feat(llm): 每个 LLM 配置支持流式/非流式切换(默认流式)
依赖 norma v0.3.0 的 Provider.Complete。为每个 llm_profile 增加 streaming 开关:开(默认)走流式 SSE;关走真·非流式(stream:false、一次性返回完整 JSON),可绕开部分网关糟糕的 SSE 实现(空帧、思考字段丢帧),代价是失去 运行中的实时进度与实时 token 计数。 - agent.Config.Stream → agentcore.Options.NonStreaming = !Stream;worker / planner / mainagent / chat / goals 经 SetNonStreaming resolver 注入,任务 路由按当前激活 profile 动态取值 - llmpool.Pool / llmrec.Recorder / taskLLMRuntime 三个 Provider 包装均实现 Complete(分别带故障转移 / 录制 / 同 provider 重试与额度 failover),使 输出、token 统计、操作统计、LLM 录制四条链路在非流式下保持兼容 - db: llm_profiles 新增 streaming 列 + ALTER 补旧库(默认 true,旧配置无感) - 系统 → LLM 配置表单增加「流式输出」开关
This commit is contained in:
Autumn-27 committed
1 parent
19a7206d12
commit
587ba422b4
22 files changed
+409
-75
No files matched your search
+15
-10
@@ -39,18 +39,23 @@ type LLMProfile struct {
|
||||
// usable when an agent/task binds it explicitly, it just never gets picked up
|
||||
// as a fallback target.
|
||||
PoolExclude bool `json:"pool_exclude"`
|
||||
// Streaming selects the wire protocol: true (default) = streaming (SSE);
|
||||
// false = real non-streaming (stream:false, single JSON response via
|
||||
// Provider.Complete). Non-streaming sidesteps flaky gateway SSE at the cost of
|
||||
// live in-run progress. Maps to agent.Config.Stream.
|
||||
Streaming bool `json:"streaming"`
|
||||
}
|
||||
|
||||
// profileCols is the read column list (hint variant, no api key) shared by the
|
||||
// list query; profileColsKey is the same with api_key for the single-row loads.
|
||||
const profileCols = `id,name,format,COALESCE(base_url,''),COALESCE(proxy,''),model,COALESCE(api_key_hint,''),rate_per_second,rate_per_minute,context_window_k,COALESCE(reasoning_effort,''),is_default,priority,pool_exclude,COALESCE(thinking_type,'')`
|
||||
const profileColsKey = `id,name,format,COALESCE(base_url,''),COALESCE(proxy,''),model,COALESCE(api_key,''),rate_per_second,rate_per_minute,context_window_k,COALESCE(reasoning_effort,''),is_default,priority,pool_exclude,COALESCE(thinking_type,'')`
|
||||
const profileCols = `id,name,format,COALESCE(base_url,''),COALESCE(proxy,''),model,COALESCE(api_key_hint,''),rate_per_second,rate_per_minute,context_window_k,COALESCE(reasoning_effort,''),is_default,priority,pool_exclude,COALESCE(thinking_type,''),COALESCE(streaming,true)`
|
||||
const profileColsKey = `id,name,format,COALESCE(base_url,''),COALESCE(proxy,''),model,COALESCE(api_key,''),rate_per_second,rate_per_minute,context_window_k,COALESCE(reasoning_effort,''),is_default,priority,pool_exclude,COALESCE(thinking_type,''),COALESCE(streaming,true)`
|
||||
|
||||
// scanProfile reads one row in the profileCols / profileColsKey column order. The
|
||||
// 7th column lands in APIKeyHint or APIKey depending on which list the caller used.
|
||||
func scanProfile(sc interface{ Scan(...any) error }, into *string, p *LLMProfile) error {
|
||||
return sc.Scan(&p.ID, &p.Name, &p.Format, &p.BaseURL, &p.Proxy, &p.Model, into,
|
||||
&p.RatePerSecond, &p.RatePerMinute, &p.ContextWindowK, &p.ReasoningEffort, &p.IsDefault, &p.Priority, &p.PoolExclude, &p.ThinkingType)
|
||||
&p.RatePerSecond, &p.RatePerMinute, &p.ContextWindowK, &p.ReasoningEffort, &p.IsDefault, &p.Priority, &p.PoolExclude, &p.ThinkingType, &p.Streaming)
|
||||
}
|
||||
|
||||
func (d *DB) ListProfiles() ([]*LLMProfile, error) {
|
||||
@@ -123,18 +128,18 @@ func (d *DB) SaveProfile(p *LLMProfile) (int64, error) {
|
||||
}
|
||||
if p.ID == 0 {
|
||||
var id int64
|
||||
err := d.QueryRow(`INSERT INTO llm_profiles(name,format,base_url,proxy,model,api_key,api_key_hint,rate_per_second,rate_per_minute,context_window_k,reasoning_effort,priority,pool_exclude,thinking_type)
|
||||
VALUES ($1,$2,NULLIF($3,''),NULLIF($4,''),$5,NULLIF($6,''),NULLIF($7,''),$8,$9,$10,$11,$12,$13,$14) RETURNING id`,
|
||||
