Version: 0.9.80.dev.260506
后端： 1. LLM 独立服务与统一计费出口落地：新增 `cmd/llm`、`client/llm` 与 `services/llm/rpc`，补齐 BillingContext、CreditBalanceGuard、价格规则解析、stream usage 归集与 `credit.charge.requested` outbox 发布，active-scheduler / agent / course / memory / gateway fallback 全部改走 llm zrpc，不再各自本地初始化模型。 2. TokenStore 收口为 Credit 权威账本：新增 credit account / ledger / product / order / price-rule / reward-rule 能力与 Redis 快照缓存，扩展 tokenstore rpc/client 支撑余额快照、消耗看板、商品、订单、流水、价格规则和奖励规则，并接入 LLM charge 事件消费完成 Credit 扣费落账。 3. 计费旧链路下线与网关切口切换：`/token-store` 语义整体切到 `/credit-store`，agent chat 移除旧 TokenQuotaGuard，userauth 的 CheckTokenQuota / AdjustTokenUsage 改为废弃，聊天历史落库不再同步旧 token 额度账本，course 图片解析请求补 user_id 进入新计费口径。 前端： 4. 计划广场从 mock 数据切到真实接口：新增 forum api/types，首页支持真实列表、标签、搜索、防抖、点赞、导入和发布计划，详情页补齐帖子详情、评论树、回复和删除评论链路，同时补上“至少一个标签”的前后端约束与默认标签兜底。 5. 商店页切到 Credit 体系并重做展示：顶部改为余额 + Credit/Token 消耗看板，支持 24h/7d/30d/all 周期切换；套餐区展示原价与当前价；历史区改为当前用户 Credit 流水并支持查看更多，整体视觉和交互同步收口。 仓库： 6. 配置与本地启动体系补齐 llm / outbox 编排：`config.example.yaml` 增加 llm rpc 和统一 outbox service 配置，`dev-common.ps1` 把 llm 纳入多服务依赖并自动建 Kafka topic，`docker-compose.yml` 同步初始化 agent/task/memory/active-scheduler/notification/taskclass-forum/llm/token-store 全量 outbox topic。
This commit is contained in:
Losita
301
backend/client/llm/client.go
Normal file
301
backend/client/llm/client.go
Normal file
@@ -0,0 +1,301 @@
|
||||
package llm
|
||||
|
||||
import (
|
||||
"context"
|
||||
"errors"
|
||||
"io"
|
||||
"strings"
|
||||
"time"
|
||||
|
||||
llmservice "github.com/LoveLosita/smartflow/backend/services/llm"
|
||||
llmrpc "github.com/LoveLosita/smartflow/backend/services/llm/rpc"
|
||||
llmcontracts "github.com/LoveLosita/smartflow/backend/shared/contracts/llm"
|
||||
"github.com/cloudwego/eino/schema"
|
||||
"github.com/zeromicro/go-zero/zrpc"
|
||||
)
|
||||
|
||||
const (
|
||||
defaultEndpoint = "127.0.0.1:9096"
|
||||
defaultTimeout = 0
|
||||
defaultPingTimeout = 2 * time.Second
|
||||
)
|
||||
|
||||
type ClientConfig struct {
|
||||
Endpoints []string
|
||||
Target string
|
||||
Timeout time.Duration
|
||||
}
|
||||
|
||||
type ServiceConfig struct {
|
||||
ClientConfig
|
||||
CourseVisionModel string
|
||||
}
|
||||
|
||||
// Client 是业务进程访问独立 LLM 服务的最小 RPC 适配层。
|
||||
type Client struct {
|
||||
rpc llmrpc.LLMClient
|
||||
}
|
||||
|
||||
func NewClient(cfg ClientConfig) (*Client, error) {
|
||||
timeout := cfg.Timeout
|
||||
if timeout < 0 {
|
||||
timeout = defaultTimeout
|
||||
}
|
||||
endpoints := normalizeEndpoints(cfg.Endpoints)
|
||||
target := strings.TrimSpace(cfg.Target)
|
||||
if len(endpoints) == 0 && target == "" {
|
||||
endpoints = []string{defaultEndpoint}
|
||||
}
|
||||
|
||||
zclient, err := zrpc.NewClient(zrpc.RpcClientConf{
|
||||
Endpoints: endpoints,
|
||||
Target: target,
|
||||
NonBlock: true,
|
||||
Timeout: int64(timeout / time.Millisecond),
|
||||
}, zrpc.WithDialOption(llmrpc.JSONCodecDialOption()))
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
client := &Client{rpc: llmrpc.NewLLMClient(zclient.Conn())}
|
||||
if err = client.ping(resolvePingTimeout(timeout)); err != nil {
|
||||
return nil, err
|
||||
}
|
||||
return client, nil
|
||||
}
|
||||
|
||||
// NewService 一次性把远端 LLM RPC 包装回旧的 *llmservice.Service 门面。
|
||||
func NewService(cfg ServiceConfig) (*llmservice.Service, error) {
|
||||
client, err := NewClient(cfg.ClientConfig)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
return client.BuildService(cfg.CourseVisionModel), nil
|
||||
}
|
||||
|
