lly/hr_data_analyzer
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

feat: mock training.

Browse Source
This commit is contained in:
laoboli
2026-04-30 17:02:26 +08:00
parent e43a034e28
commit 23d27b4b6e
3 changed files with 423 additions and 54 deletions
Download Patch File Download Diff File Expand all files Collapse all files
mockdata/main.go
+176 -54
View File
@@ -1,6 +1,7 @@
package main
import (
"encoding/json"
"fmt"
"math/rand"
"time"
@@ -9,88 +10,92 @@ import (
"hr_receiver/models"
)
const (
aiAnalysisRecordCount = 100
trainingSessionRecordCount = 80
insertBatchSize = 50
)
var (
regionIDs = []uint32{1, 3}
sourceTypes = []string{"upload", "cloud"}
appNames = []string{"HeartRate Teacher", "HeartRate Console", "HeartRate Admin"}
)
func main() {
rng := rand.New(rand.NewSource(time.Now().UnixNano()))
config.InitConfig()
config.ConnectDB()
// 生成100条测试数据
count := 100
records := make([]models.AIAnalysisRecord, 0, count)
//aiRecords := generateAIAnalysisRecords(rng, aiAnalysisRecordCount)
//if err := config.DB.CreateInBatches(aiRecords, insertBatchSize).Error; err != nil {
// panic("failed to insert AI analysis mock data: " + err.Error())
//}
for i := 0; i < count; i++ {
records = append(records, generateRecord())
trainingRecords := generateTrainingSessionRecords(rng, trainingSessionRecordCount)
if err := config.DB.CreateInBatches(trainingRecords, insertBatchSize).Error; err != nil {
panic("failed to insert training session mock data: " + err.Error())
}
if err := config.DB.CreateInBatches(records, 50).Error; err != nil {
panic("failed to insert mock data: " + err.Error())
}
fmt.Printf("成功插入 %d 条 AI 分析记录\n", count)
//fmt.Printf("成功插入 %d 条 AI 分析记录\n", len(aiRecords))
fmt.Printf("成功插入 %d 条训练会话记录\n", len(trainingRecords))
}
func generateRecord() models.AIAnalysisRecord {
// regionID 为 1 或 3
regionID := uint32(1)
if rand.Intn(2) == 1 {
regionID = 3
func generateAIAnalysisRecords(rng *rand.Rand, count int) []models.AIAnalysisRecord {
records := make([]models.AIAnalysisRecord, 0, count)
for i := 0; i < count; i++ {
records = append(records, generateAIAnalysisRecord(rng))
}
return records
}
// sourceType: upload 或 cloud
sourceType := "upload"
if rand.Intn(2) == 1 {
sourceType = "cloud"
}
func generateAIAnalysisRecord(rng *rand.Rand) models.AIAnalysisRecord {
regionID := pickRegionID(rng)
sourceType := pickOne(rng, sourceTypes)
// docx 教案原始文件大小: 50KB ~ 500KB
docxSize := int64(rand.Intn(451*1024) + 50*1024)
docxSize := int64(rng.Intn(451*1024) + 50*1024)
csvSize := int64(rng.Intn(20*1024) + 70*1024)
// 心率 csv 原始文件大小: 约 80KB (70KB ~ 90KB)
csvSize := int64(rand.Intn(20*1024) + 70*1024)
// 步数 csv 原始文件大小: 约 20KB ~ 40KB (heart_rate_with_steps 时才有)
var stepCsvSize int64
analysisType := analysisType()
analysisType := randomAnalysisType(rng)
var stepCSVSize int64
if analysisType == "heart_rate_with_steps" {
stepCsvSize = int64(rand.Intn(20*1024) + 20*1024)
stepCSVSize = int64(rng.Intn(20*1024) + 20*1024)
}
originalFileSize := docxSize + csvSize + stepCsvSize
originalFileSize := docxSize + csvSize + stepCSVSize
// 压缩后内容大小: csv 每4行保留1行，大约压缩为 25% + 表头；docx 提取文本后大约 30%~60%
compressedDocx := int64(float64(docxSize) * (0.3 + rand.Float64()*0.3))
compressedCsv := int64(float64(csvSize) * (0.22 + rand.Float64()*0.08)) // ~22%-30%
var compressedStepCsv int64
if stepCsvSize > 0 {
compressedStepCsv = int64(float64(stepCsvSize) * (0.22 + rand.Float64()*0.08))
compressedDocx := int64(float64(docxSize) * (0.3 + rng.Float64()*0.3))
compressedCSV := int64(float64(csvSize) * (0.22 + rng.Float64()*0.08))
var compressedStepCSV int64
if stepCSVSize > 0 {
compressedStepCSV = int64(float64(stepCSVSize) * (0.22 + rng.Float64()*0.08))
}
compressedContentSize := compressedDocx + compressedCsv + compressedStepCsv
compressedContentSize := compressedDocx + compressedCSV + compressedStepCSV
// prompt 大小 = 压缩后内容 + 提示词模板 (~1.5KB)
promptTemplateSize := 1500 + rand.Intn(500)
promptTemplateSize := 1500 + rng.Intn(500)
inputSizeBytes := int(compressedContentSize) + promptTemplateSize
outputSizeBytes := rng.Intn(22*1024) + 3*1024
// AI 输出大小: 3KB ~ 25KB (分析报告)
outputSizeBytes := rand.Intn(22*1024) + 3*1024
inputTokens := inputSizeBytes / (3 + rng.Intn(2))
outputTokens := outputSizeBytes / (3 + rng.Intn(2))
// token 估算: 中文混合场景，平均约 3.5 字节/token
inputTokens := inputSizeBytes / (3 + rand.Intn(2))
outputTokens := outputSizeBytes / (3 + rand.Intn(2))
// 分析时长: 主要和输出 token 数量相关，1分钟以内
