AskaEth
a6cb44b8a8
主人要求「链路流程应该有本地数据，不然每次重载之后页面就没数据了」： - 落盘：迁移 0022_flow_graphs（session_id/turn_id/graph_json/时间戳 + 唯一键 upsert）+ internal/runtime/flowgraph（RecordRound/Graph/Latest/Turns/Turn + 保留期 GC） - 写入：loop.EndRound **每轮一条**（锁外深拷贝后写；失败 WARN 且不打印图内容）；只落发布会话白名单（复用 flowSessionAllowed） - 读回：端点加 turnId + source(live/persisted/none) + turns[]（最新在前，空为 []）；**内存优先、miss 才落盘**（最新一轮永远最新） - 前端：flowGraphs 纳入 persistOptions.partialize（**刷新即刻有图**，version 3→4 + migrate）+ 历史轮选择器 + 诚实文案；顺带收 L1 转交的 flow/NodeDetail.tsx 节点日志倒序 - 新 env 键：FLOW_GRAPH_RETENTION=168h / FLOW_GRAPH_GC_INTERVAL=1h - 变异实测：每 span 一行 / 保留期守卫失效 / partialize 去掉 flowGraphs / 去掉内存优先 / 白名单失效 ⇒ 全部打红 - 3 个禁动面文件的纯追加改动（Flow.tsx 挂选择器、FlowSnapshotHint 纠已成假话的文案、其渲染断言随之更新）由主会话裁定认可 - 登记存疑：图桶持久化无体积上限；useFlowTurns 的 effect 无自动化证据（项目无 jsdom）
346 lines
14 KiB
Go
346 lines
14 KiB
Go
package gateway
|
||
|
||
// 本文件是链路图**落盘持久化 + 多轮历史回放**的跨层闭环（切片 F2；design/19 §4 P3）。
|
||
// 它用真的写入路径（`loop.FlowPublisher.EndRound` → `flowgraph.SQLStore.RecordRound`
|
||
// → 真 SQLite + 迁移 0022），再经**真的** HTTP handler 读回来，钉住三件只有端到端
|
||
// 才成立、也正是主人要的事：
|
||
//
|
||
// - **后端重启后仍能查回历史轮次的图**（本切片的内核）：换一套「进程」（新连接、
|
||
// 新 store、空内存的发布器），图仍从库里读得回来；
|
||
// - **最近一轮优先内存、miss 才落盘**：内存里有更新的那轮时，省略 turnId 不回落盘旧轮；
|
||
// - **白名单外会话既不落盘也查不到**：隔离渠道会话连图都不建，`turns` 恒空、`graph` 恒 null。
|
||
//
|
||
// 参数翻译与语义在 `admin_flow_graph_test.go`（v0.71 那批）与本文件互补。
|
||
|
||
import (
|
||
"context"
|
||
"database/sql"
|
||
"encoding/json"
|
||
"net/http"
|
||
"net/http/httptest"
|
||
"path/filepath"
|
||
"strings"
|
||
"testing"
|
||
"time"
|
||
|
||
"git.yeij.top/AskaEth/Cyrene_v2/internal/auth"
|
||
"git.yeij.top/AskaEth/Cyrene_v2/internal/db"
|
||
"git.yeij.top/AskaEth/Cyrene_v2/internal/runtime/flowgraph"
|
||
"git.yeij.top/AskaEth/Cyrene_v2/internal/runtime/loop"
|
||
"git.yeij.top/AskaEth/Cyrene_v2/internal/runtime/trace"
|
||
)
|
||
|
||
// flowGraphResponseDTO 是 v0.73 响应体的线上投影（含新增三键）。
|
||
type flowGraphResponseDTO struct {
|
||
SessionID string `json:"sessionId"`
|
||
TurnID string `json:"turnId"`
|
||
Source string `json:"source"`
|
||
Graph *struct {
|
||
ID string `json:"id"`
|
||
SessionID string `json:"sessionId"`
|
||
Nodes []struct {
|
||
Kind string `json:"kind"`
|
||
} `json:"nodes"`
|
||
} `json:"graph"`
|
||
Turns []struct {
|
||
TurnID string `json:"turnId"`
|
||
CreatedAt string `json:"createdAt"`
|
||
} `json:"turns"`
|
||
}
|
||
|
||
func decodeFlowGraphResp(t *testing.T, body []byte) flowGraphResponseDTO {