||||
func (c *Client) BuildService(courseVisionModel string) *llmservice.Service {
|
||||
if c == nil {
|
||||
return nil
|
||||
}
|
||||
|
||||
return llmservice.NewWithClients(llmservice.StaticClients{
|
||||
Lite: buildTextClient(c, llmcontracts.ModelAliasLite),
|
||||
Pro: buildTextClient(c, llmcontracts.ModelAliasPro),
|
||||
Max: buildTextClient(c, llmcontracts.ModelAliasMax),
|
||||
CourseImageResponses: llmservice.NewArkResponsesClientWithFunc(courseVisionModel, func(ctx context.Context, messages []llmservice.ArkResponsesMessage, options llmservice.ArkResponsesOptions) (*llmservice.ArkResponsesResult, error) {
|
||||
return c.GenerateResponsesText(ctx, llmcontracts.ModelAliasCourseImageResponses, messages, options)
|
||||
}),
|
||||
})
|
||||
}
|
||||
|
||||
func (c *Client) Ping(ctx context.Context) error {
|
||||
if err := c.ensureReady(); err != nil {
|
||||
return err
|
||||
}
|
||||
_, err := c.rpc.Ping(ctx, &llmcontracts.PingRequest{})
|
||||
return responseFromRPCError(err)
|
||||
}
|
||||
|
||||
func (c *Client) GenerateText(ctx context.Context, modelAlias string, messages []*schema.Message, options llmservice.GenerateOptions) (*llmservice.TextResult, error) {
|
||||
if err := c.ensureReady(); err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
resp, err := c.rpc.GenerateText(ctx, &llmcontracts.TextRequest{
|
||||
ModelAlias: modelAlias,
|
||||
Messages: messages,
|
||||
Options: toContractGenerateOptions(options),
|
||||
Billing: billingFromContext(ctx, modelAlias),
|
||||
})
|
||||
if err != nil {
|
||||
return nil, responseFromRPCError(err)
|
||||
}
|
||||
if resp == nil || resp.Result == nil {
|
||||
return nil, errors.New("llm zrpc service returned empty text response")
|
||||
}
|
||||
return &llmservice.TextResult{
|
||||
Text: resp.Result.Text,
|
||||
Usage: llmservice.CloneUsage(resp.Result.Usage),
|
||||
FinishReason: resp.Result.FinishReason,
|
||||
}, nil
|
||||
}
|
||||
|
||||
func (c *Client) StreamText(ctx context.Context, modelAlias string, messages []*schema.Message, options llmservice.GenerateOptions) (llmservice.StreamReader, error) {
|
||||
if err := c.ensureReady(); err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
stream, err := c.rpc.StreamText(ctx, &llmcontracts.StreamTextRequest{
|
||||
ModelAlias: modelAlias,
|
||||
Messages: messages,
|
||||
Options: toContractGenerateOptions(options),
|
||||
Billing: billingFromContext(ctx, modelAlias),
|
||||
})
|
||||
if err != nil {
|
||||
return nil, responseFromRPCError(err)
|
||||
}
|
||||
return &streamReader{stream: stream}, nil
|
||||
}
|
||||
|
||||
func (c *Client) GenerateResponsesText(ctx context.Context, modelAlias string, messages []llmservice.ArkResponsesMessage, options llmservice.ArkResponsesOptions) (*llmservice.ArkResponsesResult, error) {
|
||||
if err := c.ensureReady(); err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
resp, err := c.rpc.GenerateResponsesText(ctx, &llmcontracts.ResponsesRequest{
|
||||
ModelAlias: modelAlias,
|
||||
Messages: toContractResponsesMessages(messages),
|
||||
Options: toContractResponsesOptions(options),
|
||||
Billing: billingFromContext(ctx, modelAlias),
|
||||
})
|
||||
if err != nil {
|
||||
return nil, responseFromRPCError(err)
|
||||
}
|
||||
if resp == nil || resp.Result == nil {
|
||||
return nil, errors.New("llm zrpc service returned empty responses response")
|
||||
}
|
||||
return toServiceResponsesResult(resp.Result), nil
|
||||
}
|
||||
|
||||
