ["compose config validates with external network declaration", "cognitive-worker resolves llm-local-api", "MEDIUM fleet member is AVAILABLE live", "runtime /readyz remains 200", "Lifeboat fallback remains available", "no secret value is added"],
|
||||
"required_tests": ["docker compose config", "go test ./...", "go vet ./...", "preflight", "check-diff", "check-architecture", "independent", "git diff --check", "CI", "production /readyz and fleet readiness"],
|
||||
"terminal_gap_states": ["BLOCKED_BY_MISSING_CAPABILITY", "BLOCKED_BY_CONFLICT"]
|
||||
}
|
||||
@@ -2,12 +2,12 @@
|
||||
|
||||
## Current audit delta — 2026-08-22
|
||||
|
||||
- `CURRENT_PRODUCTION_REVISION = bda71caecece67e5ef225ac250a7a80c4bc84d94`; public `/healthz` is healthy.
|
||||
- `CURRENT_PRODUCTION_REVISION = 51427a246e1ccc115a6c2640790d9220beca95e9`; public `/healthz` is healthy.
|
||||
- `FAST_PROVIDER_STATUS = LIVE_SMOKE_SUPPORTED`: local CPU gateway authentication is configured through a protected file reference; four bounded local-only advisory cases completed through cognitive-worker. Evidence: `docs/evidence/wp-601-fast-gateway-live-smoke-2026-08-22-v1.json`.
|
||||
- `MEDIUM_PROVIDER_STATUS = NOT_CONFIGURED`; no medium runtime is registered in production.
|
||||
- `MEDIUM_PROVIDER_STATUS = CONFIGURABLE_NOT_CONFIGURED`; the optional OpenAI-compatible MEDIUM slot, protected CredentialRef file path, discovery and readiness wiring are deployed; no medium runtime is supplied in production.
|
||||
- `HEAVY_PROVIDER_STATUS = UNAVAILABLE`; GPU endpoint `192.168.220.91:11434` timed out during live discovery.
|
||||
- `AGENT_EXECUTION_STATUS = RECOVERY_ONLY`; optional `AGENT_EXECUTION_BRIDGE_URL` is wired into production compose but remains unset, and no Codex/bridge runtime is deployed. Durable recovery and status endpoints remain active.
|
||||
- `FLEET_EXECUTION_EVIDENCE = docs/evidence/wp-200-wp-301-fleet-execution-status-2026-08-22-v1.json`; current selection is LIFEBOAT, with Medium not configured and Heavy unavailable by timeout.
|
||||
- `FLEET_EXECUTION_EVIDENCE = docs/evidence/wp-200-wp-301-fleet-execution-status-2026-08-22-v2.json`; FAST and LIFEBOAT are ready, MEDIUM is NOT_CONFIGURED, Heavy is unavailable by timeout, and current selection is LIFEBOAT for preferred HEAVY.
|
||||
- `BACKUP_RESTORE_EVIDENCE = docs/evidence/wp-002-postgres-restore-2026-08-22-v1.json`; latest backup checksum and isolated restore verified with 11 schemas and 60 tables.
|
||||
- `UI_AUTHENTICATED_ACCEPTANCE = PENDING`; public login UI is reachable, but no authenticated browser session was available for cabinet visual acceptance.
|
||||
- `SECURITY_NOTE = one production executor secret was rotated after an accidental diagnostic disclosure; values are not retained in repository evidence.`
|
||||
|
||||
@@ -1,14 +1,14 @@
|
||||
{
|
||||
"evidence_id": "audit-completion-matrix-2026-08-22-v1",
|
||||
"captured_at": "2026-08-22",
|
||||
"production_revision": "bda71caecece67e5ef225ac250a7a80c4bc84d94",
|
||||
"production_revision": "51427a246e1ccc115a6c2640790d9220beca95e9",
|
||||
"objective": "Dovести оставшиеся пункты аудита Domovoy до проверяемого состояния",
|
||||
"matrix": [
|
||||
{"area":"Agent Execution runtime and recovery","status":"PARTIAL_EXTERNAL_GAP","repository_proof":["dev/tasks/DEV-000145-agent-execution-bridge-runtime.json","adapters/agentbridge/http.go","cmd/agent-worker/main.go"],"production_proof":["docs/evidence/wp-200-wp-301-fleet-execution-status-2026-08-22-v1.json"],"missing":"independently operated bridge endpoint is not configured; live worker is recovery-only"},
|
||||
{"area":"Multi-Agent Work Graph","status":"REPOSITORY_AND_RUNTIME_FOUNDATION_VERIFIED","repository_proof":["dev/tasks/DEV-000131-work-graph-harness.json","dev/tasks/DEV-000146-work-graph-runtime.json","tests/agent_execution/graph_runtime_test.go"],"production_proof":["docs/PROJECT_STATUS.md"],"missing":"no claim of external agent execution fan-out is made while bridge is absent"},
|
||||
