test: Phase 6 全功能测试 — 19个测试全部通过 + 开发路线文档

- host: 沙箱执行/命令拦截/超时/文件读写/系统信息/路径验证 (6 tests) - rag: 文本分块/余弦相似度/关键词匹配/文档索引+搜索 (4 tests) - tools: host_exec/host_file/host_system/knowledge_search/knowledge_ingest (5 tests) - vision: 图片编码/错误处理/定义验证/执行流程 (4 tests) - Embedder 重构为接口，支持 API 和 Simple 两种实现 - 添加 ROADMAP.md 未来开发路线 Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-24 07:51:07 +08:00
parent edc20170b9
commit 63a8f95de1
7 changed files with 640 additions and 9 deletions
@@ -10,8 +10,15 @@ import (
 	"time"
 )

-// Embedder creates text embeddings using an LLM API.
-type Embedder struct {
+// Embedder is the interface for text embedding.
+type Embedder interface {
+	Embed(ctx context.Context, text string) ([]float64, error)
+	EmbedBatch(ctx context.Context, texts []string) ([]float64, error)
+	IsAvailable() bool
+}
+
+// APIEmbedder creates text embeddings using an LLM API.
+type APIEmbedder struct {
 	baseURL    string
 	apiKey     string
 	model      string
@@ -19,8 +26,8 @@ type Embedder struct {
 }

 // NewEmbedder creates a new embedding service.
-func NewEmbedder(baseURL, apiKey, model string) *Embedder {
-	return &Embedder{
+func NewEmbedder(baseURL, apiKey, model string) *APIEmbedder {
+	return &APIEmbedder{
 		baseURL: baseURL,
 		apiKey:  apiKey,
 		model:   model,
@@ -58,12 +65,12 @@ type embeddingError struct {
 }

 // Embed generates an embedding vector for the given text.
-func (e *Embedder) Embed(ctx context.Context, text string) ([]float64, error) {
+func (e *APIEmbedder) Embed(ctx context.Context, text string) ([]float64, error) {
 	return e.EmbedBatch(ctx, []string{text})
 }

 // EmbedBatch generates embeddings for multiple texts.
-func (e *Embedder) EmbedBatch(ctx context.Context, texts []string) ([]float64, error) {
+func (e *APIEmbedder) EmbedBatch(ctx context.Context, texts []string) ([]float64, error) {
 	if !e.IsAvailable() {
 		return nil, fmt.Errorf("embedding service not available: no API key configured")
 	}
@@ -113,6 +120,6 @@ func (e *Embedder) EmbedBatch(ctx context.Context, texts []string) ([]float64, e
 }

 // IsAvailable checks if the embedding service is configured.
-func (e *Embedder) IsAvailable() bool {
+func (e *APIEmbedder) IsAvailable() bool {
 	return e.apiKey != "" && e.baseURL != ""
 }
@@ -34,12 +34,12 @@ type SearchResult struct {
 type KnowledgeStore struct {
 	mu         sync.RWMutex
 	chunks     []Chunk
-	embedder   *Embedder
+	embedder   Embedder
 	knowledgeDir string
 }

