llmgw/internal/control/adminapi/routes.go

package adminapi

import (
	"net/http"

	llmgw "gitea.mynodes.xyz/ilker/llmgw"
	"gitea.mynodes.xyz/ilker/llmgw/internal/bench"
	"gitea.mynodes.xyz/ilker/llmgw/internal/bench/eval"
	"gitea.mynodes.xyz/ilker/llmgw/internal/control/catalog"
	"gitea.mynodes.xyz/ilker/llmgw/internal/control/fabrics"
	"gitea.mynodes.xyz/ilker/llmgw/internal/control/federation"
	"gitea.mynodes.xyz/ilker/llmgw/internal/control/modelserve"
	"gitea.mynodes.xyz/ilker/llmgw/internal/control/nvlink"
	"gitea.mynodes.xyz/ilker/llmgw/internal/guard"
	"gitea.mynodes.xyz/ilker/llmgw/internal/notify"
	"gitea.mynodes.xyz/ilker/llmgw/internal/store"
)

// Route is one endpoint as the OpenAPI document describes it.
//
// A table written by hand beside Register rather than read out of the
// handlers, because what a handler decodes and answers is not visible from
// outside it. The table is checked against Register by a test, so an endpoint
// cannot be added without being described; the bodies are Go values whose
// types the generator reads, so a field added to one is in the document the
// next time it is generated.
type Route struct {
	Method string
	Path   string
	// Summary is one line; Description, when there is more to say, the rest.
	Summary     string
	Description string
	// Query is the query parameters the handler reads.
	Query []Param
	// Request is a value of the body's type; nil reads none.
	Request any
	// OptionalBody is set when the body may be left out.
	OptionalBody bool
	// Response is a value of the type a success answers with.
	Response any
	// Status is the success status; zero is 200.
	Status int
	// CSV is set when the answer is CSV unless JSON is asked for.
	CSV bool
}

// Param is a query parameter.
type Param struct {
	Name        string
	Description string
	// Type is a JSON Schema type: string when empty, or integer, number,
	// boolean.
	Type     string
	Required bool
}

// PathParams describes the parameters in Routes' paths, by the segment
// before them and their name.
var PathParams = map[string]string{
	"models/{name}":     "The model's name in the catalog.",
	"nodes/{id}":        "The node's ID, as `llmgw ctl nodes` lists it.",
	"jobs/{id}":         "The job's ID, as starting it answered.",
	"runs/{id}":         "The quality run's ID, as starting it answered.",
	"requests/{id}":     "The request's ID: the x-request-id a client was answered with.",
	"fabrics/{id}":      "The physical fabric's ID, as GET /api/v1/fabrics lists it.",
	"quotas/{scope}":    "The scope: org:NAME, team:ORG/TEAM, project:ORG/TEAM/PROJECT, or their IDs after the colon.",
	"federation/{peer}": "The federation peer's name, as `llmgw ctl federation list` lists it.",
}

// Routes is every endpoint Register mounts.
var Routes = []Route{
	// The document itself.
	{Method: "GET", Path: "/api/v1/openapi.json", Summary: "This document",
		Description: "The admin API in OpenAPI 3.1, generated from the route table and built into the binary.",
		Response:    map[string]any{}},

	// The fleet.
	{Method: "GET", Path: "/api/v1/summary", Summary: "The whole fleet in one answer",
		Response: Summary{}},
	{Method: "GET", Path: "/api/v1/nodes", Summary: "Every node, with its cards, load and fabrics",
		Response: []NodeView{}},
	{Method: "POST", Path: "/api/v1/nodes/{id}/state", Summary: "Cordon, drain or ready a node",
		Request: NodeStateRequest{}, Response: NodeStateView{}},
	{Method: "POST", Path: "/api/v1/nodes/{id}/upgrade", Summary: "Have a node run a release",
		Request: UpgradeRequest{}, Response: UpgradeView{}},
	{Method: "POST", Path: "/api/v1/nodes/{id}/engines", Summary: "Choose an engine's build for a node",
		Description: "The node installs the build chosen. Build is auto, a build such as cu129, or none.",
		Request:     NodeEngineRequest{}, Response: NodeEngineRequest{}},
	{Method: "DELETE", Path: "/api/v1/nodes/{id}", Summary: "Forget a node that has gone",
		Description: "Only a node that is lost, or unreachable with nothing on it; its certificate is revoked.",
		Response:    NodeRemoved{}},
	{Method: "GET", Path: "/api/v1/nodes/{id}/logs", Summary: "The last lines of a node's logs",
		Query: []Param{
			{Name: "lines", Type: "integer", Description: "How many lines; 300 when absent, at most 2000."},
			{Name: "source", Description: "Which log: agent when absent, or an engine the node lists in sources."},
		},
		Response: LogsView{}},
	{Method: "GET", Path: "/api/v1/fabrics", Summary: "The physical fabrics and the names nodes give them",
		Response: fabrics.Report{}},
	{Method: "GET", Path: "/api/v1/fabrics/{id}/topology", Summary: "One fabric's map, with the cluster's adapters live on it",
		Response: TopologyView{}},
	{Method: "GET", Path: "/api/v1/nvlink", Summary: "The multi-node NVLink domains and what keeps nodes out of them",
		Response: nvlink.Report{}},
	{Method: "POST", Path: "/api/v1/fabric-test", Summary: "Measure bandwidth between two nodes, rail by rail",
		Request: FabricTestRequest{}, Response: FabricTestView{}},
	{Method: "GET", Path: "/api/v1/weights/orphans", Summary: "Weights on nodes that no model uses",
		Response: []OrphanView{}},
	{Method: "POST", Path: "/api/v1/weights/collect", Summary: "Delete weights that no model uses",
		Request: CollectRequest{}, OptionalBody: true, Response: []CollectResult{}},

