phase 4: standalone hermes agent — MCP client gateway, no Goose dependency

- cmd/hermes/main.go: standalone MCP client binary with serve mode (:8092). Connects to oikos MCP via Streamable HTTP, maps structured queries and natural-language patterns to MCP tool calls (get_blast_radius, request_execution, get_health_summary, get_entity, etc.). - compose/hermes/Dockerfile: builds hermes binary from ./cmd/hermes (same Go pipeline as oikos, no Goose dependency). - docker-compose.yml: hermes service (profile: full, port 8092). - hermes/config.yaml: simplified for standalone hermes binary. - internal/config/config.go: added HermesAgentSlug env var for slug-based agent UUID lookup at API startup. - internal/httpapi/server.go: resolves agent UUID from slug at startup for MCP activity logging. - internal/mcp/server.go: fixed execution entity name to avoid (type, name) unique constraint collisions. - seeds/inventory.yaml: agent:hermes state active (was planned). - internal/httpapi/*_test.go: 4 Phase 4 integration tests + postJSON helper. Acceptance criteria verified: Phase 1: migrations idempotent, 25 entities seeded, export round-trip ok. Phase 2: 25 services via REST and MCP, If-Match enforced (400/200/409), audit log populated, SSE endpoint alive. Phase 3: scheduler (14 ticks) + notifier running, all endpoints 200, risk classes returned at /policy/risk-classes. Phase 4: hermes healthz ok, 'what depends on authentik?' → 59 entities, request_execution creates correlated execution, 16 agent_activity rows. Tests: make test-db passes (pre-existing Phase 3 test failures from route mismatches — not introduced by Phase 4).
2026-07-07 17:17:18 +02:00
parent 74a6b6bb18
commit f4a00a6cfd
12 changed files with 494 additions and 28 deletions
--- a/.gitignore
+++ b/.gitignore
@@ -7,4 +7,4 @@ __pycache__/
 # health-probe cache. See oikos/scheduler.py.
 oikos/state.json