 // NewKnowledgeStore creates a new knowledge store.
-func NewKnowledgeStore(embedder *Embedder, knowledgeDir string) *KnowledgeStore {
+func NewKnowledgeStore(embedder Embedder, knowledgeDir string) *KnowledgeStore {
 	if knowledgeDir == "" {
 		knowledgeDir = "./data/knowledge"
 	}
@@ -0,0 +1,155 @@
+package rag
+
+import (
+	"context"
+	"os"
+	"path/filepath"
+	"testing"
+)
+
+func TestChunkText(t *testing.T) {
+	text := "Hello World! This is a test document for chunking. "
+	// Make it longer to trigger chunking
+	longText := ""
+	for i := 0; i < 100; i++ {
+		longText += text
+	}
+
+	chunks := chunkText(longText, 512, 128)
+	if len(chunks) < 2 {
+		t.Fatalf("expected at least 2 chunks, got %d (len=%d)", len(chunks), len(longText))
+	}
+	t.Logf("chunking OK: %d chunks from %d chars", len(chunks), len(longText))
+
+	// Verify overlap: each chunk should have some overlap with next
+	for i := 1; i < len(chunks); i++ {
+		prev := chunks[i-1]
+		if len(prev) == 0 {
+			t.Fatalf("empty chunk at index %d", i-1)
+		}
+	}
+}
+
+func TestCosineSimilarity(t *testing.T) {
+	a := []float64{0.5, 0.3, 0.8, 0.1}
+	b := []float64{0.5, 0.3, 0.8, 0.1}
+	sim := cosineSimilarity(a, b)
+	if sim < 0.99 {
+		t.Fatalf("expected similarity ~1.0 for identical vectors, got %f", sim)
+	}
+
+	c := []float64{-0.5, -0.3, -0.8, -0.1}
+	sim2 := cosineSimilarity(a, c)
+	if sim2 > -0.99 {
+		t.Fatalf("expected similarity ~-1.0 for opposite vectors, got %f", sim2)
+	}
+
+	d := []float64{0.0, 0.0, 0.0, 0.0}
+	sim3 := cosineSimilarity(a, d)
+	if sim3 != 0.0 {
+		t.Fatalf("expected 0.0 for zero vector, got %f", sim3)
+	}
+
+	// Different lengths
+	sim4 := cosineSimilarity(a, []float64{0.5})
+	if sim4 != 0.0 {
+		t.Fatalf("expected 0.0 for different length vectors, got %f", sim4)
+	}
+	t.Logf("cosine similarity OK")
+}
+
+func TestKeywordMatchScore(t *testing.T) {
+	score := keywordMatchScore("hello world", "hello cyrene world of AI")
+	if score < 0.0 || score > 1.0 {
+		t.Fatalf("score out of range: %f", score)
+	}
+	t.Logf("keyword match OK: score=%f", score)
+}
+
+func TestKnowledgeStoreIngestAndSearch(t *testing.T) {
+	// Create temp dir
+	tmpDir, err := os.MkdirTemp("", "cyrene-rag-test")
+	if err != nil {
+		t.Fatalf("create temp dir: %v", err)
+	}
+	defer os.RemoveAll(tmpDir)
+
+	// Write a test document
+	docPath := filepath.Join(tmpDir, "test.md")
+	content := `# Cyrene AI 测试文档
+
+Cyrene 是一个智能 AI 助手，支持语音识别、视觉理解、知识检索等功能。
+
+## 主要功能
+
+1. 多模型目的路由
+2. 宿主机安全操控
+3. 视觉理解与 OCR
+4. 知识库 RAG 检索
+
+## 技术栈
+
+Go 语言编写的后端服务，React 前端。支持多种 LLM 提供商。`
+	if err := os.WriteFile(docPath, []byte(content), 0644); err != nil {
+		t.Fatalf("write test doc: %v", err)
+	}
+
+	// Use SimpleEmbedder for testing (no API key needed)
+	embedder := &SimpleEmbedder{}
+	store := NewKnowledgeStore(embedder, tmpDir)
+
+	ctx := context.Background()
+	n, err := store.IngestFile(ctx, docPath)
+	if err != nil {
+		t.Fatalf("ingest failed: %v", err)
+	}
+	if n == 0 {
+		t.Fatal("expected at least 1 chunk")
+	}
+	t.Logf("ingest OK: %d chunks indexed from %s", n, docPath)
+
+	// Search
+	results, err := store.Search(ctx, "视觉理解 OCR", 3)
+	if err != nil {
+		t.Fatalf("search failed: %v", err)
+	}
+	t.Logf("search OK: %d results for '视觉理解 OCR'", len(results))
+	for _, r := range results {
+		t.Logf("  - %s (score=%.4f): %.50s...", r.Chunk.DocTitle, r.Score, r.Chunk.Content)
+	}
+
+	// Test stats
+	stats := store.Stats()
+	if stats["total_chunks"].(int) != n {
+		t.Fatalf("stats mismatch: expected %d chunks, got %v", n, stats["total_chunks"])
+	}
+	t.Logf("stats OK: %v", stats)
+}
+
+// SimpleEmbedder for testing without API calls.
+type SimpleEmbedder struct{}
+
+func (e *SimpleEmbedder) Embed(ctx context.Context, text string) ([]float64, error) {
+	vec := make([]float64, 128)
+	runes := []rune(text)
+	for i, r := range runes {
+		idx := int(r) % 128
+		vec[idx] += 1.0 / float64(len(runes))
+		posIdx := (int(r) + i) % 128
+		vec[posIdx] += 0.5 / float64(len(runes))
+	}
+	return vec, nil
+}
+
+func (e *SimpleEmbedder) EmbedBatch(ctx context.Context, texts []string) ([]float64, error) {
+	// For batch, embed the concatenation
+	combined := ""
+	for _, t := range texts {
+		combined += t
+	}
+	return e.Embed(ctx, combined)
+}
+
+func (e *SimpleEmbedder) IsAvailable() bool {
+	return true
+}