	// The catalog.
	{Method: "GET", Path: "/api/v1/models", Summary: "Every model, with its replicas and state",
		Response: []ModelView{}},
	{Method: "POST", Path: "/api/v1/models", Summary: "Add a model to the catalog",
		Description: "Kept in the store across restarts. Refused when the name is taken, a model in " +
			"configuration has it, or the definition would make the catalog invalid.",
		Request: catalog.Model{}, Response: ModelAdded{}},
	{Method: "GET", Path: "/api/v1/models/{name}", Summary: "One model's definition, as it is in force",
		Description: "With the defaults the catalog filled in. By its name only, not an alias.",
		Response:    catalog.Model{}},
	{Method: "DELETE", Path: "/api/v1/models/{name}", Summary: "Remove a model added at runtime",
		Description: "A model declared in configuration is refused: it would come back on the next restart.",
		Response:    Removed{}},
	{Method: "POST", Path: "/api/v1/models/{name}/scale", Summary: "Set how many replicas a model runs",
		Request: ScaleRequest{}, Response: ScaleResponse{}},
	{Method: "POST", Path: "/api/v1/models/{name}/resources", Summary: "Set what a model's profiles reserve of a node",
		Request: ResourcesRequest{}, Response: catalog.Model{}},
	{Method: "POST", Path: "/api/v1/models/{name}/class", Summary: "Set a model's capacity class",
		Request: ClassRequest{}, Response: ClassView{}},
	{Method: "POST", Path: "/api/v1/models/{name}/owner", Summary: "Set the part of the tenancy tree a model's capacity counts against",
		Request: OwnerRequest{}, Response: OwnerView{}},
	{Method: "POST", Path: "/api/v1/models/{name}/adapters", Summary: "Replace a model's LoRA adapters",
		Description: "Each new file is read first, to check it is an adapter for the model's architecture.",
		Request:     AdaptersRequest{}, Response: AdaptersRequest{}},
	{Method: "POST", Path: "/api/v1/models/{name}/retry", Summary: "Try a parked model again",
		Response: RetryView{}},
	{Method: "GET", Path: "/api/v1/models/{name}/target", Summary: "How many replicas a throughput needs, and whether they fit",
		Query: []Param{{Name: "tokens_per_sec", Type: "number", Required: true,
			Description: "The output tokens a second wanted of the model."}},
		Response: TargetView{}},
	{Method: "GET", Path: "/api/v1/models/{name}/headroom", Summary: "How many more replicas of a model fit now, and where",
		Query: []Param{{Name: "limit", Type: "integer",
			Description: "The most to try placing; 16 when absent or past 64."}},
		Response: HeadroomView{}},
	{Method: "PUT", Path: "/api/v1/models/{name}/capacity", Summary: "Record what a replica was measured to serve",
		Description: "Replaces an earlier measurement of the same model on the same profile.",
		Request:     bench.Capacity{}, Response: bench.Capacity{}},
	{Method: "GET", Path: "/api/v1/capacity", Summary: "Every recorded measurement",
		Response: []CapacityView{}},
	{Method: "GET", Path: "/api/v1/replicas", Summary: "Every replica, with its shards",
		Query:    []Param{{Name: "model", Description: "Only this model's replicas."}},
		Response: []ReplicaView{}},
	{Method: "GET", Path: "/api/v1/explain-placement", Summary: "Why a model could not be placed",
		Query:    []Param{{Name: "model", Required: true, Description: "The model's name or an alias."}},
		Response: ExplainView{}},
	{Method: "POST", Path: "/api/v1/restart", Summary: "Restart replicas, every replica of some models, or all",
		Request: RestartRequest{}, Response: RestartResult{}, Status: http.StatusAccepted},

