memmy-agent: Let every AI remember the same you - V1.0.5

.github/CODEOWNERS

@@ -0,0 +1,4 @@
+*                  @ZongYue99
+/App/memmy-agent/  @Wang-Daoji @ZongYue99
+/Memory/           @hijzy @ZongYue99
+/scripts/          @wustzdy @ZongYue99

.github/release-notes/v1.0.5.md

@@ -0,0 +1,30 @@
+# Memmy v1.0.5
+
+## Highlights
+
+- Improved time-aware Memory recall. Requests with a concrete time range can now retrieve matching L1 trace summaries in chronological context.
+- Made episodic Memory capture more reliable by committing routing, raw turns, and L1 records only after a turn completes. Cancelled, interrupted, and structurally incomplete turns no longer reserve an episode or create durable L1 memories.
+- Added QR-based Feishu bot setup through the official Lark flow, while retaining manual App ID and App Secret configuration as a fallback.
+
+## Agent and integration improvements
+
+- Refined the first-report handoff for detected Cursor, Claude Code, Codex, OpenCode, OpenClaw, Hermes, and WorkBuddy installations, with prompt-copy fallback and Memory lookup verification when permission allows.
+- Codex integration setup now registers, trusts, and verifies only the two user-level Memmy hooks it installs.
+- Structured quota-exhaustion errors are now propagated consistently across supported providers and the desktop UI, without mistaking ordinary response text for a quota failure.
+- Invalid existing Agent configuration files now fail with a clear error instead of silently reverting to defaults; missing configuration files still retain first-run behavior.
+- When the Memmy Memory service is unavailable, Agent turns continue but show an explicit warning instead of making a silent fallback look like an empty recall.
+
+## Desktop and Memory improvements
+
+- Reworked the onboarding report around user preferences and the latest task, and now preserves the generated report as a real chat and handoff memory instead of rerunning the Agent after onboarding.
+- Expanded the guided product tour across Memory logs, connected Agents, sync settings, the Memory overview, and tools; the flow adapts when scan permission is declined.
+- Memory detail views now provide direct navigation between related traces, tasks, policies, world-model entries, and skills.
+- Strengthened L3 evidence validation and episode reward processing, including safer handling of negative outcomes and reusable avoidance guidance.
+- Agent-source scan permission is now installation-scoped, so the same local installation uses one consistent choice across account contexts.
+- Settings now show configured invitation reward amounts and clearer BYOK usage copy.
+
+## Packaging notes
+
+- Consolidated macOS and Windows packaging behind parameterized scripts for version, architecture, edition, and signing mode.
+- Desktop packages now exclude dependency tests, documentation, examples, and coverage files from the runtime payload.
+- Packaged builds now prepare and include the default local embedding model as an application resource.

App/backend/local-api-contracts/src/index.ts

@@ -355,26 +355,6 @@ export const AgentSourceScanInputSchema = z.preprocess(
 );
 export type AgentSourceScanInput = z.infer<typeof AgentSourceScanInputSchema>;
 
-export const OnboardingInsightActionTypeSchema = z.enum([
-    "continue_task",
-    "cross_agent_synthesis",
-    "decision_doc",
-    "problem_diagnosis",
-    "open_ended"
-]);
-export type OnboardingInsightActionType = z.infer<typeof OnboardingInsightActionTypeSchema>;
-
-export const OnboardingInsightActionSchema = z.object({
-    type: OnboardingInsightActionTypeSchema,
-    buttonLabel: z.string().min(1),
-    description: z.string().min(1),
-    contextSummary: z.string().min(1),
-    relatedAgents: z.array(z.string().min(1)).default([]),
-    topicKeywords: z.array(z.string().min(1)).default([]),
-    suggestedPrompt: z.string().min(1)
-});
-export type OnboardingInsightAction = z.infer<typeof OnboardingInsightActionSchema>;
-
 export const OnboardingInsightReportInputSchema = z.object({
     locale: z.enum(["zh-CN", "en-US"]).optional(),
     stream: z.boolean().optional()
@@ -386,6 +366,8 @@ export const OnboardingInsightDiagnosticsSchema = z.object({
     sampledQueryCount: z.number().int().nonnegative(),
     usedLlm: z.boolean(),
     elapsedMs: z.number().int().nonnegative(),
+    reportLanguage: z.enum(["zh-CN", "en-US"]).optional(),
+    latestWorkspacePath: z.string().nullable().optional(),
     agents: z.array(z.object({
         sourceId: z.string().min(1),
         displayName: z.string().min(1),
@@ -399,8 +381,6 @@ export type OnboardingInsightDiagnostics = z.infer<typeof OnboardingInsightDiagn
 export const OnboardingInsightReportResponseSchema = z.object({
     status: z.enum(["ready", "fallback", "skipped"]),
     reportMarkdown: z.string(),
-    primaryAction: OnboardingInsightActionSchema.optional(),
-    secondaryActions: z.array(OnboardingInsightActionSchema).default([]),
     diagnostics: OnboardingInsightDiagnosticsSchema
 });
 export type OnboardingInsightReportResponse = z.infer<typeof OnboardingInsightReportResponseSchema>;