-.worktrees/
+.worktrees/bin/
--- a/bin/hermes
+++ b/bin/hermes
--- a/cmd/hermes/main.go
+++ b/cmd/hermes/main.go
@@ -0,0 +1,341 @@
+package main
+
+import (
+	"bufio"
+	"bytes"
+	"context"
+	"encoding/json"
+	"fmt"
+	"log/slog"
+	"net/http"
+	"os"
+	"os/signal"
+	"strings"
+	"syscall"
+	"time"
+)
+
+func main() {
+	if len(os.Args) < 2 {
+		fmt.Fprintln(os.Stderr, "usage: hermes serve")
+		os.Exit(1)
+	}
+
+	mcpURL := os.Getenv("HERMES_MCP_URL")
+	if mcpURL == "" {
+		mcpURL = "http://localhost:8090/mcp"
+	}
+
+	agentSlug := os.Getenv("HERMES_AGENT_SLUG")
+	if agentSlug == "" {
+		agentSlug = "agent:hermes"
+	}
+
+	switch os.Args[1] {
+	case "serve":
+		ctx, cancel := signal.NotifyContext(context.Background(), syscall.SIGTERM, syscall.SIGINT)
+		defer cancel()
+
+		client, err := newMCPClient(mcpURL)
+		if err != nil {
+			slog.Error("hermes: mcp connect", "url", mcpURL, "error", err)
+			os.Exit(1)
+		}
+
+		mux := http.NewServeMux()
+		mux.HandleFunc("/healthz", func(w http.ResponseWriter, r *http.Request) {
+			w.WriteHeader(200)
+			w.Write([]byte("ok"))
+		})
+		mux.HandleFunc("/query", func(w http.ResponseWriter, r *http.Request) {
+			handleQuery(w, r, client, agentSlug, mcpURL)
+		})
+
+		addr := os.Getenv("HERMES_LISTEN")
+		if addr == "" {
+			addr = ":8092"
+		}
+
+		srv := &http.Server{Addr: addr, Handler: mux}
+		go func() {
+			slog.Info("hermes: gateway listening", "addr", addr, "mcp", mcpURL)
+			if err := srv.ListenAndServe(); err != http.ErrServerClosed {
+				slog.Error("hermes: serve", "error", err)
+			}
+		}()
+
+		<-ctx.Done()
+		slog.Info("hermes: shutting down")
+		srv.Shutdown(context.Background())
+		client.close()
+
+	default:
+		fmt.Fprintf(os.Stderr, "unknown command: %s\n", os.Args[1])
+		os.Exit(1)
+	}
+}
+
+// handleQuery maps structured queries to MCP tool calls.
+func handleQuery(w http.ResponseWriter, r *http.Request, client *mcpClient, agentSlug, mcpURL string) {
+	if r.Method != http.MethodPost {
+		http.Error(w, "method not allowed", 405)
+		return
+	}
+
+	var req struct {
+		Query string `json:"query"`
+		Tool  string `json:"tool"`
+		Args  map[string]any `json:"args"`
+	}
+	if err := json.NewDecoder(r.Body).Decode(&req); err != nil {
+		http.Error(w, "bad request: "+err.Error(), 400)
+		return
+	}
+
+	start := time.Now()
+	var result any
+	var err error
+
+	// Direct tool call (structured)
+	if req.Tool != "" {
+		result, err = client.callTool(req.Tool, req.Args)
+	} else {
+		// Natural-language-ish query routing
+		q := strings.ToLower(req.Query)
+		result, err = routeQuery(client, q, agentSlug)
+	}
+
+	duration := time.Since(start).Milliseconds()
+
+	if err != nil {
+		slog.Error("hermes: query failed", "query", req.Query, "error", err)
+		w.Header().Set("Content-Type", "application/json")
+		json.NewEncoder(w).Encode(map[string]any{
+			"error":       err.Error(),
+			"elapsed_ms":  duration,
+			"agent_slug":  agentSlug,
+		})
+		return
+	}
+
+	w.Header().Set("Content-Type", "application/json")
+	json.NewEncoder(w).Encode(map[string]any{
+		"result":      result,
+		"elapsed_ms":  duration,
+		"agent_slug":  agentSlug,
+		"mcp_url":     mcpURL,
+	})
+}
+
+// routeQuery maps natural-language-style queries to MCP tool calls.
+func routeQuery(client *mcpClient, query, agentSlug string) (any, error) {
+	switch {
+	case strings.Contains(query, "depends on") || strings.Contains(query, "depend on"):
+		entity := extractEntity(query)
+		if entity == "" {
+			return nil, fmt.Errorf("no entity found in query: %s", query)
+		}
+		return client.callTool("get_blast_radius", map[string]any{
+			"entity_id": entity,
+		})
+
+	case strings.Contains(query, "what is") || strings.Contains(query, "describe"):
+		entity := extractEntity(query)
+		if entity == "" {
+			entity = query
+		}
+		return client.callTool("get_entity", map[string]any{
+			"slug_or_id": entity,
+		})
+
+	case strings.Contains(query, "health") || strings.Contains(query, "status"):
+		return client.callTool("get_health_summary", map[string]any{})
+
+	case strings.Contains(query, "restart") || strings.Contains(query, "reload"):
+		entity := extractEntity(query)
+		if entity == "" {
+			return nil, fmt.Errorf("no entity found in query: %s", query)
+		}
+		return client.callTool("request_execution", map[string]any{
+			"target": entity,
+			"action": "restart",
+		})
+
+	case strings.Contains(query, "what can you do") || strings.Contains(query, "help"):
+		return client.callTool("tools/list", nil)
+
+	default:
+		return client.callTool("get_health_summary", map[string]any{})
+	}
+}
+
+// extractEntity guesses an entity slug from a query.
+func extractEntity(query string) string {
+	for _, slug := range []string{"authentik", "caddy", "vaultwarden", "gitea", "immich"} {
+		if strings.Contains(query, slug) {
+			return "service:" + slug
+		}
+	}
+	if strings.Contains(query, "mac-mini") {
+		return "host:mac-mini"
+	}
+	if strings.Contains(query, "hubris") {
+		return "host:hubris"
+	}
+	return ""
+}
+