func (c *Client) ensureReady() error {
|
||||
if c == nil || c.rpc == nil {
|
||||
return errors.New("llm zrpc client is not initialized")
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
func (c *Client) ping(timeout time.Duration) error {
|
||||
ctx, cancel := context.WithTimeout(context.Background(), timeout)
|
||||
defer cancel()
|
||||
return c.Ping(ctx)
|
||||
}
|
||||
|
||||
type streamReader struct {
|
||||
stream llmrpc.LLM_StreamTextClient
|
||||
}
|
||||
|
||||
func (r *streamReader) Recv() (*schema.Message, error) {
|
||||
if r == nil || r.stream == nil {
|
||||
return nil, errors.New("llm zrpc stream is not initialized")
|
||||
}
|
||||
|
||||
chunk, err := r.stream.Recv()
|
||||
if err != nil {
|
||||
if errors.Is(err, io.EOF) {
|
||||
return nil, io.EOF
|
||||
}
|
||||
return nil, responseFromRPCError(err)
|
||||
}
|
||||
if chunk == nil {
|
||||
return nil, errors.New("llm zrpc service returned empty stream chunk")
|
||||
}
|
||||
return chunk.Message, nil
|
||||
}
|
||||
|
||||
func (r *streamReader) Close() error {
|
||||
return nil
|
||||
}
|
||||
|
||||
func buildTextClient(remote *Client, modelAlias string) *llmservice.Client {
|
||||
return llmservice.NewClient(
|
||||
func(ctx context.Context, messages []*schema.Message, options llmservice.GenerateOptions) (*llmservice.TextResult, error) {
|
||||
return remote.GenerateText(ctx, modelAlias, messages, options)
|
||||
},
|
||||
func(ctx context.Context, messages []*schema.Message, options llmservice.GenerateOptions) (llmservice.StreamReader, error) {
|
||||
return remote.StreamText(ctx, modelAlias, messages, options)
|
||||
},
|
||||
)
|
||||
}
|
||||
|
||||
func billingFromContext(ctx context.Context, modelAlias string) *llmcontracts.BillingContext {
|
||||
billing, ok := llmservice.BillingContextFromContext(ctx)
|
||||
if !ok {
|
||||
return nil
|
||||
}
|
||||
if strings.TrimSpace(billing.ModelAlias) == "" {
|
||||
billing.ModelAlias = strings.TrimSpace(modelAlias)
|
||||
}
|
||||
return &llmcontracts.BillingContext{
|
||||
UserID: billing.UserID,
|
||||
EventID: billing.EventID,
|
||||
Scene: billing.Scene,
|
||||
RequestID: billing.RequestID,
|
||||
ConversationID: billing.ConversationID,
|
||||
ModelAlias: billing.ModelAlias,
|
||||
SkipCharge: billing.SkipCharge,
|
||||
}
|
||||
}
|
||||
|
||||
func toContractGenerateOptions(input llmservice.GenerateOptions) llmcontracts.GenerateOptions {
|
||||
return llmcontracts.GenerateOptions{
|
||||
Temperature: input.Temperature,
|
||||
MaxTokens: input.MaxTokens,
|
||||
Thinking: string(input.Thinking),
|
||||
Metadata: input.Metadata,
|
||||
}
|
||||
}
|
||||
|
||||
func toContractResponsesMessages(input []llmservice.ArkResponsesMessage) []llmcontracts.ResponsesMessage {
|
||||
if len(input) == 0 {
|
||||
return nil
|
||||
}
|
||||
output := make([]llmcontracts.ResponsesMessage, 0, len(input))
|
||||
for _, item := range input {
|
||||
output = append(output, llmcontracts.ResponsesMessage{
|
||||
Role: item.Role,
|
||||
Text: item.Text,
|
||||
ImageURL: item.ImageURL,
|
||||
ImageDetail: item.ImageDetail,
|
||||
})
|
||||
}
|
||||
return output
|
||||
}
|
||||
|
||||
func toContractResponsesOptions(input llmservice.ArkResponsesOptions) llmcontracts.ResponsesOptions {
|
||||
return llmcontracts.ResponsesOptions{
|
||||
Model: input.Model,
|
||||
Temperature: input.Temperature,
|
||||
MaxOutputTokens: input.MaxOutputTokens,
|
||||
Thinking: string(input.Thinking),
|
||||
TextFormat: input.TextFormat,