|
||
t.Helper()
|
||
var out flowGraphResponseDTO
|
||
if err := json.Unmarshal(decodeBody(t, body)["data"], &out); err != nil {
|
||
t.Fatalf("decode flow graph data: %v (body=%s)", err, body)
|
||
}
|
||
return out
|
||
}
|
||
|
||
// openFlowGraphStore 打开一个真实的 SQLite 文件库、跑迁移 0022，返回连接与存储句柄。
|
||
// caller 可显式 Close 连接以模拟进程退出（`t.Cleanup` 会兜底再关一次，无害）。
|
||
func openFlowGraphStore(t *testing.T, path string) (*sql.DB, *flowgraph.SQLStore) {
|
||
t.Helper()
|
||
ctx := context.Background()
|
||
conn, err := db.Open(ctx, path, gwTestLogger())
|
||
if err != nil {
|
||
t.Fatalf("open db: %v", err)
|
||
}
|
||
if err := db.MigrateUp(ctx, conn, gwTestLogger()); err != nil {
|
||
_ = conn.Close()
|
||
t.Fatalf("migrate: %v", err)
|
||
}
|
||
store, err := flowgraph.NewStore(flowgraph.Config{DB: conn, Log: gwTestLogger()})
|
||
if err != nil {
|
||
_ = conn.Close()
|
||
t.Fatalf("new flowgraph store: %v", err)
|
||
}
|
||
t.Cleanup(func() { _ = conn.Close() })
|
||
return conn, store
|
||
}
|
||
|
||
// flowPublisherWithStore 造一个把每轮收尾落进 store 的发布器（生产那条路径）。
|
||
func flowPublisherWithStore(store *flowgraph.SQLStore, allow func(string) bool) *loop.FlowPublisher {
|
||
return loop.NewFlowPublisher(func(string, any) {}, allow,
|
||
loop.WithRoundRecorder(store), loop.WithFlowLogger(gwTestLogger()))
|
||
}
|
||
|
||
// runFlowRound 让发布器跑完一整轮（起轮 → 一个 llm span 起止 → 收轮）。
|
||
func runFlowRound(p *loop.FlowPublisher, session, input, output string) {
|
||
p.StartRound(session, input)
|
||
sp := &trace.Span{ID: "span-llm", Name: "llm.call", Kind: trace.SpanKindLLM, StartedAt: time.Now()}
|
||
p.SpanStarted(session, sp)
|
||
sp.Finish(trace.SpanStatusOK)
|
||
p.SpanFinished(session, sp)
|
||
p.EndRound(session, output)
|
||
}
|
||
|
||
// mkFlowGraph 造一轮最小图，供直接写库（不经发布器）的用例使用。
|
||
func mkFlowGraph(session string, round int) *trace.FlowGraph {
|
||
return &trace.FlowGraph{
|
||
ID: session + "-" + itoa(round),
|
||
SessionID: session,
|
||
Nodes: []trace.FlowNode{
|
||
{ID: "input", Label: "你好", Kind: "input", Status: trace.FlowStatusDone},
|
||
{ID: "output", Label: "回复", Kind: "output", Status: trace.FlowStatusDone},
|
||
},
|
||
Edges: []trace.FlowEdge{{From: "input", To: "output"}},
|
||
}
|
||
}
|
||
|
||
func itoa(n int) string {
|
||
if n == 0 {
|
||
return "0"
|
||
}
|
||
digits := make([]byte, 0, 4)
|
||
for n > 0 {
|
||
digits = append([]byte{byte('0' + n%10)}, digits...)
