feat: extract structured data from notifications (#617)

* feat: extract structured data from notifications

* fix: typing errors

* chore: add work-log

* chore: update todo list

* fix: harden notification extraction with type safety, caching, and unit tests

- Cache extractStructuredData result instead of calling twice per notification
- Replace unsafe double-cast (as unknown as LinkedInNotification) with properly typed object
- Remove dead notification_category filter condition from isMatch()
- Add module-level _extractNotificationStructuredData export for testing
- Add --types option to CLI notifications list command for parity with MCP
- Add 18 unit tests covering all 9 regex parsers and edge cases

Author: sigvardt

.opencode/todo.md

@@ -1,30 +1,22 @@
-# Mission Tasks for Issue #609 - Newsletters & Articles: make it production-grade
+# Mission: Notifications: make it production-grade (#613)
 
-## Phase 1: Support Rich Text and Images in Articles
-- [ ] Investigate how LinkedIn rich text editor works (contenteditable vs specific DOM structure)
-- [ ] Add support for cover image URL in `PrepareCreateArticleInput` and `PrepareCreateNewsletterInput`
-- [ ] Add support for HTML/Markdown body parsing and inserting rich text
+## M1: Analysis & Infrastructure
+### T1.1: Identify Notification Types & Extracted Data | agent:Planner
+- [x] S1.1.1: Document parsing logic for the 9 specified notification types | size:M
+- [x] S1.1.2: Define updated `LinkedInNotification` interface with `extracted_data` | size:S
 
-## Phase 2: Newsletter Metadata Editing
-- [ ] Create `PrepareUpdateNewsletterInput` interface
-- [ ] Implement `prepareUpdate` in `LinkedInNewslettersService`
-- [ ] Implement `UpdateNewsletterActionExecutor`
-- [ ] Register tool in MCP `linkedin.newsletter.prepare_update`
-- [ ] Add CLI command for updating newsletter
+## M2: Implementation
+### T2.1: Implement Data Extraction (Rich Data) | agent:Worker | depends:T1.1
+- [x] S2.1.1: Update `extractNotificationSnapshots` to extract structured data for all required types | size:L
+- [x] S2.1.2: Add parsing utilities for metrics (view counts, names, etc) | size:M
 
-## Phase 3: Newsletter Editions List and Stats
-- [ ] Update `list` method in `LinkedInNewslettersService` to fetch stats (subscribers, views)
-- [ ] Add new interface `ListNewsletterEditionsInput` and `listEditions` method to list individual editions for a newsletter
-- [ ] Register `linkedin.newsletter.list_editions` tool in MCP
-- [ ] Add CLI command for listing editions
+### T2.2: Implement Type Filtering & Pagination | agent:Worker | depends:T2.1
+- [x] S2.2.1: Add `types` filtering to `ListNotificationsInput` and `listNotifications` method | size:S
+- [x] S2.2.2: Add `types` parameter to MCP tool `notifications_list` | size:S
+- [x] S2.2.3: Update pagination logic in `loadNotificationSnapshots` to support fetching longer feeds without fixed scroll limits | size:M
 
-## Phase 4: Share Newsletter
-- [ ] Investigate how sharing works for newsletters (Share button -> modal -> post)
-- [ ] Implement `prepareShare` in `LinkedInNewslettersService`
-- [ ] Implement `ShareNewsletterActionExecutor`
-- [ ] Register MCP tool and CLI command
-
-## Phase 5: Testing and Polish
-- [ ] Add unit tests for new methods in `linkedinPublishing.test.ts`
-- [ ] Run e2e tests
-- [ ] Address edge cases: draft vs published, editing after publish, newsletter with zero editions
+## M3: Verification & Integration
+### T3.1: Testing & Quality Gates | agent:Reviewer | depends:M2
+- [x] S3.1.1: Run unit tests and e2e tests for notifications | size:M
+- [x] S3.1.2: Run lint and typecheck | size:S
+- [x] S3.1.3: Wait for CI and create PR | size:S

.opencode/work-log.md

@@ -0,0 +1,11 @@
+# Work Log
+
+## Active Sessions
+
+## Completed Units (Ready for Integration)
+| File | Session | Unit Test | Timestamp |
+|------|---------|-----------|-----------|
+| packages/core/src/linkedinNotifications.ts | Worker | pass | 2026-03-23T15:42:58.981Z |
+| packages/mcp/src/bin/linkedin-mcp.ts | Worker | pass | 2026-03-23T15:42:58.981Z |
+
+## Pending Integration

packages/cli/src/bin/linkedin.ts

@@ -8549,6 +8549,7 @@ async function runNotificationsList(
   input: {
     profileName: string;
     limit: number;
+    types?: string[];
   },
   cdpUrl?: string,
 ): Promise<void> {
@@ -8558,11 +8559,13 @@ async function runNotificationsList(
     runtime.logger.log("info", "cli.notifications.list.start", {
       profileName: input.profileName,
       limit: input.limit,
+      types: input.types,
     });
 
     const notifications = await runtime.notifications.listNotifications({
       profileName: input.profileName,
       limit: input.limit,
+      ...(input.types ? { types: input.types } : {}),
     });
 