|
||||
}
|
||||
}
|
||||
|
||||
func toServiceResponsesResult(result *llmcontracts.ResponsesResult) *llmservice.ArkResponsesResult {
|
||||
if result == nil {
|
||||
return nil
|
||||
}
|
||||
output := &llmservice.ArkResponsesResult{
|
||||
Text: result.Text,
|
||||
Status: result.Status,
|
||||
IncompleteReason: result.IncompleteReason,
|
||||
ErrorCode: result.ErrorCode,
|
||||
ErrorMessage: result.ErrorMessage,
|
||||
}
|
||||
if result.Usage != nil {
|
||||
output.Usage = &llmservice.ArkResponsesUsage{
|
||||
InputTokens: result.Usage.InputTokens,
|
||||
OutputTokens: result.Usage.OutputTokens,
|
||||
TotalTokens: result.Usage.TotalTokens,
|
||||
}
|
||||
}
|
||||
return output
|
||||
}
|
||||
|
||||
func normalizeEndpoints(values []string) []string {
|
||||
endpoints := make([]string, 0, len(values))
|
||||
for _, value := range values {
|
||||
trimmed := strings.TrimSpace(value)
|
||||
if trimmed != "" {
|
||||
endpoints = append(endpoints, trimmed)
|
||||
}
|
||||
}
|
||||
return endpoints
|
||||
}
|
||||
|
||||
func resolvePingTimeout(timeout time.Duration) time.Duration {
|
||||
if timeout > 0 && timeout < defaultPingTimeout {
|
||||
return timeout
|
||||
}
|
||||
return defaultPingTimeout
|
||||
}
|
||||
73
backend/client/llm/errors.go
Normal file
73
backend/client/llm/errors.go
Normal file
@@ -0,0 +1,73 @@
|
||||
package llm
|
||||
|
||||
import (
|
||||
"errors"
|
||||
"fmt"
|
||||
"strings"
|
||||
|
||||
"google.golang.org/genproto/googleapis/rpc/errdetails"
|
||||
"google.golang.org/grpc/codes"
|
||||
"google.golang.org/grpc/status"
|
||||
)
|
||||
|
||||
func responseFromRPCError(err error) error {
|
||||
if err == nil {
|
||||
return nil
|
||||
}
|
||||
|
||||
st, ok := status.FromError(err)
|
||||
if !ok {
|
||||
return wrapRPCError(err)
|
||||
}
|
||||
if message, ok := messageFromStatus(st); ok {
|
||||
switch st.Code() {
|
||||
case codes.InvalidArgument, codes.ResourceExhausted, codes.FailedPrecondition:
|
||||
return errors.New(message)
|
||||
}
|
||||
}
|
||||
|
||||
switch st.Code() {
|
||||
case codes.InvalidArgument, codes.ResourceExhausted, codes.FailedPrecondition:
|
||||
return errors.New(strings.TrimSpace(st.Message()))
|
||||
case codes.Internal, codes.Unknown, codes.Unavailable, codes.DeadlineExceeded, codes.DataLoss, codes.Unimplemented:
|
||||
msg := strings.TrimSpace(st.Message())
|
||||
if msg == "" {
|
||||
msg = "llm zrpc service internal error"
|
||||
}
|
||||
return wrapRPCError(errors.New(msg))
|
||||
default:
|
||||
msg := strings.TrimSpace(st.Message())
|
||||
if msg == "" {
|
||||
msg = "llm zrpc service rejected request"
|
||||
}
|
||||
return errors.New(msg)
|
||||
}
|
||||
}
|
||||
|
||||
func messageFromStatus(st *status.Status) (string, bool) {
|
||||
if st == nil {
|
||||
return "", false
|
||||
}
|
||||
for _, detail := range st.Details() {
|
||||
info, ok := detail.(*errdetails.ErrorInfo)
|
||||
if !ok {
|
||||
continue
|
||||
}
|
||||
message := strings.TrimSpace(st.Message())
|
||||
if message == "" && info.Metadata != nil {
|
||||
message = strings.TrimSpace(info.Metadata["info"])
|
||||
}
|
||||
if message == "" {
|
||||
message = strings.TrimSpace(info.Reason)
|
||||
}
|
||||
return message, message != ""
|
||||
}
|
||||
return "", false
|
||||
}
|
||||
|
||||
func wrapRPCError(err error) error {
|
||||
if err == nil {
|
||||
return nil
|
||||
}
|
||||
return fmt.Errorf("调用 llm zrpc 服务失败: %w", err)
|
||||
}
|
||||
Reference in New Issue
Block a user