@badger_badger/badger_hq 1.5.1 → 1.7.0

package/CHANGELOG.md

@@ -1,5 +1,49 @@
 # badger_hq
 
+## 1.7.0
+
+### Minor Changes
+
+- e85a4c8: Monitor tasks as a batch of sources.
+
+  - `AgentTaskSchema.sources` (`AgentTaskSourceSchema[]`, default `[]`): the sources attached to a Monitor task other than its `target` — `{ platform, url, label? }`, http(s) only, no internal source id.
+    Each also carries `lastCapturedAt?`, `dueAt?` (under current urgency; absent when parked) and `accessProblem?`.
+  - `AgentAssignmentVersionSchema.sourcesUpdatedAt?`: the latest change to a Monitor task's attached sources, so a visit refreshes agents without bumping the task version.
+  - `MONITOR_TASK_MAX_SOURCES` (20): the most sources one Monitor task may have attached (enforced by hero-service).
+  - `Source.visitIntervalMs?`: the activity-based wait set by the last visit, before urgency.
+  - Source schedule helpers (pure, browser-safe): `effectiveDueAt` applies current urgency (`URGENCY_FACTOR`: emergency ×0.5; mission priority critical ×0.5, high ×0.75, medium ×1, low ×1.5) to the base interval, clamped to 1–72 h; `effectiveInterval` (that interval, for `overdueLevel`); `overdueLevel` (`none` / `late` / `missed` after one more interval; `OverdueLevelSchema`); `baseVisitInterval`, `urgencyFactor`, `clampInterval`.
+    A source without `visitIntervalMs` (scheduled before it existed) falls back to its last `nextVisitAt − lastCollectedAt` gap, which already includes the urgency of that visit and is not scaled again.
+  - Sources index `{ orgId, taskIds }` (multikey): the sources attached to a task.
+
+  Wire-compatible: existing documents and payloads still parse. The inferred `AgentTask` type now always has `sources`, so code that builds `AgentTask` objects must set it (`[]` when none).
+
+## 1.6.0
+
+### Minor Changes
+
+- d6d284c: Two task kinds with their own shape.
+
+  - `Task.target?` (`TaskTargetSchema`): the one source a task points at — `{ sourceId?, platform, url, label?, accessIssue?, note? }`. Required on Monitor tasks by hero-service; optional on Discovery tasks (search inside this source).
+  - `Task.captureWindow?` (`CaptureWindowSchema`): what a Monitor task captures — `{ mode: 'last_days', days: 1..90 }` or `{ mode: 'range', start, end }` with `end > start`. Stored flat in Mongo (`captureWindowSchema`), validated as a discriminated union in Zod.
+  - New enums `ACCESS_ISSUE_VALUES` (`login_required`, `private`, `bot_detection`, `captcha`, `geo_blocked`, `rate_limited`, `not_found`, `no_collector`, `collector_full`, `other`) and `CAPTURE_WINDOW_MODE_VALUES`.
+  - `AgentTaskSchema` gains `target?` (as `AgentTaskTargetSchema`, without `sourceId`) and `captureWindow?`.
+  - `AgentSession.captureWindow?` (`{ start, end }`): the task's window resolved to dates when the session started.
+  - `AgentSessionViewSchema` gains `target?` and `captureWindow?` (resolved) so the session header can show what is being captured.
+  - Task-tagged captures (agent-service analyses them in the background, with no live session):
+    - `Capture.sessionId` is now optional.
+    - New optional `assignmentId`, `taskId`, `missionId` and `userId`.
+    - New `attempts` (default 0) and `claimedAt` for the capture worker.
+    - New indexes `{status, createdAt}` and `{orgId, assignmentId, userId, capturedAt}`.
+  - `HoursLog.assignmentId?` and `HoursLog.source?` (`HOURS_LOG_SOURCE_VALUES`: `manual`, `inferred`), plus a unique partial index for one inferred row per person, assignment and day.
+
+  All fields are optional; existing documents and consumers are unaffected.
+
+  Validation:
+
+  - `TaskTargetSchema.url` accepts only `http:` / `https:` URLs (new `isHttpUrl` helper).
+  - `CaptureWindowSchema` `range` dates must be a `Date` or an ISO 8601 string with an explicit offset; `end` is exclusive and the range is at most `MAX_CAPTURE_WINDOW_DAYS` long.
+  - New index `{ orgId, target.sourceId }` on `tasks` (partial on `target.sourceId`).
+
 ## 1.4.0
 
 ### Minor Changes

package/README.md

@@ -1,4 +1,4 @@
-# badger_hq
+# @badger_badger/badger_hq
 
 Shared types and schemas for **Harvester V2** (Badger HQ). Every service imports it: the Next.js web app,
 the Chrome extension, the Fastify core API, the worker, the agent and the scrapers.
@@ -7,10 +7,10 @@ It ships two entry points:
 
 | Entry | Import | Contents | Safe in |
 |---|---|---|---|
-| `.` | `badger_hq` | Zod schemas, inferred TS types, enums, `COLLECTIONS`, pagination / `ApiError` helpers | Browser **and** Node |
-| `./mongo` | `badger_hq/mongo` | Mongoose schemas, index definitions, `registerModels(connection)` | Node only |
+| `.` | `@badger_badger/badger_hq` | Zod schemas, inferred TS types, enums, `COLLECTIONS`, pagination / `ApiError` helpers | Browser **and** Node |
+| `./mongo` | `@badger_badger/badger_hq/mongo` | Mongoose schemas, index definitions, `registerModels(connection)` | Node only |
 
-Nothing reachable from `badger_hq` imports `mongoose`, `bson` or a Node built-in; a test bundles it with esbuild
+Nothing reachable from `@badger_badger/badger_hq` imports `mongoose`, `bson` or a Node built-in; a test bundles it with esbuild
 for `platform: 'browser'` and fails if that ever changes.
 