|
||
n /= 10
|
||
}
|
||
return string(digits)
|
||
}
|
||
|
||
// flowURL 拼出该端点的 URL（带可选 turnId）。
|
||
func flowURL(ts *httptest.Server, session, turn string) string {
|
||
u := ts.URL + "/api/v1/admin/flow/graph?sessionId=" + session
|
||
if turn != "" {
|
||
u += "&turnId=" + turn
|
||
}
|
||
return u
|
||
}
|
||
|
||
// TestAdminFlowGraph_PersistedHistoryReadBack_正常：省略 turnId 回内存最近一轮
|
||
// （source live），带 turnId 回该历史轮（source persisted），且 turns 列全三轮、最新在前。
|
||
func TestAdminFlowGraph_PersistedHistoryReadBack_正常(t *testing.T) {
|
||
t.Parallel()
|
||
path := filepath.Join(t.TempDir(), "cyrene.db")
|
||
_, store := openFlowGraphStore(t, path)
|
||
pub := flowPublisherWithStore(store, func(string) bool { return true })
|
||
_, ts, jwt := buildEnv(t, cfgDefault(), Deps{Flow: pub, FlowGraphs: store})
|
||
tok := signAny(t, jwt, "acct", auth.RoleOwner)
|
||
|
||
for _, io := range []struct{ in, out string }{
|
||
{"第一轮", "答一"}, {"第二轮", "答二"}, {"第三轮", "答三"},
|
||
} {
|
||
runFlowRound(pub, "web", io.in, io.out)
|
||
}
|
||
|
||
resp, body := doReq(t, http.MethodGet, flowURL(ts, "web", ""), nil, tok)
|
||
if resp.StatusCode != http.StatusOK {
|
||
t.Fatalf("latest status = %d, want 200 (body=%s)", resp.StatusCode, body)
|
||
}
|
||
latest := decodeFlowGraphResp(t, body)
|
||
if latest.Source != "live" {
|
||
t.Fatalf("最新一轮 source = %q, want live（先查内存）", latest.Source)
|
||
}
|
||
if latest.TurnID != "web-3" {
|
||
t.Fatalf("最新一轮 turnId = %q, want web-3", latest.TurnID)
|
||
}
|
||
if len(latest.Turns) != 3 {
|
||
t.Fatalf("turns = %d, want 3（每轮一行，不是每 span 一行）", len(latest.Turns))
|
||
}
|
||
for i, want := range []string{"web-3", "web-2", "web-1"} {
|
||
if latest.Turns[i].TurnID != want {
|
||
t.Fatalf("turns[%d] = %q, want %q（最新在前）", i, latest.Turns[i].TurnID, want)
|
||
}
|
||
}
|
||
|
||
resp, body = doReq(t, http.MethodGet, flowURL(ts, "web", "web-1"), nil, tok)
|
||
if resp.StatusCode != http.StatusOK {
|
||
t.Fatalf("history status = %d, want 200 (body=%s)", resp.StatusCode, body)
|
||
}
|
||
hist := decodeFlowGraphResp(t, body)
|
||
if hist.Source != "persisted" {
|
||
t.Fatalf("历史轮 source = %q, want persisted（内存里没有 web-1）", hist.Source)
|
||
}
|
||
if hist.Graph == nil || hist.Graph.ID != "web-1" {
|
||
t.Fatalf("历史轮 graph = %+v, want web-1", hist.Graph)
|
||
}
|
||
kinds := map[string]bool{}
|
||
for _, n := range hist.Graph.Nodes {
|
||
kinds[n.Kind] = true
|
||
}
|
||
if !kinds["input"] || !kinds["output"] {
|
||
t.Fatalf("历史轮图缺主干端点: %+v", hist.Graph.Nodes)
|
||
}
|
||
}
|
||
|
||
// TestAdminFlowGraph_SurvivesRestart_正常 是本切片的**内核**（主人要求）：后端重启后
|
||
// 仍能查回历史轮次。换一套「进程」（新连接 + 新 store + **空内存**的新发布器），
|
||
// 唯一能拿到图的方式就是从库读回。
|
||
func TestAdminFlowGraph_SurvivesRestart_正常(t *testing.T) {
|
||
t.Parallel()
|
||
path := filepath.Join(t.TempDir(), "cyrene.db")
|
||
|
||
// 第一个「进程」：跑两轮，图进库。
|
||
conn1, store1 := openFlowGraphStore(t, path)
|
||
pub1 := flowPublisherWithStore(store1, func(string) bool { return true })
|
||
_, ts1, _ := buildEnv(t, cfgDefault(), Deps{Flow: pub1, FlowGraphs: store1})
|
||
runFlowRound(pub1, "web", "第一轮", "答一")
|
||
runFlowRound(pub1, "web", "第二轮", "答二")
|
||
