91 lines
2.7 KiB
Go
91 lines
2.7 KiB
Go
package main
|
||
|
||
import (
|
||
"fmt"
|
||
"os"
|
||
)
|
||
|
||
func main() {
|
||
if err := runOnce(); err != nil {
|
||
logError(err.Error())
|
||
os.Exit(1)
|
||
}
|
||
}
|
||
|
||
func runOnce() error {
|
||
logInfo("Go RPA agent demo 启动")
|
||
|
||
config, err := loadConfig("config.toml")
|
||
if err != nil {
|
||
return fmt.Errorf("读取 config.toml 失败：%w", err)
|
||
}
|
||
logInfo("已读取 config.toml，模型=" + config.Volcengine.Model)
|
||
|
||
regionsPath := appDataPath(config, config.Agent.RegionsPath)
|
||
screenshotPath := appDataPath(config, config.Agent.ScreenshotPath)
|
||
annotation, err := loadAnnotation(regionsPath)
|
||
if err != nil {
|
||
return fmt.Errorf("读取标注文件失败：%w", err)
|
||
}
|
||
summaries := summarizeRegions(annotation.Regions)
|
||
summaries, inferenceWarnings := inferMissingRegionTypes(summaries)
|
||
logInfo(fmt.Sprintf("已读取标注文件，regions=%d", len(summaries)))
|
||
for _, warning := range inferenceWarnings {
|
||
logWarning(warning)
|
||
}
|
||
if err := validateRequiredRegions(summaries); err != nil {
|
||
return err
|
||
}
|
||
|
||
for _, summary := range summaries {
|
||
logThink(fmt.Sprintf("区域识别：%s center=[%.1f, %.1f]", summary.Type, summary.CenterScreen[0], summary.CenterScreen[1]))
|
||
}
|
||
|
||
imageDataURL, err := loadImageDataURL(screenshotPath)
|
||
if err != nil {
|
||
return fmt.Errorf("读取截图失败：%w", err)
|
||
}
|
||
logInfo("已读取截图 WeChat.jpg")
|
||
|
||
prompt, err := buildPrompt(summaries)
|
||
if err != nil {
|
||
return fmt.Errorf("构造 prompt 失败：%w", err)
|
||
}
|
||
|
||
logThink("正在请求豆包多模态模型识别聊天内容并生成任务计划")
|
||
rawResponse, err := requestLLM(config, prompt, imageDataURL)
|
||
if err != nil {
|
||
return fmt.Errorf("请求 LLM 失败：%w", err)
|
||
}
|
||
|
||
task, _, err := parseTaskFromLLM(rawResponse)
|
||
if err != nil {
|
||
_ = saveRawResponse(config, rawResponse)
|
||
return fmt.Errorf("解析 LLM task JSON 失败，已保存 latest_task_raw.txt：%w", err)
|
||
}
|
||
|
||
task = finalizeTask(task, summaries, config, regionsPath, screenshotPath)
|
||
logJSON("summary", fmt.Sprintf(
|
||
"LLM 摘要：联系人=%s；最新消息=%s；回复=%s；confidence=%.2f",
|
||
task.LLMSummary.ContactName,
|
||
task.LLMSummary.LatestUserMessage,
|
||
task.LLMSummary.ReplyText,
|
||
task.Confidence,
|
||
))
|
||
logTask(fmt.Sprintf("已生成任务 %s，type=%s，actions=%d，dry_run=%v", task.TaskID, task.TaskType, len(task.Actions), config.Agent.DryRun))
|
||
|
||
for _, action := range task.Actions {
|
||
logTask(fmt.Sprintf("动作记录：%s %s status=%s target=%s", action.ActionID, action.Name, action.Status, action.TargetRegion))
|
||
}
|
||
|
||
latestPath, historyPath, err := saveTaskFiles(task, config)
|
||
if err != nil {
|
||
return fmt.Errorf("保存 task 文件失败：%w", err)
|
||
}
|
||
logInfo("已保存 latest task: " + latestPath)
|
||
logInfo("已保存 history task: " + historyPath)
|
||
logInfo("Go RPA agent demo 完成，未真实执行鼠标键盘")
|
||
|
||
return nil
|
||
}
|