p.Name, p.Format, p.BaseURL, p.Proxy, p.Model, p.APIKey, hint, p.RatePerSecond, p.RatePerMinute, p.ContextWindowK, p.ReasoningEffort, p.Priority, p.PoolExclude, p.ThinkingType).Scan(&id)
|
||||
err := d.QueryRow(`INSERT INTO llm_profiles(name,format,base_url,proxy,model,api_key,api_key_hint,rate_per_second,rate_per_minute,context_window_k,reasoning_effort,priority,pool_exclude,thinking_type,streaming)
|
||||
VALUES ($1,$2,NULLIF($3,''),NULLIF($4,''),$5,NULLIF($6,''),NULLIF($7,''),$8,$9,$10,$11,$12,$13,$14,$15) RETURNING id`,
|
||||
p.Name, p.Format, p.BaseURL, p.Proxy, p.Model, p.APIKey, hint, p.RatePerSecond, p.RatePerMinute, p.ContextWindowK, p.ReasoningEffort, p.Priority, p.PoolExclude, p.ThinkingType, p.Streaming).Scan(&id)
|
||||
return id, err
|
||||
}
|
||||
if p.APIKey == "" {
|
||||
_, err := d.Exec(`UPDATE llm_profiles SET name=$1,format=$2,base_url=NULLIF($3,''),proxy=NULLIF($4,''),model=$5,rate_per_second=$6,rate_per_minute=$7,context_window_k=$8,reasoning_effort=$9,priority=$10,pool_exclude=$11,thinking_type=$12 WHERE id=$13`,
|
||||
p.Name, p.Format, p.BaseURL, p.Proxy, p.Model, p.RatePerSecond, p.RatePerMinute, p.ContextWindowK, p.ReasoningEffort, p.Priority, p.PoolExclude, p.ThinkingType, p.ID)
|
||||
_, err := d.Exec(`UPDATE llm_profiles SET name=$1,format=$2,base_url=NULLIF($3,''),proxy=NULLIF($4,''),model=$5,rate_per_second=$6,rate_per_minute=$7,context_window_k=$8,reasoning_effort=$9,priority=$10,pool_exclude=$11,thinking_type=$12,streaming=$13 WHERE id=$14`,
|
||||
p.Name, p.Format, p.BaseURL, p.Proxy, p.Model, p.RatePerSecond, p.RatePerMinute, p.ContextWindowK, p.ReasoningEffort, p.Priority, p.PoolExclude, p.ThinkingType, p.Streaming, p.ID)
|
||||
return p.ID, err
|
||||
}
|
||||
_, err := d.Exec(`UPDATE llm_profiles SET name=$1,format=$2,base_url=NULLIF($3,''),proxy=NULLIF($4,''),model=$5,api_key=$6,api_key_hint=$7,rate_per_second=$8,rate_per_minute=$9,context_window_k=$10,reasoning_effort=$11,priority=$12,pool_exclude=$13,thinking_type=$14 WHERE id=$15`,
|
||||
p.Name, p.Format, p.BaseURL, p.Proxy, p.Model, p.APIKey, hint, p.RatePerSecond, p.RatePerMinute, p.ContextWindowK, p.ReasoningEffort, p.Priority, p.PoolExclude, p.ThinkingType, p.ID)
|
||||
_, err := d.Exec(`UPDATE llm_profiles SET name=$1,format=$2,base_url=NULLIF($3,''),proxy=NULLIF($4,''),model=$5,api_key=$6,api_key_hint=$7,rate_per_second=$8,rate_per_minute=$9,context_window_k=$10,reasoning_effort=$11,priority=$12,pool_exclude=$13,thinking_type=$14,streaming=$15 WHERE id=$16`,
|
||||
p.Name, p.Format, p.BaseURL, p.Proxy, p.Model, p.APIKey, hint, p.RatePerSecond, p.RatePerMinute, p.ContextWindowK, p.ReasoningEffort, p.Priority, p.PoolExclude, p.ThinkingType, p.Streaming, p.ID)
|
||||
return p.ID, err
|
||||
}
|
||||
|
||||
|
||||
@@ -303,6 +303,8 @@ CREATE TABLE IF NOT EXISTS llm_profiles (
|
||||
-- pool_exclude true=不作为故障转移目标(仍可被 agent/任务显式绑定使用)。
|
||||
priority INTEGER NOT NULL DEFAULT 0,
|
||||
pool_exclude BOOLEAN NOT NULL DEFAULT false,
|
||||
-- streaming=true(默认)走流式 SSE；false 走真·非流式(stream:false，一次性 JSON)。
|
||||
streaming BOOLEAN NOT NULL DEFAULT true,
|
||||
created_at TIMESTAMPTZ NOT NULL DEFAULT now(),
|
||||
updated_at TIMESTAMPTZ NOT NULL DEFAULT now()
|
||||
);
|
||||
@@ -313,6 +315,8 @@ CREATE TRIGGER trg_llm_upd BEFORE UPDATE ON llm_profiles
|
||||
-- 轮询顺位/排除标记；补旧库。默认 0 / false = 全部配置都参与轮询。
|
||||
ALTER TABLE llm_profiles ADD COLUMN IF NOT EXISTS priority INTEGER NOT NULL DEFAULT 0;
|
||||
ALTER TABLE llm_profiles ADD COLUMN IF NOT EXISTS pool_exclude BOOLEAN NOT NULL DEFAULT false;
|
||||
-- 流式开关；补旧库。默认 true = 保持既有的流式行为，旧配置无感升级。
|
||||
ALTER TABLE llm_profiles ADD COLUMN IF NOT EXISTS streaming BOOLEAN NOT NULL DEFAULT true;
|
||||
|
||||
-- 思考开关字段 thinking_type，从旧的单一 reasoning_effort 语义一次性拆分而来。
|
||||
-- schema.sql 每次启动都执行，故迁移必须只跑一次：仅当该列尚不存在时才回填，
|
||||
|
||||
Reference in new issue
Block a user