	// Serving from the registry, and newer weights.
	{Method: "GET", Path: "/api/v1/plan", Summary: "How a registry repository would be served here",
		Query: []Param{
			{Name: "repo", Required: true, Description: "The repository, owner/name."},
			{Name: "format", Description: "gguf or safetensors; gguf when absent."},
			{Name: "context", Type: "integer", Description: "The context length to plan for, in tokens."},
		},
		Response: modelserve.Plan{}},
	{Method: "POST", Path: "/api/v1/models/serve", Summary: "Fetch a registry repository and serve it",
		Request: modelserve.Request{}, Response: modelserve.Job{}, Status: http.StatusAccepted},
	{Method: "GET", Path: "/api/v1/jobs", Summary: "Every job since the control plane started",
		Response: []modelserve.Job{}},
	{Method: "GET", Path: "/api/v1/jobs/{id}", Summary: "One job",
		Response: modelserve.Job{}},
	{Method: "GET", Path: "/api/v1/updates", Summary: "Newer weights the registry has for models",
		Response: []modelserve.UpdateInfo{}},
	{Method: "POST", Path: "/api/v1/updates/check", Summary: "Ask the registry now, about one model or every one",
		Request: CheckRequest{}, OptionalBody: true, Response: []modelserve.UpdateInfo{}},
	{Method: "POST", Path: "/api/v1/models/{name}/update", Summary: "Move a model to its repository's newer weights",
		Description: "With a quality suite, or the model's standing gate, the new weights are asked it first as a hidden canary.",
		Request:     modelserve.Change{}, OptionalBody: true, Response: modelserve.Job{}, Status: http.StatusAccepted},
	{Method: "POST", Path: "/api/v1/models/{name}/rollback", Summary: "Go back to the weights the last update replaced",
		Response: Result{}},
	{Method: "POST", Path: "/api/v1/models/{name}/origin", Summary: "Record the repository a model's weights came from",
		Request: OriginRequest{}, Response: OriginView{}},
	{Method: "GET", Path: "/api/v1/canaries", Summary: "Every canary, and how its two sides serve",
		Response: []modelserve.CanaryView{}},
	{Method: "POST", Path: "/api/v1/models/{name}/canary", Summary: "Start a canary of a model's newer weights",
		Request: catalog.Canary{}, Response: modelserve.Job{}, Status: http.StatusAccepted},
	{Method: "PUT", Path: "/api/v1/models/{name}/canary", Summary: "Change the share of requests a canary takes",
		Request: CanaryPercentRequest{}, Response: Result{}},
	{Method: "DELETE", Path: "/api/v1/models/{name}/canary", Summary: "Take a canary out: its share first, then the candidate",
		Response: Result{}},
	{Method: "POST", Path: "/api/v1/models/{name}/promote", Summary: "Make a canary's weights the model's",
		Description: "With a quality suite, or the model's standing gate, the canary is asked it first and kept below the score.",
		Request:     modelserve.Change{}, OptionalBody: true, Response: modelserve.Job{}, Status: http.StatusAccepted},

	// Quality suites.
	{Method: "GET", Path: "/api/v1/eval/suites", Summary: "The quality suites the control plane can run by name",
		Response: []eval.SuiteInfo{}},
	{Method: "POST", Path: "/api/v1/models/{name}/eval", Summary: "Ask a model a quality suite",
		Description: "Started at once and answered with the run, to be followed at /api/v1/eval/runs/{id}.",
		Request:     eval.Request{}, Response: eval.Live{}, Status: http.StatusAccepted},
	{Method: "GET", Path: "/api/v1/eval/runs", Summary: "Quality runs, the newest first",
		Query: []Param{
			{Name: "model", Description: "Only this model's runs."},
			{Name: "limit", Type: "integer", Description: "How many; a sensible number when absent."},
		},
		Response: []eval.Live{}},
	{Method: "GET", Path: "/api/v1/eval/runs/{id}", Summary: "One quality run, with each case",
		Response: eval.Live{}},
	{Method: "PUT", Path: "/api/v1/models/{name}/eval-gate", Summary: "Hold a model's updates and promotions to a quality suite",
		Request: EvalGateRequest{}, Response: EvalGateView{}},
	{Method: "DELETE", Path: "/api/v1/models/{name}/eval-gate", Summary: "Stop holding a model's updates to a suite",
		Response: EvalGateView{}},

