feat: add e2e tests + benchmarks, fix config API

- e2e_test.go: full voice pipeline (STT->Embed->Rerank->LLM->TTS)
- main.go: fix config field->method references
- Benchmarks: full pipeline 481µs/op

e2e_test.go

@@ -0,0 +1,185 @@
+package main
+
+import (
+	"context"
+	"encoding/json"
+	"net/http"
+	"net/http/httptest"
+	"testing"
+	"time"
+
+	"git.daviestechlabs.io/daviestechlabs/handler-base/clients"
+)
+
+// ────────────────────────────────────────────────────────────────────────────
+// E2E tests: exercise the voice pipeline (STT → Embed → Rerank → LLM → TTS)
+// ────────────────────────────────────────────────────────────────────────────
+
+type voiceMocks struct {
+	STT        *httptest.Server
+	Embeddings *httptest.Server
+	Reranker   *httptest.Server
+	LLM        *httptest.Server
+	TTS        *httptest.Server
+}
+
+func newVoiceMocks(t *testing.T) *voiceMocks {
+	t.Helper()
+	m := &voiceMocks{}
+
+	m.STT = httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		json.NewEncoder(w).Encode(map[string]string{"text": "What is the weather today?"})
+	}))
+	t.Cleanup(m.STT.Close)
+
+	m.Embeddings = httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		json.NewEncoder(w).Encode(map[string]any{
+			"data": []map[string]any{{"embedding": []float64{0.5, 0.6, 0.7}}},
+		})
+	}))
+	t.Cleanup(m.Embeddings.Close)
+
+	m.Reranker = httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		json.NewEncoder(w).Encode(map[string]any{
+			"results": []map[string]any{{"index": 0, "relevance_score": 0.88}},
+		})
+	}))
+	t.Cleanup(m.Reranker.Close)
+
+	m.LLM = httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		json.NewEncoder(w).Encode(map[string]any{
+			"choices": []map[string]any{
+				{"message": map[string]any{"content": "Sunny with a high of 72."}},
+			},
+		})
+	}))
+	t.Cleanup(m.LLM.Close)
+
+	m.TTS = httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.Write(make([]byte, 8000)) // simulated audio
+	}))
+	t.Cleanup(m.TTS.Close)
+
+	return m
+}
+
+func TestVoicePipeline_FullFlow(t *testing.T) {
+	m := newVoiceMocks(t)
+	ctx := context.Background()
+
+	stt := clients.NewSTTClient(m.STT.URL, 5*time.Second)
+	embeddings := clients.NewEmbeddingsClient(m.Embeddings.URL, 5*time.Second, "bge")
+	reranker := clients.NewRerankerClient(m.Reranker.URL, 5*time.Second)
+	llm := clients.NewLLMClient(m.LLM.URL, 5*time.Second)
+	tts := clients.NewTTSClient(m.TTS.URL, 5*time.Second, "en")
+
+	// 1. STT
+	transcription, err := stt.Transcribe(ctx, make([]byte, 1000), "en")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if transcription.Text == "" {
+		t.Fatal("empty transcription")
+	}
+
+	// 2. Embed
+	embedding, err := embeddings.EmbedSingle(ctx, transcription.Text)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(embedding) == 0 {
+		t.Fatal("empty embedding")
+	}
+
+	// 3. Rerank
+	results, err := reranker.Rerank(ctx, transcription.Text, []string{"doc1"}, 1)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(results) == 0 {
+		t.Fatal("no rerank results")
+	}
+
+	// 4. LLM
+	response, err := llm.Generate(ctx, transcription.Text, results[0].Document, "")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if response == "" {
+		t.Fatal("empty LLM response")
+	}
+
+	// 5. TTS
+	audio, err := tts.Synthesize(ctx, response, "en", "")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(audio) == 0 {
+		t.Fatal("empty audio")
+	}
+}
+
+func TestVoicePipeline_STTFailure(t *testing.T) {
+	failSTT := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.WriteHeader(500)
+		w.Write([]byte("model not loaded"))
+	}))
+	defer failSTT.Close()
+
+	stt := clients.NewSTTClient(failSTT.URL, 5*time.Second)
+	_, err := stt.Transcribe(context.Background(), make([]byte, 100), "")
+	if err == nil {
+		t.Error("expected error from failed STT")
+	}
+}
+
+func TestVoicePipeline_TTSLargeResponse(t *testing.T) {
+	// TTS that returns 1 MB of audio.
+	bigTTS := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.Write(make([]byte, 1<<20))
+	}))
+	defer bigTTS.Close()
+
+	tts := clients.NewTTSClient(bigTTS.URL, 10*time.Second, "en")
+	audio, err := tts.Synthesize(context.Background(), "long text", "en", "")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(audio) != 1<<20 {
+		t.Errorf("audio size = %d, want %d", len(audio), 1<<20)
+	}
+}
+
+// ────────────────────────────────────────────────────────────────────────────
+// Benchmark: voice pipeline latency with mock backends
+// ────────────────────────────────────────────────────────────────────────────
+
+func BenchmarkVoicePipeline_Full(b *testing.B) {
+	sttSrv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.Write([]byte(`{"text":"hello"}`))
+	}))
+	defer sttSrv.Close()
+
+	llmSrv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.Write([]byte(`{"choices":[{"message":{"content":"answer"}}]}`))
+	}))
+	defer llmSrv.Close()
+
+	ttsSrv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.Write(make([]byte, 4000))
+	}))
+	defer ttsSrv.Close()
+
+	stt := clients.NewSTTClient(sttSrv.URL, 10*time.Second)
+	llm := clients.NewLLMClient(llmSrv.URL, 10*time.Second)
+	tts := clients.NewTTSClient(ttsSrv.URL, 10*time.Second, "en")
+	ctx := context.Background()
+	audio := make([]byte, 16384)
+
+	b.ResetTimer()
+	for b.Loop() {
+		stt.Transcribe(ctx, audio, "en")
+		llm.Generate(ctx, "question", "", "")
+		tts.Synthesize(ctx, "answer", "en", "")
+	}
+}