ts1.Close()
|
||
_ = conn1.Close() // 模拟进程退出
|
||
|
||
// 第二个「进程」：空内存的新发布器 + 同一个文件库的新连接 / 新 store。
|
||
_, store2 := openFlowGraphStore(t, path)
|
||
pub2 := flowPublisherWithStore(store2, func(string) bool { return true })
|
||
_, ts2, jwt2 := buildEnv(t, cfgDefault(), Deps{Flow: pub2, FlowGraphs: store2})
|
||
tok := signAny(t, jwt2, "acct", auth.RoleOwner)
|
||
|
||
resp, body := doReq(t, http.MethodGet, flowURL(ts2, "web", ""), nil, tok)
|
||
if resp.StatusCode != http.StatusOK {
|
||
t.Fatalf("after restart status = %d, want 200 (body=%s)", resp.StatusCode, body)
|
||
}
|
||
got := decodeFlowGraphResp(t, body)
|
||
if got.Graph == nil {
|
||
t.Fatalf("重启后 graph = null，但该会话刚跑完两轮并已落盘（这正是主人要的「本地数据」）: %s", body)
|
||
}
|
||
if got.Source != "persisted" || got.TurnID != "web-2" {
|
||
t.Fatalf("重启后 source/turnId = %q/%q, want persisted/web-2", got.Source, got.TurnID)
|
||
}
|
||
|
||
// 更早那一轮也能按 turnId 取回。
|
||
resp, body = doReq(t, http.MethodGet, flowURL(ts2, "web", "web-1"), nil, tok)
|
||
if resp.StatusCode != http.StatusOK {
|
||
t.Fatalf("restart history status = %d, want 200 (body=%s)", resp.StatusCode, body)
|
||
}
|
||
hist := decodeFlowGraphResp(t, body)
|
||
if hist.Graph == nil || hist.Graph.ID != "web-1" {
|
||
t.Fatalf("重启后 web-1 = %+v, want web-1", hist.Graph)
|
||
}
|
||
}
|
||
|
||
// TestAdminFlowGraph_LatestPrefersMemory_正常 钉住「最新一轮永远最新」：内存里有更新的
|
||
// 那轮（web-2）而库里的最新只有 web-1 时，省略 turnId **必须**回内存那轮。
|
||
func TestAdminFlowGraph_LatestPrefersMemory_正常(t *testing.T) {
|
||
t.Parallel()
|
||
path := filepath.Join(t.TempDir(), "cyrene.db")
|
||
_, store := openFlowGraphStore(t, path)
|
||
if err := store.RecordRound(mkFlowGraph("web", 1)); err != nil {
|
||
t.Fatalf("seed persisted round: %v", err)
|
||
}
|
||
// 发布器**不挂 recorder**：内存跑两轮（web-2），但不再落盘 ⇒ 库里仍旧只有 web-1。
|
||
pub := loop.NewFlowPublisher(func(string, any) {}, func(string) bool { return true })
|
||
_, ts, jwt := buildEnv(t, cfgDefault(), Deps{Flow: pub, FlowGraphs: store})
|
||
tok := signAny(t, jwt, "acct", auth.RoleOwner)
|
||
|
||
runFlowRound(pub, "web", "第一轮", "答一")
|
||
runFlowRound(pub, "web", "第二轮", "答二")
|
||
|
||
resp, body := doReq(t, http.MethodGet, flowURL(ts, "web", ""), nil, tok)
|
||
if resp.StatusCode != http.StatusOK {
|
||
t.Fatalf("status = %d, want 200 (body=%s)", resp.StatusCode, body)
|
||
}
|
||
got := decodeFlowGraphResp(t, body)
|
||
if got.Source != "live" || got.TurnID != "web-2" {
|
||
t.Fatalf("source/turnId = %q/%q, want live/web-2（省略 turnId 时先查内存，绝不回落盘的旧轮）",
|
||
got.Source, got.TurnID)
|
||
}
|
||
}
|
||
|
||
// TestAdminFlowGraph_OutsideWhitelistNotPersisted_正常 钉住白名单复用：发布会话范围外
|
||
// 的会话（隔离渠道会话）连图都不建 ⇒ 既不落盘、也查不到，`turns` 恒空。
|
||
func TestAdminFlowGraph_OutsideWhitelistNotPersisted_正常(t *testing.T) {
|
||
t.Parallel()
|
||
path := filepath.Join(t.TempDir(), "cyrene.db")
|
||
_, store := openFlowGraphStore(t, path)
|
||
pub := flowPublisherWithStore(store, func(session string) bool { return session == "web" })
|
||
_, ts, jwt := buildEnv(t, cfgDefault(), Deps{Flow: pub, FlowGraphs: store})
|
||
tok := signAny(t, jwt, "acct", auth.RoleOwner)
|
||
|
||
runFlowRound(pub, "sess-im-stranger", "陌生人说话", "回应")