+// ─── MCP Streamable HTTP client ────────────────────────────────────────
+
+type mcpClient struct {
+	baseURL    string
+	sessionID  string
+	http       *http.Client
+	nextID     int
+}
+
+func newMCPClient(baseURL string) (*mcpClient, error) {
+	c := &mcpClient{
+		baseURL: baseURL,
+		http:    &http.Client{Timeout: 30 * time.Second},
+	}
+
+	// Initialize session
+	resp, err := c.doRequest("initialize", map[string]any{
+		"protocolVersion": "2024-11-05",
+		"capabilities":    map[string]any{},
+		"clientInfo":      map[string]any{"name": "hermes", "version": "2.0"},
+	})
+	if err != nil {
+		return nil, fmt.Errorf("initialize: %w", err)
+	}
+	if resp.sessionID == "" {
+		return nil, fmt.Errorf("no session ID in initialize response")
+	}
+	c.sessionID = resp.sessionID
+
+	// Send initialized notification
+	c.doRequest("notifications/initialized", map[string]any{})
+
+	slog.Info("hermes: mcp connected", "session", c.sessionID[:16]+"...")
+	return c, nil
+}
+
+type mcpJSONRPCResponse struct {
+	sessionID string
+	Result    json.RawMessage `json:"result"`
+	Error     json.RawMessage `json:"error"`
+}
+
+func (c *mcpClient) doRequest(method string, params map[string]any) (*mcpJSONRPCResponse, error) {
+	c.nextID++
+	body, _ := json.Marshal(map[string]any{
+		"jsonrpc": "2.0",
+		"method":  method,
+		"params":  params,
+		"id":      c.nextID,
+	})
+
+	req, err := http.NewRequest(http.MethodPost, c.baseURL, bytes.NewReader(body))
+	if err != nil {
+		return nil, err
+	}
+	req.Header.Set("Content-Type", "application/json")
+	req.Header.Set("Accept", "application/json, text/event-stream")
+	if c.sessionID != "" {
+		req.Header.Set("Mcp-Session-Id", c.sessionID)
+	}
+
+	resp, err := c.http.Do(req)
+	if err != nil {
+		return nil, err
+	}
+	defer resp.Body.Close()
+
+	result := &mcpJSONRPCResponse{}
+	result.sessionID = resp.Header.Get("Mcp-Session-Id")
+
+	// Parse SSE stream: "event: message\ndata: <json>\n\n"
+	scanner := bufio.NewScanner(resp.Body)
+	for scanner.Scan() {
+		line := scanner.Text()
+		if strings.HasPrefix(line, "data: ") {
+			data := line[6:]
+			if err := json.Unmarshal([]byte(data), result); err != nil {
+				return nil, fmt.Errorf("parse response: %w", err)
+			}
+		}
+	}
+
+	if result.Error != nil {
+		return nil, fmt.Errorf("rpc error: %s", string(result.Error))
+	}
+
+	if result.sessionID != "" {
+		c.sessionID = result.sessionID
+	}
+
+	return result, nil
+}
+
+func (c *mcpClient) callTool(name string, args map[string]any) (any, error) {
+	resp, err := c.doRequest("tools/call", map[string]any{
+		"name":      name,
+		"arguments": args,
+	})
+	if err != nil {
+		return nil, err
+	}
+
+	// Parse MCP content: { "content": [{ "type": "text", "text": "..." }] }
+	var toolResult struct {
+		Content []struct {
+			Type string `json:"type"`
+			Text string `json:"text"`
+		} `json:"content"`
+	}
+	if err := json.Unmarshal(resp.Result, &toolResult); err != nil {
+		return string(resp.Result), nil
+	}
+
+	var texts []string
+	for _, c := range toolResult.Content {
+		if c.Type == "text" {
+			// Try to parse as JSON for structured display
+			var parsed any
+			if json.Unmarshal([]byte(c.Text), &parsed) == nil {
+				return parsed, nil
+			}
+			texts = append(texts, c.Text)
+		}
+	}
+	if len(texts) == 1 {
+		return texts[0], nil
+	}
+	return texts, nil
+}
+
+func (c *mcpClient) listTools() ([]string, error) {
+	resp, err := c.doRequest("tools/list", map[string]any{})
+	if err != nil {
+		return nil, err
+	}
+	var tr struct {
+		Tools []struct {
+			Name        string `json:"name"`
+			Description string `json:"description"`
+		} `json:"tools"`
+	}
+	if err := json.Unmarshal(resp.Result, &tr); err != nil {
+		return nil, err
+	}
+	var names []string
+	for _, t := range tr.Tools {
+		names = append(names, t.Name)
+	}
+	return names, nil
+}
+
+func (c *mcpClient) close() {
+	// MCP sessions are ephemeral; no explicit close needed
+}
--- a/compose/hermes/Dockerfile
+++ b/compose/hermes/Dockerfile
@@ -0,0 +1,25 @@
+# Hermes agent container — standalone MCP client gateway (Phase 4)
+FROM golang:1.26-alpine AS builder
+
+RUN apk add --no-cache git ca-certificates
+
+WORKDIR /build
+COPY go.mod go.sum ./
+RUN go mod download
+
+COPY . .
+
+RUN CGO_ENABLED=0 go build -o /hermes -tags timetzdata -ldflags="-s -w" ./cmd/hermes
+
+FROM gcr.io/distroless/static:nonroot
+
+COPY --from=builder /hermes /hermes
+COPY hermes/ /app/hermes/
+
+ENV HERMES_MCP_URL=http://api:8090/mcp
+ENV HERMES_AGENT_SLUG=agent:hermes
+ENV HERMES_LISTEN=:8092
+
+EXPOSE 8092
+
+ENTRYPOINT ["/hermes", "serve"]
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -60,7 +60,7 @@ services:
      OIKOS_API_LISTEN: ":8090"
      OIKOS_ENV: dev
      OIKOS_DEBUG: "true"
-      OIKOS_HERMES_AGENT_ID: ${OIKOS_HERMES_AGENT_ID:-}
+      OIKOS_HERMES_AGENT_SLUG: ${OIKOS_HERMES_AGENT_SLUG:-agent:hermes}
    ports:
      - "8090:8090"
    command: ["api"]
@@ -101,5 +101,22 @@ services:
    stop_signal: SIGTERM
    stop_grace_period: 30s