     runtime.logger.log("info", "cli.notifications.list.done", {
@@ -12889,11 +12892,13 @@ export function createCliProgram(): Command {
     .description("List your LinkedIn notifications")
     .option("-p, --profile <profile>", "Profile name", "default")
     .option("-l, --limit <limit>", "Max notifications to return", "20")
-    .action(async (options: { profile: string; limit: string }) => {
+    .option("-t, --types <types...>", "Filter by notification types")
+    .action(async (options: { profile: string; limit: string; types?: string[] }) => {
       await runNotificationsList(
         {
           profileName: options.profile,
           limit: coercePositiveInt(options.limit, "limit"),
+          ...(options.types ? { types: options.types } : {}),
         },
         readCdpUrl(),
       );

packages/core/src/__tests__/linkedinNotifications.test.ts

@@ -4,6 +4,7 @@ import {
   LINKEDIN_NOTIFICATION_PREFERENCE_CHANNELS,
   _hashNotificationFingerprint as hashNotificationFingerprint,
   _legacyHashNotificationFingerprint as legacyHashNotificationFingerprint,
+  _extractNotificationStructuredData as extractNotificationStructuredData,
   _normalizeNotificationLink as normalizeNotificationLink,
   _stripVolatileContent as stripVolatileContent,
   LinkedInNotificationsService,
@@ -1083,3 +1084,122 @@ describe("legacyHashNotificationFingerprint", () => {
     expect(id).toMatch(/^notif_[0-9a-f]{16}$/);
   });
 });
+
+describe("extractNotificationStructuredData", () => {
+  it("extracts post analytics from 'Your post has ... views'", () => {
+    expect(extractNotificationStructuredData("Your post has 1,234 views")).toEqual({
+      views: 1234,
+    });
+  });
+
+  it("extracts post analytics from 'people viewed your post'", () => {
+    expect(extractNotificationStructuredData("500 people viewed your post")).toEqual({
+      views: 500,
+    });
+  });
+
+  it("extracts profile views from 'people viewed your profile'", () => {
+    expect(extractNotificationStructuredData("42 people viewed your profile")).toEqual({
+      profile_views: 42,
+    });
+  });
+
+  it("extracts profile views from 'Your profile was viewed by ... people'", () => {
+    expect(
+      extractNotificationStructuredData("Your profile was viewed by 100 people"),
+    ).toEqual({
+      profile_views: 100,
+    });
+  });
+
+  it("extracts search appearances", () => {
+    expect(extractNotificationStructuredData("You appeared in 15 searches")).toEqual({
+      search_appearances: 15,
+    });
+  });
+
+  it("extracts mention sender", () => {
+    expect(extractNotificationStructuredData("John Smith mentioned you")).toEqual({
+      mentioned_by: "John Smith",
+    });
+  });
+
+  it("extracts connection sender from connection request", () => {
+    expect(
+      extractNotificationStructuredData("Jane Doe sent you a connection request"),
+    ).toEqual({
+      sender: "Jane Doe",
+    });
+  });
+
+  it("extracts connection sender from accepted connection", () => {
+    expect(extractNotificationStructuredData("Bob accepted your connection")).toEqual({
+      sender: "Bob",
+    });
+  });
+
+  it("extracts newsletter subscriber count", () => {
+    expect(extractNotificationStructuredData("1,500 people subscribed to")).toEqual({
+      subscriber_count: 1500,
+    });
+  });
+
+  it("extracts newsletter subscriber name", () => {
+    expect(extractNotificationStructuredData("Alice subscribed to")).toEqual({
+      subscriber: "Alice",
+    });
+  });
+
+  it("extracts job alert count and title", () => {
+    expect(
+      extractNotificationStructuredData('5 new jobs for "Software Engineer"'),
+    ).toEqual({
+      job_count: 5,
+      job_title: "Software Engineer",
+    });
+  });
+
+  it("extracts single job alert title", () => {
+    expect(extractNotificationStructuredData('new job for "Designer"')).toEqual({
+      job_title: "Designer",
+    });
+  });
+
+  it("extracts company name from posted notification", () => {
+    expect(extractNotificationStructuredData("Google posted:")).toEqual({
+      company_name: "Google",
+    });
+  });
+
+  it("extracts company name from shared post notification", () => {
+    expect(extractNotificationStructuredData("Microsoft shared a post:")).toEqual({
+      company_name: "Microsoft",
+    });
+  });
+
+  it("extracts trending topic", () => {
+    expect(
+      extractNotificationStructuredData("Trending: AI advances in 2025"),
+    ).toEqual({
+      topic: "AI advances in 2025",
+    });
+  });
+
+  it("returns undefined when message does not match known parsers", () => {
+    expect(
+      extractNotificationStructuredData("Someone liked your comment"),
+    ).toBeUndefined();
+  });
+
+  it("returns undefined for empty string", () => {
+    expect(extractNotificationStructuredData("")).toBeUndefined();
+  });
+
+  it("normalizes excessive whitespace before parsing", () => {
+    expect(
+      extractNotificationStructuredData("   John   Smith   mentioned\n\t you   "),
+    ).toEqual({
+      mentioned_by: "John Smith",
+    });
+  });
+});