Files
Memoh/internal/memory/indexer_test.go
T
BBQ 6aebbe9279 feat: refactor User/Bot architecture and implement multi-channel gateway
Major changes:
1. Core Architecture: Decoupled Bots from Users. Bots now have independent lifecycles, member management (bot_members), and dedicated configurations.
2. Channel Gateway:
   - Implemented a unified Channel Manager supporting Feishu, Telegram, and Local (Web/CLI) adapters.
   - Added message processing pipeline to normalize interactions across different platforms.
   - Introduced a Contact system for identity binding and guest access policies.
3. Database & Tooling:
   - Consolidated all migrations into 0001_init with updated schema for bots, channels, and contacts.
   - Optimized sqlc.yaml to automatically track the migrations directory.
4. Agent Enhancements:
   - Introduced ToolContext to provide Agents with platform-aware execution capabilities (e.g., messaging, contact lookups).
   - Added tool logging and fallback mechanisms for toolChoice execution.
5. UI & Docs: Updated frontend stores, UI components, and Swagger documentation to align with the new Bot-centric model.
2026-02-04 23:49:50 +08:00

151 lines
4.2 KiB
Go

package memory
import (
"reflect"
"testing"
)
func TestBM25Indexer_TermFrequencies(t *testing.T) {
indexer := NewBM25Indexer(nil)
tests := []struct {
name string
lang string
text string
want map[string]int
docLen int
wantErr bool
}{
{
name: "English text",
lang: "en",
text: "The quick brown fox jumps over the lazy dog",
// Note: Bleve English analyzer stems words (jumps -> jump, lazy -> lazi) and removes stop words (the, over)
want: map[string]int{"quick": 1, "brown": 1, "fox": 1, "jump": 1, "lazi": 1, "dog": 1},
docLen: 6,
},
{
name: "CJK text",
lang: "cjk",
text: "你好世界",
// Note: Bleve CJK analyzer uses bigrams
want: map[string]int{"你好": 1, "好世": 1, "世界": 1},
docLen: 3,
},
{
name: "Mixed text with standard analyzer",
lang: "",
text: "Go 语言 123",
// Note: Standard analyzer splits CJK characters individually
want: map[string]int{"go": 1, "语": 1, "言": 1, "123": 1},
docLen: 4,
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
got, gotLen, err := indexer.TermFrequencies(tt.lang, tt.text)
if (err != nil) != tt.wantErr {
t.Errorf("TermFrequencies() error = %v, wantErr %v", err, tt.wantErr)
return
}
if !reflect.DeepEqual(got, tt.want) {
t.Errorf("TermFrequencies() got = %v, want %v", got, tt.want)
}
if gotLen != tt.docLen {
t.Errorf("TermFrequencies() gotLen = %v, want %v", gotLen, tt.docLen)
}
})
}
}
func TestBM25Indexer_BM25Logic(t *testing.T) {
indexer := NewBM25Indexer(nil)
// 1. 添加一个包含 "golang" 的文档
lang := "en"
tf1 := map[string]int{"golang": 1, "programming": 1}
len1 := 2
indices1, values1 := indexer.AddDocument(lang, tf1, len1)
// 2. 添加另一个包含 "golang" 但更长的文档
tf2 := map[string]int{"golang": 1, "tutorial": 1, "advanced": 1, "topics": 1}
len2 := 4
indices2, values2 := indexer.AddDocument(lang, tf2, len2)
// 验证:在 BM25 中,相同词项在短文档中的权重应该比在长文档中高(惩罚长文档)
var weight1, weight2 float32
for i, idx := range indices1 {
if idx == termHash("golang") {
weight1 = values1[i]
}
}
for i, idx := range indices2 {
if idx == termHash("golang") {
weight2 = values2[i]
}
}
if weight1 <= weight2 {
t.Errorf("Expected weight in shorter doc (%f) to be higher than in longer doc (%f)", weight1, weight2)
}
// 3. 添加一个不包含 "golang" 的文档,增加文档总数,验证 IDF 变化
// IDF 应该随着包含该词的文档比例减少而增加
oldWeight1 := weight1
indexer.AddDocument(lang, map[string]int{"rust": 1}, 1)
indices3, values3 := indexer.AddDocument(lang, tf1, len1) // 再次生成相同文档的向量
for i, idx := range indices3 {
if idx == termHash("golang") {
weight1 = values3[i]
}
}
if weight1 <= oldWeight1 {
t.Errorf("Expected weight to increase as IDF increases (more docs without the term), got %f -> %f", oldWeight1, weight1)
}
}
func TestBM25Indexer_RemoveDocument(t *testing.T) {
indexer := NewBM25Indexer(nil)
lang := "en"
term := "test"
// 添加文档
tf, docLen, _ := indexer.TermFrequencies(lang, term)
indexer.AddDocument(lang, tf, docLen)
indexer.mu.RLock()
stats := indexer.stats["en"]
if stats.DocCount != 1 || stats.DocFreq[term] != 1 {
t.Errorf("Expected stats to be updated after add, got count=%d, freq=%d", stats.DocCount, stats.DocFreq[term])
}
indexer.mu.RUnlock()
// 删除文档
indexer.RemoveDocument(lang, tf, docLen)
indexer.mu.RLock()
if stats.DocCount != 0 || stats.DocFreq[term] != 0 {
t.Errorf("Expected stats to be cleared after remove, got count=%d, freq=%d", stats.DocCount, stats.DocFreq[term])
}
indexer.mu.RUnlock()
}
func TestTermHash_CollisionResistance(t *testing.T) {
// 验证不同词项生成的哈希索引在 20bit 空间内是否分布合理(简单检查不冲突)
h1 := termHash("apple")
h2 := termHash("orange")
h3 := termHash("banana")
if h1 == h2 || h2 == h3 || h1 == h3 {
t.Errorf("Detected unexpected hash collision in small sample: %d, %d, %d", h1, h2, h3)
}
// 验证掩码是否生效
if h1 > sparseDimMask {
t.Errorf("Hash %d exceeds mask %d", h1, sparseDimMask)
}
}