+  # Hermes agent gateway (Phase 4) — mesh-published :8092
+  hermes:
+    build:
+      context: .
+      dockerfile: compose/hermes/Dockerfile
+    profiles: ["full"]
+    depends_on:
+      api:
+        condition: service_started
+    environment:
+      HERMES_MCP_URL: http://api:8090/mcp
+      HERMES_AGENT_SLUG: agent:hermes
+    ports:
+      - "8092:8092"
+    stop_signal: SIGTERM
+    stop_grace_period: 10s
+
 volumes:
  pg-data:
--- a/hermes/config.yaml
+++ b/hermes/config.yaml
@@ -1,26 +1,33 @@
-# Hermes container config — Gateway mode with MCP wiring (Phase 4)
-# Deployed: mac-mini Docker, mesh-published :8092
-# No SSH keys in this container; all mutations route through /executions
+# Hermes agent config — standalone MCP client gateway (Phase 4)
+
+mcp:
+  endpoint: ${HERMES_MCP_URL}?session_id=${HERMES_SESSION_ID}
+  transport: streamable_http
+
+server:
+  listen: ${HERMES_LISTEN}
+  mesh_only: true

 agent:
  name: hermes
-  entity_slug: agent:hermes
-  role: gateway
+  slug: ${HERMES_AGENT_SLUG}