App/backend/local-api-contracts/src/memory-runtime.ts

@@ -300,7 +300,6 @@ export const StartTurnOutputSchema = z.object({
   turnId: NonEmptyStringSchema,
   contextPacketId: NonEmptyStringSchema,
   sessionId: NonEmptyStringSchema,
-  episodeId: NonEmptyStringSchema,
   injectedContext: InjectedContextSchema,
   searchEventId: NonEmptyStringSchema,
   sourceMemoryIds: z.array(NonEmptyStringSchema),
@@ -333,11 +332,14 @@ export const CompleteTurnOutputSchema = z.object({
   sessionId: NonEmptyStringSchema,
   episodeId: NonEmptyStringSchema,
   rawTurnId: NonEmptyStringSchema,
-  l1MemoryId: NonEmptyStringSchema,
+  l1MemoryId: z.string(),
+  l1MemoryIds: z.array(NonEmptyStringSchema),
+  closedEpisodeIds: z.array(NonEmptyStringSchema),
   scheduledEvolution: z.boolean(),
   jobs: z.array(JobRefSchema),
   changeSeq: z.number().int().nonnegative(),
-  serverTime: IsoTimeSchema
+  serverTime: IsoTimeSchema,
+  duplicate: z.boolean().optional()
 });
 export type CompleteTurnOutput = z.infer<typeof CompleteTurnOutputSchema>;
 
@@ -423,12 +425,15 @@ export const GetMemoryOutputSchema = z.object({
     worldModel: z
       .object({
         sourceMemoryIds: z.array(NonEmptyStringSchema),
-        confidence: z.number().optional()
+        confidence: z.number().optional(),
+        summary: z.string().optional()
       })
       .optional(),
     skill: z
       .object({
         invocationGuide: z.string(),
+        retrievalBlurb: z.string().optional(),
+        triggerContext: z.string().optional(),
         procedure: z.array(z.string()).optional(),
         sourcePolicyIds: z.array(NonEmptyStringSchema),
         sourceWorldModelIds: z.array(NonEmptyStringSchema),

App/backend/src/adapters/inbound/local-api/routes/onboarding-insight.ts

@@ -30,7 +30,6 @@ export function registerOnboardingInsightRoutes(
         return reply.send(OnboardingInsightReportResponseSchema.parse({
           status: "skipped",
           reportMarkdown: "",
-          secondaryActions: [],
           diagnostics: {
             discoveredAgentCount: 0,
             sampledQueryCount: 0,
@@ -60,7 +59,6 @@ export function registerOnboardingInsightRoutes(
           response: {
             status: "skipped",
             reportMarkdown: "",
-            secondaryActions: [],
             diagnostics: {
               discoveredAgentCount: 0,
               sampledQueryCount: 0,

