internal/server,shared: support request metadata (#850)

- add support for http handlers in the request chain to append metadata
to the request
- metrics middleware will include metadata in the activity log 
- update Activity UI to support metadata, drag sort columns
- update Activity UI capture dialog to use more screen space

Updates #834

internal/server/api.go

@@ -271,7 +271,7 @@ func (s *Server) startPreload() {
 			if err != nil {
 				continue
 			}
-			req = req.WithContext(shared.SetContext(req.Context(), shared.ReqContextData{Model: modelID, ModelID: modelID}))
+			req = req.WithContext(shared.SetContext(req.Context(), shared.ReqContextData{Model: modelID, ModelID: modelID, Metadata: make(map[string]string)}))
 
 			dw := &discardResponseWriter{status: http.StatusOK}
 			s.local.ServeHTTP(dw, req)
@@ -338,7 +338,7 @@ func (s *Server) handleUpstream(w http.ResponseWriter, r *http.Request) {
 	// Strip the /upstream/<model> prefix before forwarding.
 	r.URL.Path = remainingPath
 	// Pin the resolved model so the router skips body/query extraction.
-	*r = *r.WithContext(shared.SetContext(r.Context(), shared.ReqContextData{Model: searchName, ModelID: modelID}))
+	*r = *r.WithContext(shared.SetContext(r.Context(), shared.ReqContextData{Model: searchName, ModelID: modelID, Metadata: make(map[string]string)}))
 
 	switch {
 	case s.local.Handles(modelID):

internal/server/metrics.go

@@ -33,15 +33,16 @@ type TokenMetrics struct {
 
 // ActivityLogEntry represents parsed token statistics from llama-server logs.
 type ActivityLogEntry struct {
-	ID              int          `json:"id"`
-	Timestamp       time.Time    `json:"timestamp"`
-	Model           string       `json:"model"`
-	ReqPath         string       `json:"req_path"`
-	RespContentType string       `json:"resp_content_type"`
-	RespStatusCode  int          `json:"resp_status_code"`
-	Tokens          TokenMetrics `json:"tokens"`
-	DurationMs      int          `json:"duration_ms"`
-	HasCapture      bool         `json:"has_capture"`
+	ID              int               `json:"id"`
+	Timestamp       time.Time         `json:"timestamp"`
+	Model           string            `json:"model"`
+	ReqPath         string            `json:"req_path"`
+	RespContentType string            `json:"resp_content_type"`
+	RespStatusCode  int               `json:"resp_status_code"`
+	Tokens          TokenMetrics      `json:"tokens"`
+	DurationMs      int               `json:"duration_ms"`
+	HasCapture      bool              `json:"has_capture"`
+	Metadata        map[string]string `json:"metadata,omitempty"`
 }
 
 // ActivityLogEvent carries a single activity log entry to event subscribers.
@@ -135,6 +136,13 @@ func (mp *metricsMonitor) record(modelID string, r *http.Request, recorder *resp
 		DurationMs:      int(time.Since(recorder.StartTime()).Milliseconds()),
 	}
 
+	if ctxData, ok := shared.ReadContext(r.Context()); ok && len(ctxData.Metadata) > 0 {
+		tm.Metadata = make(map[string]string, len(ctxData.Metadata))
+		for k, v := range ctxData.Metadata {
+			tm.Metadata[k] = v
+		}
+	}
+
 	queueAndEmit := func() {
 		tm.ID = mp.queueMetrics(tm)
 		mp.emitMetric(tm)

internal/server/metrics_test.go

@@ -1,9 +1,13 @@
 package server
 
 import (
+	"net/http"
+	"net/http/httptest"
+	"strings"
 	"testing"
 	"time"
 
+	"github.com/mostlygeek/llama-swap/internal/shared"
 	"github.com/tidwall/gjson"
 )
 
@@ -56,6 +60,33 @@ func TestServer_ProcessStreamingResponse_NoData(t *testing.T) {
 	}
 }
 
+func TestMetricsMonitor_RecordMetadata(t *testing.T) {
+	mm := newMetricsMonitor(nil, 10, 0)
+	r := httptest.NewRequest(http.MethodPost, "/v1/chat/completions", strings.NewReader(`{"usage":{}}`))
+	r = r.WithContext(shared.SetContext(r.Context(), shared.ReqContextData{
+		ModelID:  "m",
+		Metadata: map[string]string{"client": "web", "trace": "abc"},
+	}))
+
+	w := httptest.NewRecorder()
+	copier := newBodyCopier(w)
+	copier.WriteHeader(http.StatusOK)
+	copier.Write([]byte(`{"usage":{"prompt_tokens":1,"completion_tokens":2}}`))
+
+	mm.record("m", r, copier, 0, nil, nil)
+
+	entries := mm.getMetrics()
+	if len(entries) != 1 {
+		t.Fatalf("want 1 entry, got %d", len(entries))
+	}
+	if entries[0].Metadata["client"] != "web" {
+		t.Errorf("client = %q, want web", entries[0].Metadata["client"])
+	}
+	if entries[0].Metadata["trace"] != "abc" {
+		t.Errorf("trace = %q, want abc", entries[0].Metadata["trace"])
+	}
+}
+
 func TestServer_ParseMetrics_Infill(t *testing.T) {
 	// /infill responses are arrays; timings live in the last element.
 	body := `[{"content":"a"},{"content":"b","timings":{"prompt_n":5,"predicted_n":9,"prompt_ms":10,"predicted_ms":20}}]`