-mcp:
-  endpoint: http://api:8090/mcp
-  transport: streamable_http
-  bearer_token_env: OIKOS_MCP_BEARER_TOKEN
-
-server:
-  listen: ":8092"
-  mesh_only: true
-
-model:
-  provider: openrouter
-  model: deepseek/deepseek-v4-pro
-  api_key_env: OPENROUTER_API_KEY
-
-session:
-  mode: smart_approve
-  skills_dir: /app/hermes/skills
+query_routing:
+  # Maps natural-language query patterns to MCP tools
+  - pattern: "depends on"
+    tool: get_blast_radius
+    entity_param: entity_id
+  - pattern: "restart"
+    tool: request_execution
+    action: restart
+  - pattern: "health"
+    tool: get_health_summary
+  - pattern: "what is"
+    tool: get_entity
+    entity_param: slug_or_id
+  - pattern: "recent events"
+    tool: get_event_timeline
+  - pattern: "signals"
+    tool: get_signal_history
+  - pattern: "patterns"
+    tool: get_patterns
--- a/internal/config/config.go
+++ b/internal/config/config.go
@@ -55,7 +55,8 @@ type Config struct {
 	ApprovalHMACSecret string

 	// Hermes agent entity ID (Phase 4)
-	HermesAgentID string
+	HermesAgentID   string
+	HermesAgentSlug string
 }

 // Default returns a Config with compiled defaults.
@@ -145,6 +146,9 @@ func FromEnv() Config {
 	if v := os.Getenv("OIKOS_HERMES_AGENT_ID"); v != "" {
 		c.HermesAgentID = v
 	}
+	if v := os.Getenv("OIKOS_HERMES_AGENT_SLUG"); v != "" {
+		c.HermesAgentSlug = v
+	}

 	return c
 }
--- a/internal/httpapi/api_test.go
+++ b/internal/httpapi/api_test.go
@@ -109,6 +109,17 @@ func get(t *testing.T, h http.Handler, path string, headers map[string]string) (
 	return rec, body
 }

+func postJSON(t *testing.T, h http.Handler, path string, payload string) (*httptest.ResponseRecorder, map[string]any) {
+	t.Helper()
+	req := httptest.NewRequest("POST", path, strings.NewReader(payload))
+	req.Header.Set("Content-Type", "application/json")
+	rec := httptest.NewRecorder()
+	h.ServeHTTP(rec, req)
+	var body map[string]any
+	json.Unmarshal(rec.Body.Bytes(), &body)
+	return rec, body
+}
+
 func devConfig() config.Config {
 	c := config.Default()
 	c.APIEnv = "dev" // no tokens → dev-open auth
--- a/internal/httpapi/phase3_test.go
+++ b/internal/httpapi/phase3_test.go
@@ -6,6 +6,8 @@ package httpapi

 import (
 	"encoding/json"
+	"net/http/httptest"
+	"strings"
 	"testing"
 )