|
||
runFlowRound(pub, "web", "主人说话", "回应")
|
||
|
||
resp, body := doReq(t, http.MethodGet, flowURL(ts, "sess-im-stranger", ""), nil, tok)
|
||
if resp.StatusCode != http.StatusOK {
|
||
t.Fatalf("stranger status = %d, want 200 (body=%s)", resp.StatusCode, body)
|
||
}
|
||
stranger := decodeFlowGraphResp(t, body)
|
||
if stranger.Graph != nil {
|
||
t.Fatalf("白名单外会话不该有图，却拿到 %+v", stranger.Graph)
|
||
}
|
||
if stranger.Source != "none" {
|
||
t.Fatalf("source = %q, want none", stranger.Source)
|
||
}
|
||
if len(stranger.Turns) != 0 {
|
||
t.Fatalf("白名单外会话不该落盘（turns=%+v）", stranger.Turns)
|
||
}
|
||
// 反向：白名单内的会话照常可查、有落盘轮（保证上面的空不是端点整体坏了）。
|
||
_, body = doReq(t, http.MethodGet, flowURL(ts, "web", ""), nil, tok)
|
||
if inScope := decodeFlowGraphResp(t, body); inScope.Graph == nil || len(inScope.Turns) != 1 {
|
||
t.Fatalf("白名单内会话应有图且有一轮落盘: %s", body)
|
||
}
|
||
}
|
||
|
||
// TestAdminFlowGraph_UnknownTurnIsHonestNull_边界 钉住「该 turnId 不存在」是诚实空态：
|
||
// 200 + `graph:null` + `source:"none"`，绝不用别的轮次顶替。
|
||
func TestAdminFlowGraph_UnknownTurnIsHonestNull_边界(t *testing.T) {
|
||
t.Parallel()
|
||
path := filepath.Join(t.TempDir(), "cyrene.db")
|
||
_, store := openFlowGraphStore(t, path)
|
||
pub := flowPublisherWithStore(store, func(string) bool { return true })
|
||
_, ts, jwt := buildEnv(t, cfgDefault(), Deps{Flow: pub, FlowGraphs: store})
|
||
tok := signAny(t, jwt, "acct", auth.RoleOwner)
|
||
|
||
runFlowRound(pub, "web", "第一轮", "答一")
|
||
|
||
resp, body := doReq(t, http.MethodGet, flowURL(ts, "web", "web-999"), nil, tok)
|
||
if resp.StatusCode != http.StatusOK {
|
||
t.Fatalf("status = %d, want 200 (body=%s)", resp.StatusCode, body)
|
||
}
|
||
got := decodeFlowGraphResp(t, body)
|
||
if got.Graph != nil {
|
||
t.Fatalf("不存在的轮次 graph = %+v, want null", got.Graph)
|
||
}
|
||
if got.Source != "none" || got.TurnID != "" {
|
||
t.Fatalf("source/turnId = %q/%q, want none/\"\"", got.Source, got.TurnID)
|
||
}
|
||
// `graph` 键必须恒在（客户端靠它区分「没有图」与「字段缺失」）。
|
||
var envelope map[string]json.RawMessage
|
||
if err := json.Unmarshal(decodeBody(t, body)["data"], &envelope); err != nil {
|
||
t.Fatalf("decode envelope: %v", err)
|
||
}
|
||
if _, present := envelope["graph"]; !present {
|
||
t.Fatalf("`graph` 键必须恒在（空时是 null）: %s", body)
|
||
}
|
||
}
|
||
|
||
// TestAdminFlowGraph_EmptyTurnsSerialiseAsArray_边界 钉住线上空形状：没有落盘轮次时
|
||
// `turns` 是 `[]`（客户端不必在 null 与空数组之间分叉）。
|
||
func TestAdminFlowGraph_EmptyTurnsSerialiseAsArray_边界(t *testing.T) {
|
||
t.Parallel()
|
||
path := filepath.Join(t.TempDir(), "cyrene.db")
|
||
_, store := openFlowGraphStore(t, path)
|
||
pub := flowPublisherWithStore(store, func(string) bool { return true })
|
||
_, ts, jwt := buildEnv(t, cfgDefault(), Deps{Flow: pub, FlowGraphs: store})
|
||
tok := signAny(t, jwt, "acct", auth.RoleOwner)
|
||
|
||
resp, body := doReq(t, http.MethodGet, flowURL(ts, "web", ""), nil, tok)
|
||
if resp.StatusCode != http.StatusOK {
|
||
t.Fatalf("status = %d, want 200 (body=%s)", resp.StatusCode, body)
|
||
}
|
||
if !strings.Contains(string(body), `"turns":[]`) {
|
||
t.Fatalf("空 turns 必须序列化为 [] 而非 null: %s", body)
|
||
}
|
||
}
|