{"area":"Truth/Research runtime","status":"REPOSITORY_AND_RUNTIME_FOUNDATION_VERIFIED","repository_proof":["dev/tasks/DEV-000134-truth-bundle-compiler.json","dev/tasks/DEV-000143-research-runtime-orchestration.json","tests/research_runtime"],"production_proof":["docs/PROJECT_STATUS.md"],"missing":"authenticated end-to-end research acceptance is not claimed here"},
|
||||
{"area":"Learning reflection loop","status":"REPOSITORY_AND_RUNTIME_FOUNDATION_VERIFIED","repository_proof":["dev/tasks/DEV-000135-reflection-worker.json","cmd/reflection-worker/main.go","tests/research_contracts/reflection_worker_sql_test.go"],"production_proof":["docs/PROJECT_STATUS.md"],"missing":"no claim of fully closed historical candidate backlog"},
|
||||
{"area":"Model fleet/failover","status":"PARTIAL_EXTERNAL_GAP","repository_proof":["dev/tasks/DEV-000144-fleet-readiness-runtime.json","cmd/cognitive-worker/fleet_status.go"],"production_proof":["docs/evidence/wp-200-wp-301-fleet-execution-status-2026-08-22-v1.json"],"missing":"Medium is not configured; Heavy GPU probe times out; Lifeboat is selected fallback"},
|
||||
{"area":"Model fleet/failover","status":"PARTIAL_EXTERNAL_GAP","repository_proof":["dev/tasks/DEV-000144-fleet-readiness-runtime.json","dev/tasks/DEV-000168-configurable-medium-fleet-slot.json","dev/tasks/DEV-000169-optional-medium-readiness.json","cmd/cognitive-worker/fleet_status.go"],"production_proof":["docs/evidence/wp-200-wp-301-fleet-execution-status-2026-08-22-v2.json"],"missing":"Medium runtime is not configured; Heavy GPU probe times out; FAST and Lifeboat are ready and Lifeboat is selected fallback"},
|
||||
{"area":"Governed self-development","status":"REPOSITORY_FOUNDATION_VERIFIED","repository_proof":["dev/tasks/DEV-000137-governed-self-development-gate.json","tests/semantic_evolution"],"production_proof":["docs/PROJECT_STATUS.md"],"missing":"real bridge-backed self-development execution is not claimed"},
|
||||
{"area":"UI/operations acceptance","status":"PARTIAL_EXTERNAL_GAP","repository_proof":["dev/tasks/DEV-000138-ui-operations-acceptance-contract.json","tests/creator_web/cabinet_acceptance_contract_test.go","dev/tasks/DEV-000150-agent-operations-ui.json","dev/tasks/DEV-000151-agent-worker-status-ui.json"],"production_proof":["docs/PROJECT_STATUS.md"],"missing":"authenticated browser visual acceptance is pending because no signed-in session is available"}
|
||||
],
|
||||
|
||||
@@ -0,0 +1,36 @@
|
||||
{
|
||||
"evidence_id": "wp-200-wp-301-fleet-execution-status-2026-08-22-v2",
|
||||
"captured_at": "2026-08-22",
|
||||
"production_revision": "51427a246e1ccc115a6c2640790d9220beca95e9",
|
||||
"source": "live internal status probes on docker.cin.su after DEV-000168 and DEV-000169 deployment",
|
||||
"fleet_readiness": {
|
||||
"preferred_tier": "HEAVY",
|
||||
"selected_tier": "LIFEBOAT",
|
||||
"ready_tiers": ["FAST", "LIFEBOAT"],
|
||||
"members": [
|
||||
{"provider_instance_id": "local-cpu-primary", "tier": "FAST", "ready": true, "reason": "AVAILABLE"},
|
||||
{"provider_instance_id": "local-medium-primary", "tier": "MEDIUM", "ready": false, "reason": "NOT_CONFIGURED"},
|
||||
{"provider_instance_id": "local-gpu-opportunistic", "tier": "HEAVY", "ready": false, "reason": "context deadline exceeded while awaiting headers"},
|
||||
{"provider_instance_id": "cognitive-lifeboat", "tier": "LIFEBOAT", "ready": true, "reason": "AVAILABLE"}
|
||||
]
|
||||
},
|
||||
"runtime_readiness": {
|
||||
"status": "ready",
|
||||
"cpu": "READY/AVAILABLE",
|
||||
"lifeboat": "READY/AVAILABLE",
|
||||
"gpu": "DEGRADED/DEPENDENCY_UNAVAILABLE",
|
||||
"medium": "not included in /readyz while unconfigured"
|
||||
},
|
||||
"agent_execution": {
|
||||
"runtime": "agent-worker",
|
||||
"executor_configured": false,
|
||||
"mode": "recovery-only",
|
||||
"ready": false,
|
||||
"reason": "AGENT_EXECUTION_BRIDGE_URL is not configured"
|
||||
},
|
||||
"interpretation": [
|
||||
"MEDIUM is now a real optional protected-file-authenticated configuration slot, but no runtime is claimed without a supplied endpoint.",
|
||||
"An absent optional MEDIUM slot remains explicit in fleet inventory and does not degrade cognitive-worker /readyz.",
|
||||
"This artifact records observed state only and contains no secret values."
