feat(server): hot-reload config store and reloadable infra clients
- Replace single Load() with a watching Store that re-reads config files
(and config.local.yaml) and fans out a ReloadEvent with a per-field diff
so consumers can decide whether the change is hot-applicable or requires
a process restart.
- Wrap llm, retrieval, vector store, and object storage clients in
Reloadable* shells so the bootstrap can swap their underlying impls when
config changes without re-instantiating handlers.
- Make jwt.Manager and ops TokenIssuer mutable under a lock so secrets and
TTLs can be rotated live; thread default plan code through a setter on
the ops AdminUserService.
- Wire ConfigStore through bootstrap and every cmd/main.go, scheduler /
worker / tenant-api / ops-api start the watcher; services and handlers
take a config.Provider so they always read current values for things
like generation.stream_enabled, scheduler dispatch, retrieval, etc.
- Switch shared/config decoding off viper to a Kratos-derived runtime
package so env placeholders (\${VAR:default}) resolve consistently and
the same source machinery powers both the loader and the watcher.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
@@ -23,13 +23,14 @@ import (
|
||||
const kolGenerationTaskType = "kol"
|
||||
|
||||
type KolGenerationWorker struct {
|
||||
pool *pgxpool.Pool
|
||||
llm llm.Client
|
||||
knowledge *tenantapp.KnowledgeService
|
||||
logger *zap.Logger
|
||||
articleTimeout time.Duration
|
||||
maxOutputTokens int64
|
||||
cache sharedcache.Cache
|
||||
pool *pgxpool.Pool
|
||||
llm llm.Client
|
||||
knowledge *tenantapp.KnowledgeService
|
||||
logger *zap.Logger
|
||||
cfg config.GenerationConfig
|
||||
llmCfg config.LLMConfig
|
||||
provider config.Provider
|
||||
cache sharedcache.Cache
|
||||
}
|
||||
|
||||
func NewKolGenerationWorker(
|
||||
@@ -40,18 +41,13 @@ func NewKolGenerationWorker(
|
||||
cfg config.GenerationConfig,
|
||||
llmMaxOutputTokens int64,
|
||||
) *KolGenerationWorker {
|
||||
articleTimeout := cfg.ArticleTimeout
|
||||
if articleTimeout <= 0 {
|
||||
articleTimeout = 8 * time.Minute
|
||||
}
|
||||
|
||||
return &KolGenerationWorker{
|
||||
pool: pool,
|
||||
llm: llmClient,
|
||||
knowledge: knowledgeSvc,
|
||||
logger: logger,
|
||||
articleTimeout: articleTimeout,
|
||||
maxOutputTokens: llmMaxOutputTokens,
|
||||
pool: pool,
|
||||
llm: llmClient,
|
||||
knowledge: knowledgeSvc,
|
||||
logger: logger,
|
||||
cfg: cfg,
|
||||
llmCfg: config.LLMConfig{MaxOutputTokens: llmMaxOutputTokens},
|
||||
}
|
||||
}
|
||||
|
||||
@@ -60,6 +56,23 @@ func (w *KolGenerationWorker) WithCache(c sharedcache.Cache) *KolGenerationWorke
|
||||
return w
|
||||
}
|
||||
|
||||
func (w *KolGenerationWorker) WithConfigProvider(provider config.Provider) *KolGenerationWorker {
|
||||
w.provider = provider
|
||||
return w
|
||||
}
|
||||
|
||||
func (w *KolGenerationWorker) runtimeConfig() (time.Duration, int64) {
|
||||
if w != nil && w.provider != nil {
|
||||
if cfg := w.provider.Current(); cfg != nil {
|
||||
return kolArticleTimeout(cfg.Generation), cfg.LLM.MaxOutputTokens
|
||||
}
|
||||
}
|
||||
if w == nil {
|
||||
return 8 * time.Minute, 0
|
||||
}
|
||||
return kolArticleTimeout(w.cfg), w.llmCfg.MaxOutputTokens
|
||||
}
|
||||
|
||||
func (w *KolGenerationWorker) Process(ctx context.Context, task tenantapp.ClaimedGenerationTask) {
|
||||
renderedPrompt := strings.TrimSpace(kolStringInput(task.InputParams, "rendered_prompt"))
|
||||
if renderedPrompt == "" {
|
||||
@@ -114,10 +127,11 @@ func (w *KolGenerationWorker) Process(ctx context.Context, task tenantapp.Claime
|
||||
prompt = strings.TrimSpace(renderedPrompt + "\n\n" + knowledgePrompt)
|
||||
}
|
||||
|
||||
articleTimeout, maxOutputTokens := w.runtimeConfig()
|
||||
generateReq := llm.GenerateRequest{
|
||||
Prompt: prompt,
|
||||
Timeout: w.articleTimeout,
|
||||
MaxOutputTokens: w.maxOutputTokens,
|
||||
Timeout: articleTimeout,
|
||||
MaxOutputTokens: maxOutputTokens,
|
||||
}
|
||||
if kolBoolInput(task.InputParams, "enable_web_search") {
|
||||
generateReq.WebSearch = &llm.WebSearchOptions{Enabled: true}
|
||||
@@ -236,6 +250,13 @@ var (
|
||||
errEmptyGeneratedContent = errors.New("kol generation returned empty content")
|
||||
)
|
||||
|
||||
func kolArticleTimeout(cfg config.GenerationConfig) time.Duration {
|
||||
if cfg.ArticleTimeout > 0 {
|
||||
return cfg.ArticleTimeout
|
||||
}
|
||||
return 8 * time.Minute
|
||||
}
|
||||
|
||||
func kolStringInput(input map[string]interface{}, key string) string {
|
||||
if input == nil {
|
||||
return ""
|
||||
|
||||
Reference in New Issue
Block a user