Version: 0.9.65.dev.260503

后端:
1. 阶段 1.5/1.6
收口 llm-service / rag-service,统一模型出口与检索基础设施入口,清退 backend/infra/llm 与 backend/infra/rag 旧实现;
2. 同步更新相关调用链与微服务迁移计划文档
This commit is contained in:
Losita
2026-05-03 23:21:03 +08:00
parent a6c1e5d077
commit 9902ca3563
65 changed files with 550 additions and 376 deletions

View File

@@ -6,9 +6,9 @@ import (
"log"
"strings"
infrallm "github.com/LoveLosita/smartflow/backend/infra/llm"
newagentprompt "github.com/LoveLosita/smartflow/backend/newAgent/prompt"
newagentstream "github.com/LoveLosita/smartflow/backend/newAgent/stream"
llmservice "github.com/LoveLosita/smartflow/backend/services/llm"
)
const reasoningSummaryMaxTokens = 700
@@ -24,7 +24,7 @@ type reasoningSummaryLLMResponse struct {
// 1. service 层负责选择模型与 promptstream 层只负责调度和闸门;
// 2. 这里不持久化摘要,持久化统一走 ChunkEmitter 的 extra hook
// 3. 摘要失败时返回 error由 ReasoningDigestor 吞掉并等待下一次水位线/Flush 兜底。
func (s *AgentService) makeReasoningSummaryFunc(client *infrallm.Client) newagentstream.ReasoningSummaryFunc {
func (s *AgentService) makeReasoningSummaryFunc(client *llmservice.Client) newagentstream.ReasoningSummaryFunc {
if client == nil {
return nil
}
@@ -47,14 +47,14 @@ func (s *AgentService) makeReasoningSummaryFunc(client *infrallm.Client) newagen
DurationSeconds: input.DurationSeconds,
})
resp, rawResult, err := infrallm.GenerateJSON[reasoningSummaryLLMResponse](
resp, rawResult, err := llmservice.GenerateJSON[reasoningSummaryLLMResponse](
ctx,
client,
messages,
infrallm.GenerateOptions{
llmservice.GenerateOptions{
Temperature: 0.1,
MaxTokens: reasoningSummaryMaxTokens,
Thinking: infrallm.ThinkingModeDisabled,
Thinking: llmservice.ThinkingModeDisabled,
Metadata: map[string]any{
"stage": "reasoning_summary",
"candidate_seq": input.CandidateSeq,
@@ -99,7 +99,7 @@ func limitReasoningDetailSummary(text string, maxRunes int) string {
return string(runes[:maxRunes])
}
func truncateReasoningSummaryRaw(raw *infrallm.TextResult) string {
func truncateReasoningSummaryRaw(raw *llmservice.TextResult) string {
if raw == nil {
return ""
}