App/backend/src/adapters/inbound/local-api/tests/agent-runtime-routes.test.ts

@@ -439,7 +439,6 @@ function startTurnOutput() {
     turnId: "turn-1",
     contextPacketId: "context-1",
     sessionId: "session-1",
-    episodeId: "episode-1",
     injectedContext: { markdown: "", sections: [] },
     searchEventId: "search-1",
     sourceMemoryIds: [],
@@ -454,6 +453,8 @@ function completeTurnOutput() {
     turnId: "turn-1",
     sessionId: "session-1",
     l1MemoryId: "memory-1",
+    l1MemoryIds: ["memory-1"],
+    closedEpisodeIds: [],
     rawTurnId: "raw-1",
     episodeId: "episode-1",
     scheduledEvolution: false,

App/backend/src/adapters/inbound/local-api/tests/local-app-route-inventory.test.ts

@@ -238,7 +238,6 @@ function createServer(): FastifyInstance {
         return {
           status: "ready",
           reportMarkdown: "初见报告",
-          secondaryActions: [],
           diagnostics: {
             discoveredAgentCount: 1,
             sampledQueryCount: 1,
@@ -254,7 +253,6 @@ function createServer(): FastifyInstance {
           response: {
             status: "ready",
             reportMarkdown: "初见报告",
-            secondaryActions: [],
             diagnostics: {
               discoveredAgentCount: 1,
               sampledQueryCount: 1,

App/backend/src/adapters/inbound/local-api/tests/onboarding-insight-route.test.ts

@@ -37,7 +37,6 @@ describe("onboarding insight local api routes", () => {
     expect(response.json()).toEqual({
       status: "skipped",
       reportMarkdown: "",
-      secondaryActions: [],
       diagnostics: {
         discoveredAgentCount: 0,
         sampledQueryCount: 0,
@@ -49,7 +48,7 @@ describe("onboarding insight local api routes", () => {
     expect(generateReport).not.toHaveBeenCalled();
   });
 
-  it("streams first-login report chunks and final actions when scan permission is granted", async () => {
+  it("streams first-login report chunks and the final report when scan permission is granted", async () => {
     app = createServer({
       permissionManager: createPermissionManager("scan_only"),
       onboardingInsight: {
@@ -63,16 +62,6 @@ describe("onboarding insight local api routes", () => {
             response: {
               status: "ready",
               reportMarkdown: "你好",
-              primaryAction: {
-                type: "continue_task",
-                buttonLabel: "继续",
-                description: "继续任务",
-                contextSummary: "上下文",
-                relatedAgents: ["Codex"],
-                topicKeywords: ["Memory"],
-                suggestedPrompt: "继续任务"
-              },
-              secondaryActions: [],
               diagnostics: {
                 discoveredAgentCount: 1,
                 sampledQueryCount: 1,

App/backend/src/adapters/outbound/agent-source/insight-sampler-types.ts

@@ -18,12 +18,30 @@ export interface OnboardingSampledQuery {
   workspacePath: string | null;
 }
 
+export interface OnboardingSampledMessage extends OnboardingSampledQuery {
+  role: "user" | "assistant" | "tool";
+}
+
+export interface OnboardingConversationReference {
+  sourceId: string;
+  displayName: string;
+  conversationId: string;
+  latestActivityAt: string;
+  workspacePath: string | null;
+}
+
+export interface OnboardingConversationWindow extends OnboardingConversationReference {
+  messages: OnboardingSampledMessage[];
+}
+
 export interface OnboardingSampleResult {
   sourceId: string;
   displayName: string;
   recentSessionCount: number;
   latestActivityAt: string | null;
   queries: OnboardingSampledQuery[];
+  /** Recent visible messages used only to identify the newest conversation. */
+  recentMessages?: OnboardingSampledMessage[];
   errors: Array<{ target: string; reason: string }>;
 }
 
@@ -34,6 +52,13 @@ export interface OnboardingInsightSampler {
   sampleRecentUserQueries(options: OnboardingInsightSampleOptions): Promise<OnboardingSampleResult>;
 }
 
+export interface OnboardingConversationWindowReader {
+  readConversation(
+    reference: OnboardingConversationReference,
+    options: Pick<OnboardingInsightSampleOptions, "maxQueryChars" | "deadlineMs" | "signal">
+  ): Promise<OnboardingConversationWindow | null>;
+}
+
 export function emptyOnboardingSampleResult(input: {
   sourceId: string;
   displayName: string;