@mean-weasel/lineage 0.1.6 → 0.1.8

package/CHANGELOG.md

@@ -1,5 +1,19 @@
 # Changelog
 
+## 0.1.8
+
+- Add durable selection packet export so agents can hand selected Lineage assets to GrowthOps without scraping UI state or copying local paths.
+- Add Agent OS adoption guidance for Lineage agents and operators.
+- Improve popover media previews, node actions, and image expansion controls for faster asset inspection.
+- Clean up the Lineage shell navigation, toolbar, and side-panel layout for a more focused workspace.
+
+## 0.1.7
+
+- Add a claim-aware lineage task queue for per-image iteration and re-roll work, including task instructions, comments, cancellation, and human override controls.
+- Add visible agent/task state in the lineage canvas and side panel so humans can see when a task is pending, locked, or actively claimed.
+- Add QA seed guardrails that distinguish basic SVG placeholder media from the Swissifier rich PNG demo seed.
+- Add deterministic rich-seed verification and browser coverage to fail when QA is pointed at the wrong seed or invisible placeholder previews.
+
 ## 0.1.6
 
 - Add per-image re-roll attempt history with stack inspection, previous-attempt selection, and promotion back to the current/top attempt.

package/README.md

@@ -60,6 +60,20 @@ Agents can inspect the current lineage graph without a claim:
 lineage agent graph --project demo-project --root <root-asset-id> --json
 ```
 
+Agents can also export the current active workspace selection as a durable JSON
+packet for downstream tools such as GrowthOps:
+
+```bash
+lineage selection packet --project demo-project --channel linkedin --campaign 2026-07-launch --out ./lineage-selection-packet.json --json
+```
+
+The packet is schema-versioned as `lineage.selection_packet.v1` and includes the
+workspace/root binding, selected asset IDs, local media paths when known, S3 key
+metadata when known, context labels/notes, warnings, and a stable `packet_id`.
+Lineage only exports the packet. GrowthOps or another downstream tool remains
+responsible for importing it, creating posts, checking media readiness, preparing
+public URLs, scheduling, and recording placement receipts.
+
 Keep the claim fresh and pass it to mutating commands:
 
 ```bash
@@ -68,6 +82,9 @@ lineage link-child --project demo-project --root <root-asset-id> --child <child-
 lineage agent release --claim-token "$LINEAGE_CLAIM_TOKEN" --json
 ```
 
+`lineage link-child` creates a new visible descendant in the lineage graph. Do
+not use it for re-rolls.
+
 For re-roll work, mark the target, have the agent plan/import one replacement
 attempt, and cancel only when the request should be abandoned:
 
@@ -79,6 +96,23 @@ lineage reroll import --project demo-project --job-id <job-id> --file <.asset-sc
 lineage reroll cancel --project demo-project --root <root-asset-id> --target <target-asset-id> --confirm-write --json
 ```
 
+`lineage reroll import` updates the target node's current attempt and should
+not add a visible child edge.
+
+Lineage task queue commands expose the same work as explicit tasks that agents
+can inspect, claim, start, comment on, and cancel:
+
+```bash
+lineage tasks list --project demo-project --root <root-asset-id> --json
+lineage tasks inspect --project demo-project --task <task-id> --json
+lineage tasks claim --project demo-project --task <task-id> --agent-name "Task worker" --json
+lineage tasks start --project demo-project --task <task-id> --claim-token "$LINEAGE_CLAIM_TOKEN" --json
+lineage tasks comment --project demo-project --task <task-id> --message "Blocked on source art review." --json
+lineage tasks cancel --project demo-project --task <task-id> --confirm-write --override --json
+lineage tasks instructions --project demo-project --task <task-id> --instructions "Preserve palette; replace unreadable text." --json
+lineage tasks override --project demo-project --task <task-id> --reason "Human is reassigning the work." --instructions "Use the updated brief." --json
+```
+
 The app-created claim-aware handoff packet includes the same token export,
 heartbeat, inspect, and write commands. Raw claim tokens are not shown in the
 read-only Agents view.