// 基础延迟 500ms + 每token约 15~40ms
tokenLatency := int64(15 + rand.Intn(26))
tokenLatency := int64(15 + rng.Intn(26))
durationMs := 500 + int64(outputTokens)*tokenLatency
if durationMs > 60000 {
durationMs = 60000 - int64(rand.Intn(5000))
durationMs = 60000 - int64(rng.Intn(5000))
}
// 上传时间: 最近 90 天内随机
uploadTime := time.Now().Add(-time.Duration(rand.Intn(90*24)) * time.Hour).Add(-time.Duration(rand.Intn(60)) * time.Minute).UnixMilli()
uploadTime := randomRecentMillis(rng, 90)
totalCost, costJSON := buildMockCost(inputTokens, outputTokens)
return models.AIAnalysisRecord{
RegionID: &regionID,
SourceType: sourceType,
AnalysisType: analysisType,
AnalysisResult: "mock analysis result",
CostJSON: costJSON,
TotalCost: totalCost,
InputTokens: inputTokens,
OutputTokens: outputTokens,
InputSizeBytes: inputSizeBytes,
@@ -102,10 +107,127 @@ func generateRecord() models.AIAnalysisRecord {
}
}
func analysisType() string {
// 约 30% 的带步数分析
if rand.Intn(100) < 30 {
func generateTrainingSessionRecords(rng *rand.Rand, count int) []models.MqttTrainingSessionRecord {
records := make([]models.MqttTrainingSessionRecord, 0, count)
sessionCount := count / 2
if sessionCount == 0 {
sessionCount = 1
}
for i := 0; i < sessionCount; i++ {
sessionRecords := generateTrainingSessionPair(rng, i)
records = append(records, sessionRecords...)
if len(records) >= count {
break
}
}
if len(records) > count {
records = records[:count]
}
return records
}
func generateTrainingSessionPair(rng *rand.Rand, index int) []models.MqttTrainingSessionRecord {
regionID := pickRegionID(rng)
appName := pickOne(rng, appNames)
testID := fmt.Sprintf("mock-test-%04d", index+1)
startedAt := randomRecentMillis(rng, 60)
publishedAt := startedAt + int64(rng.Intn(30_000))
startRecord := models.MqttTrainingSessionRecord{
Identifier: fmt.Sprintf("mock-heartrate-%d-%s", regionID, testID),
Topic: fmt.Sprintf("/whgw/v2/region/test/%d", regionID),
TestID: testID,
EventType: "start_test",
RegionID: regionID,
FlavorType: "heartrate",
RawFlavor: "heartrate",
AppName: appName,
StartedAt: int64Ptr(startedAt),
PublishedAt: publishedAt,
ReceivedAt: publishedAt + int64(rng.Intn(3000)),
RawPayload: buildTrainingPayloadJSON(testID, regionID, appName, "start_test", publishedAt),
}
records := []models.MqttTrainingSessionRecord{startRecord}
// 保留少量“开始未结束”的样本，用于测试进行中统计。
if rng.Intn(100) < 15 {
return records
}
durationMs := int64((5+rng.Intn(41))*60*1000 + rng.Intn(60_000))
endedAt := startedAt + durationMs
stopPublishedAt := endedAt + int64(rng.Intn(20_000))
stopRecord := models.MqttTrainingSessionRecord{
Identifier: fmt.Sprintf("mock-heartrate-%d-%s-stop", regionID, testID),
Topic: fmt.Sprintf("/whgw/v2/region/test/%d", regionID),
TestID: testID,
EventType: "stop_test",
RegionID: regionID,
FlavorType: "heartrate",
RawFlavor: "heartrate",
AppName: appName,
StartedAt: int64Ptr(startedAt),
EndedAt: int64Ptr(endedAt),
PublishedAt: stopPublishedAt,
ReceivedAt: stopPublishedAt + int64(rng.Intn(3000)),
RawPayload: buildTrainingPayloadJSON(testID, regionID, appName, "stop_test", stopPublishedAt),
}
return append(records, stopRecord)
}
func buildMockCost(inputTokens, outputTokens int) (float64, string) {
inputCost := float64(inputTokens) * 0.000002
outputCost := float64(outputTokens) * 0.000008
totalCost := inputCost + outputCost
payload := map[string]float64{
"inputCost": inputCost,
"outputCost": outputCost,
}
data, _ := json.Marshal(payload)
return totalCost, string(data)
}
func buildTrainingPayloadJSON(testID string, regionID uint32, appName, eventType string, publishedAt int64) string {
payload := map[string]string{
"appName": appName,
"eventType": eventType,
"flavor": "heartrate",
"regionId": fmt.Sprintf("%d", regionID),
"testId": testID,
"timestamp": time.UnixMilli(publishedAt).UTC().Format("2006-01-02T15:04:05.000Z"),
"type": "mqtt_test",
}
data, _ := json.Marshal(payload)
return string(data)
}
func randomAnalysisType(rng *rand.Rand) string {
if rng.Intn(100) < 30 {
return "heart_rate_with_steps"
}
return "heart_rate_only"
}
func randomRecentMillis(rng *rand.Rand, days int) int64 {
return time.Now().
Add(-time.Duration(rng.Intn(days*24)) * time.Hour).
Add(-time.Duration(rng.Intn(60)) * time.Minute).
UnixMilli()
}
func pickRegionID(rng *rand.Rand) uint32 {
return regionIDs[rng.Intn(len(regionIDs))]
}
func pickOne(rng *rand.Rand, values []string) string {
return values[rng.Intn(len(values))]
}
func int64Ptr(v int64) *int64 {
return &v
}