	// Federation: other llmgw clusters as providers.
	{Method: "GET", Path: "/api/v1/federation", Summary: "This cluster's id, and its federation peers with their health and models",
		Response: FederationView{}},
	{Method: "POST", Path: "/api/v1/federation", Summary: "Add another llmgw cluster as a peer, or change one",
		Description: "The peer is checked first: its /ready, timed, and its /v1/models with the key. Its models join the catalog as PEER/MODEL.",
		Request:     federation.AddRequest{}, Response: federation.Status{}, Status: http.StatusCreated},
	{Method: "DELETE", Path: "/api/v1/federation/{peer}", Summary: "Remove a federation peer and its models",
		Description: "Refused while a model here falls back to one of the peer's.",
		Response:    Removed{}},
	{Method: "POST", Path: "/api/v1/federation/{peer}/sync", Summary: "Check a peer now, rather than at the next interval",
		Response: federation.Status{}},
	{Method: "PUT", Path: "/api/v1/models/{name}/fallbacks", Summary: "Replace the models a model falls back to when it has no capacity",
		Description: "A model added at runtime only. A federation peer's model, PEER/MODEL, sends the overflow to the other cluster.",
		Request:     FallbacksRequest{}, Response: FallbacksView{}},

	// Tenancy: quotas and guardrails.
	{Method: "GET", Path: "/api/v1/quotas", Summary: "Every capacity quota, with what its scope holds now",
		Response: []QuotaView{}},
	{Method: "PUT", Path: "/api/v1/quotas", Summary: "Set a scope's capacity quota",
		Description: "Replaces the scope's quota whole. The scope may be named by path or by ID; the answer names it by ID.",
		Request:     QuotaRequest{}, Response: QuotaSet{}},
	{Method: "DELETE", Path: "/api/v1/quotas/{scope}", Summary: "Remove a scope's capacity quota",
		Response: Removed{}},
	{Method: "GET", Path: "/api/v1/guard", Summary: "The guardrail policies and where they are attached",
		Response: GuardView{}},
	{Method: "POST", Path: "/api/v1/guard/attachments", Summary: "Attach a guardrail policy",
		Request: GuardAttachmentRequest{}, Response: guard.Attachment{}},
	{Method: "DELETE", Path: "/api/v1/guard/attachments", Summary: "Detach a guardrail policy",
		Request: GuardAttachmentRequest{}, Response: guard.Attachment{}},

	// Usage, requests and the record of changes.
	{Method: "GET", Path: "/api/v1/usage", Summary: "What each group of the tenancy tree spent",
		Query: []Param{
			{Name: "by", Description: "org, team or project; team when absent."},
			{Name: "since", Description: "A duration back from now, like 24h; 24h when absent."},
		},
		Response: []ScopeUsageView{}},
	{Method: "GET", Path: "/api/v1/usage/export", Summary: "Usage totals by period and dimension, as CSV or JSON",
		Query: []Param{
			{Name: "from", Description: "A date (2026-09-01) or an RFC 3339 time; the first of this month when absent."},
			{Name: "to", Description: "A date or an RFC 3339 time; now when absent."},
			{Name: "period", Description: "hour, day or month; day when absent."},
			{Name: "by", Description: "Dimensions, comma-separated: org, team, project, user, key, model."},
			{Name: "format", Description: "json for JSON; CSV otherwise."},
		},
		Response: []store.ExportRow{}, CSV: true},
	{Method: "GET", Path: "/api/v1/requests/{id}", Summary: "What one request went through",
		Response: RequestView{}},
	{Method: "GET", Path: "/api/v1/requests/{id}/body", Summary: "A kept request and its answer",
		Response: store.RequestBody{}},
	{Method: "POST", Path: "/api/v1/requests/{id}/replay", Summary: "Send a kept request again, to any model",
		Request: ReplayRequest{}, Response: ReplayView{}},
	{Method: "GET", Path: "/api/v1/audit", Summary: "Changes made, newest first",
		Query: []Param{
			{Name: "since", Description: "How far back: 90m, 24h, 7d; 7d when absent."},
			{Name: "limit", Type: "integer", Description: "The most events to answer with."},
			{Name: "actor", Description: "Only changes by this actor."},
			{Name: "contains", Description: "Only events whose action or target contains this."},
		},
		Response: []store.AuditEvent{}},

	// Operations.
	{Method: "POST", Path: "/api/v1/notify/test", Summary: "Send a test notification to every webhook",
		Response: []notify.Result{}},
	{Method: "POST", Path: "/api/v1/metrics/range", Summary: "Several Prometheus range queries at once",
		Description: "Read from the Prometheus monitoring.prometheus names; configured is false without one.",
		Request:     MetricsRequest{}, Response: MetricsResult{}},
}

// serveOpenAPI answers with the document generated from Routes, as it was
// when this binary was built.
func serveOpenAPI(w http.ResponseWriter, _ *http.Request) {
	w.Header().Set("Content-Type", "application/json")
	_, _ = w.Write(llmgw.AdminOpenAPI)
}