 ## Install
@@ -18,7 +18,7 @@ for `platform: 'browser'` and fails if that ever changes.
 The package is published publicly to the npm registry.
 
 ```sh
-npm install badger_hq zod
+npm install @badger_badger/badger_hq zod
 # server-side packages also need:
 npm install mongoose
 ```
@@ -37,7 +37,7 @@ import {
   PaginationQuerySchema,
   type Paginated,
   type ApiError,
-} from 'badger_hq';
+} from '@badger_badger/badger_hq';
 ```
 
 Per entity (`Mission` shown; the same pattern applies to every entity):
@@ -57,7 +57,7 @@ Dates use `z.coerce.date()`, so JSON ISO strings parse to `Date`.
 ```tsx
 import { useForm } from 'react-hook-form';
 import { zodResolver } from '@hookform/resolvers/zod';
-import { MissionCreateSchema, type MissionCreateInput, MISSION_PRIORITY_VALUES } from 'badger_hq';
+import { MissionCreateSchema, type MissionCreateInput, MISSION_PRIORITY_VALUES } from '@badger_badger/badger_hq';
 
 export function NewMissionForm() {
   const form = useForm<MissionCreateInput>({
@@ -86,7 +86,7 @@ export function NewMissionForm() {
 ### Example: parsing an API response
 
 ```ts
-import { PaginatedSchema, TaskSchema } from 'badger_hq';
+import { PaginatedSchema, TaskSchema } from '@badger_badger/badger_hq';
 
 const PageOfTasks = PaginatedSchema(TaskSchema);
 const page = PageOfTasks.parse(await res.json()); // page.items[0].createdAt is a Date
@@ -95,7 +95,7 @@ const page = PageOfTasks.parse(await res.json()); // page.items[0].createdAt is
 ## Entry point `./mongo` — Mongoose
 
 ```ts
-import { registerModels, COLLECTIONS, INDEXES, incrementTaskVersion, type HarvesterModels } from 'badger_hq/mongo';
+import { registerModels, COLLECTIONS, INDEXES, incrementTaskVersion, type HarvesterModels } from '@badger_badger/badger_hq/mongo';
 ```
 
 The package **never connects**. You own the `mongoose.Connection`; `registerModels(conn)` registers the models on it
@@ -139,8 +139,8 @@ with the Zod `MissionSchema`, `TaskSchema`, … from the browser entry.
 ```ts
 import Fastify from 'fastify';
 import mongoose from 'mongoose';
-import { registerModels } from 'badger_hq/mongo';
-import { MissionCreateSchema, MissionSchema, type Mission } from 'badger_hq';
+import { registerModels } from '@badger_badger/badger_hq/mongo';
+import { MissionCreateSchema, MissionSchema, type Mission } from '@badger_badger/badger_hq';
 
 const app = Fastify();
 const conn = mongoose.createConnection(process.env.MONGO_URL!);
@@ -205,6 +205,10 @@ Each enum is defined once in `enums.ts` as `z.enum([...])` plus a `*_VALUES` con
 - **Mission** — a priority (PIR) in plain words; the top-level unit of work.
 - **Task** — a unit of work under a mission, `source_discovery` or `monitor`; what gets assigned to a scout or analyst.
   There is no "subtask" concept.
+  - **Discovery** — a free search: phrases (`keywords`), a short explanation (`description`), search queries, and
+    optionally one `target` source to search inside.
+  - **Monitor** — collect **one** `target` source a scraper can't (see `accessIssue`) over a `captureWindow`
+    (the last N days, or a fixed range).
 
 ## Versioning
 
@@ -268,7 +272,7 @@ Choices made where the spec left room, kept consistent with the rest of it:
    `suppressReservedKeysWarning` set on that schema.
 8. **Mongoose schema export names** are lowerCamelCase (`missionSchema`) to avoid colliding with the Zod
    `MissionSchema` when both entry points are imported in one file.
-9. **Package name** is `badger_hq` (set in `package.json`), published publicly to the npm registry.
+9. **Package name** is `@badger_badger/badger_hq` (set in `package.json`), published publicly to the npm registry.
 10. **Multiple assignees per task (1.0.0).** The "one active assignee per task" unique index was replaced by
     `{ taskId, assigneeUserId }` (partial, `status: 'active'`). Limits are enforced by `assigneeLoads`, not by an index,
     because a scout's limit spans organizations while an analyst's is per org.