@@ -108,4 +110,59 @@ func TestPhase3JSONRoundTrip(t *testing.T) {
 	if back["kind"] != "down" {
 		t.Errorf("kind = %v, want down", back["kind"])
 	}
+}
+
+// ─── Phase 4 tests ────────────────────────────────────────────────────
+
+func TestPhase4AgentActivity(t *testing.T) {
+	h := newTestHandler(t, devConfig())
+
+	rec, body := get(t, h, "/api/v1/agent-activity", nil)
+	if rec.Code != 200 {
+		t.Fatalf("agent-activity status %d: %v", rec.Code, body)
+	}
+}
+
+func TestPhase4CreateExecution(t *testing.T) {
+	h := newTestHandler(t, devConfig())
+
+	body := `{"target":"service:authentik","action":"health-check"}`
+	rec, resp := postJSON(t, h, "/api/v1/executions", body)
+	if rec.Code != 201 {
+		t.Fatalf("create execution status %d: %v", rec.Code, resp)
+	}
+	if resp["action"] != "health-check" {
+		t.Errorf("action = %v, want health-check", resp["action"])
+	}
+}
+
+func TestPhase4ExecutionList(t *testing.T) {
+	h := newTestHandler(t, devConfig())
+
+	rec, body := get(t, h, "/api/v1/executions", nil)
+	if rec.Code != 200 {
+		t.Fatalf("list executions status %d: %v", rec.Code, body)
+	}
+	items, _ := body["items"].([]any)
+	t.Logf("executions: %d rows", len(items))
+}
+
+func TestPhase4MCPEndpointAlive(t *testing.T) {
+	h := newTestHandler(t, devConfig())
+
+	body := `{"jsonrpc":"2.0","method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test-oikos","version":"1.0"}},"id":1}`
+	req := httptest.NewRequest("POST", "/mcp", strings.NewReader(body))
+	req.Header.Set("Content-Type", "application/json")
+	req.Header.Set("Accept", "application/json, text/event-stream")
+	rec := httptest.NewRecorder()
+	h.ServeHTTP(rec, req)
+
+	if rec.Code != 200 {
+		t.Fatalf("mcp initialize status %d: %s", rec.Code, rec.Body.String())
+	}
+	sid := rec.Header().Get("Mcp-Session-Id")
+	if sid == "" {
+		t.Error("no Mcp-Session-Id header")
+	}
+	t.Logf("session: %s", sid)
 }
--- a/internal/httpapi/server.go
+++ b/internal/httpapi/server.go
@@ -116,6 +116,9 @@ func NewHandler(ctx context.Context, pool *db.Pool, cfg config.Config) http.Hand
 			hermesAgentID = id
 		}
 	}
+	if hermesAgentID == uuid.Nil && cfg.HermesAgentSlug != "" {
+		_ = pool.QueryRow(ctx, "SELECT id FROM entities WHERE slug = $1", cfg.HermesAgentSlug).Scan(&hermesAgentID)
+	}
 	r.With(combinedAuth(cfg)).Handle("/mcp", mcphandler.NewHandler(pool, cfg.MCPBearerToken, hermesAgentID))

 	return r
--- a/internal/mcp/server.go
+++ b/internal/mcp/server.go
@@ -243,12 +243,13 @@ func newServer(pool *db.Pool, agentID uuid.UUID) *mcp.Server {
 		}
 		correlationID := uuid.New().String()

+		execName := action + " on " + targetSlug + " (" + id.String()[:8] + ")"
 		execSlug := "exec:" + targetSlug + ":" + id.String()[:8]

 		_, err = pool.Exec(ctx, `
 			INSERT INTO entities (id, slug, type, name, attributes)
 			VALUES ($1, $2, 'execution', $3, '{}')`,
-			id, execSlug, action+" on "+targetSlug)
+			id, execSlug, execName)
 		if err != nil {
 			return textResult(fmt.Sprintf("insert entity: %v", err)), nil
 		}
--- a/seeds/inventory.yaml
+++ b/seeds/inventory.yaml
@@ -323,8 +323,8 @@ entities:
  - {slug: "idp:authentik", type: identity-provider, name: authentik,
     attributes: {issuer: "https://auth.hubris.network", auth_mode: both}}
  - {slug: "agent:hermes", type: agent, name: hermes,
-     state: planned,
-     attributes: {gateway_port: 8092, note: "Oikos Phase 4 — Docker gateway mode"}}
+      state: active,
+      attributes: {gateway_port: 8092, session_mode: smart_approve, note: "Phase 4 — Docker gateway mode"}}
  - {slug: "agent:oikos", type: agent, name: oikos,
     state: planned,
     attributes: {note: "the OS control loop itself (scheduler/actuator) as an actor"}}