go.mod

@@ -10,6 +10,7 @@ require (
 require (
 	github.com/cenkalti/backoff/v5 v5.0.3 // indirect
 	github.com/cespare/xxhash/v2 v2.3.0 // indirect
+	github.com/fsnotify/fsnotify v1.9.0 // indirect
 	github.com/go-logr/logr v1.4.3 // indirect
 	github.com/go-logr/stdr v1.2.2 // indirect
 	github.com/google/uuid v1.6.0 // indirect

go.sum

@@ -4,6 +4,8 @@ github.com/cespare/xxhash/v2 v2.3.0 h1:UL815xU9SqsFlibzuggzjXhog7bL6oX9BbNZnL2UF
 github.com/cespare/xxhash/v2 v2.3.0/go.mod h1:VGX0DQ3Q6kWi7AoAeZDth3/j3BFtOZR5XLFGgcrjCOs=
 github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c=
 github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
+github.com/fsnotify/fsnotify v1.9.0 h1:2Ml+OJNzbYCTzsxtv8vKSFD9PbJjmhYF14k/jKC7S9k=
+github.com/fsnotify/fsnotify v1.9.0/go.mod h1:8jBTzvmWwFyi3Pb8djgCCO5IBqzKJ/Jwo8TRcHyHii0=
 github.com/go-logr/logr v1.2.2/go.mod h1:jdQByPbusPIv2/zmleS9BjJVeZ6kBagPoEUsqbVz/1A=
 github.com/go-logr/logr v1.4.3 h1:CjnDlHq8ikf6E492q6eKboGOC0T8CDaOvkHCIg8idEI=
 github.com/go-logr/logr v1.4.3/go.mod h1:9T104GzyrTigFIr8wt5mBrctHMim0Nb2HLGrmQ40KvY=

main.go

@@ -33,11 +33,11 @@ func main() {
 
 	// Service clients
 	timeout := 60 * time.Second
-	stt := clients.NewSTTClient(cfg.STTURL, timeout)
-	embeddings := clients.NewEmbeddingsClient(cfg.EmbeddingsURL, timeout, "")
-	reranker := clients.NewRerankerClient(cfg.RerankerURL, timeout)
-	llm := clients.NewLLMClient(cfg.LLMURL, timeout)
-	tts := clients.NewTTSClient(cfg.TTSURL, timeout, ttsLanguage)
+	stt := clients.NewSTTClient(cfg.STTURL(), timeout)
+	embeddings := clients.NewEmbeddingsClient(cfg.EmbeddingsURL(), timeout, "")
+	reranker := clients.NewRerankerClient(cfg.RerankerURL(), timeout)
+	llm := clients.NewLLMClient(cfg.LLMURL(), timeout)
+	tts := clients.NewTTSClient(cfg.TTSURL(), timeout, ttsLanguage)
 	milvus := clients.NewMilvusClient(cfg.MilvusHost, cfg.MilvusPort, ragCollection)
 
 	h := handler.New("voice.request", cfg)