|
||||
]
|
||||
}
|
||||
@@ -29,6 +29,7 @@ type Probe struct {
|
||||
Name string
|
||||
Check func(context.Context) error
|
||||
Optional bool
|
||||
Timeout time.Duration
|
||||
}
|
||||
|
||||
type Result struct {
|
||||
@@ -56,7 +57,11 @@ func (h ReadinessHandler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
|
||||
if probe.Check == nil {
|
||||
result.State, result.Reason, result.Detail = "DEGRADED", "PROBE_NOT_CONFIGURED", "probe is not configured"
|
||||
} else {
|
||||
ctx, cancel := context.WithTimeout(r.Context(), timeout)
|
||||
probeTimeout := timeout
|
||||
if probe.Timeout > 0 {
|
||||
probeTimeout = probe.Timeout
|
||||
}
|
||||
ctx, cancel := context.WithTimeout(r.Context(), probeTimeout)
|
||||
err := probe.Check(ctx)
|
||||
cancel()
|
||||
if err != nil {
|
||||
|
||||
@@ -7,6 +7,7 @@ import (
|
||||
"net/http/httptest"
|
||||
"strings"
|
||||
"testing"
|
||||
"time"
|
||||
)
|
||||
|
||||
func TestReadinessReturns503WithTypedDependencyReason(t *testing.T) {
|
||||
@@ -41,6 +42,17 @@ func TestHTTPProbeWithHeadersSendsHeaders(t *testing.T) {
|
||||
}
|
||||
}
|
||||
|
||||
func TestReadinessUsesPerProbeTimeout(t *testing.T) {
|
||||
handler := ReadinessHandler{Runtime: "test", Timeout: time.Millisecond, Probes: []Probe{{
|
||||
Name: "slow-but-required", Timeout: time.Second, Check: func(context.Context) error { return nil },
|
||||
}}}
|
||||
recorder := httptest.NewRecorder()
|
||||
handler.ServeHTTP(recorder, httptest.NewRequest(http.MethodGet, "/readyz", nil))
|
||||
if recorder.Code != http.StatusOK {
|
||||
t.Fatalf("readiness response = %d %s", recorder.Code, recorder.Body.String())
|
||||
}
|
||||
}
|
||||
|
||||
func TestReadinessStaysAvailableWithOptionalDependencyDegraded(t *testing.T) {
|
||||
handler := ReadinessHandler{Runtime: "test", Probes: []Probe{
|
||||
{Name: "lifeboat", Check: func(context.Context) error { return nil }},
|
||||
|
||||
@@ -3,6 +3,11 @@ Set-StrictMode -Version Latest
|
||||
$script:AllowedCognitiveWorkerEnvironment = @{
|
||||
'CORE_RUNTIME_URL' = 'http_url'
|
||||
'LOCAL_CPU_PROVIDER_URL' = 'http_url'
|
||||
'LOCAL_CPU_PROVIDER_API_KEY_FILE' = 'secret_file'
|
||||
'LOCAL_CPU_PROVIDER_CREDENTIAL_REF' = 'nonempty'
|
||||
'LOCAL_MEDIUM_PROVIDER_URL' = 'http_url'
|
||||
'LOCAL_MEDIUM_PROVIDER_API_KEY_FILE' = 'secret_file'
|
||||
'LOCAL_MEDIUM_PROVIDER_CREDENTIAL_REF' = 'nonempty'
|
||||
'LOCAL_GPU_PROVIDER_URL' = 'http_url'
|
||||
'LOCAL_OLLAMA_PROVIDER_URL' = 'http_url'
|
||||
'COGNITIVE_OWNER_PRINCIPAL_ID' = 'nonempty'
|
||||
|
||||
@@ -15,6 +15,8 @@ function Assert-PolicyFail([string]$Name, [string[]]$Environment) {
|
||||
}
|
||||
|
||||
Assert-PolicyPass 'CPU provider endpoint' @('LOCAL_CPU_PROVIDER_URL=http://docker.cin.su:18082/v1')
|
||||
Assert-PolicyPass 'optional medium provider endpoint' @('LOCAL_MEDIUM_PROVIDER_URL=http://medium.example/v1')
|
||||
Assert-PolicyPass 'approved medium CredentialRef file' @('LOCAL_MEDIUM_PROVIDER_API_KEY_FILE=/run/secrets/medium_api_key', 'LOCAL_MEDIUM_PROVIDER_CREDENTIAL_REF=credential://domovoy/local-medium/provider-api-key/v1')
|
||||
Assert-PolicyPass 'Ollama provider endpoint' @('LOCAL_OLLAMA_PROVIDER_URL=http://192.168.220.91:11434')
|
||||
Assert-PolicyFail 'plaintext OpenAI API key' @('OPENAI_API_KEY=secret')
|
||||
Assert-PolicyFail 'plaintext provider token' @('PROVIDER_TOKEN=secret')
|
||||
|
||||
Reference in New Issue
Block a user