ur-agent 1.78.7 → 1.78.9

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/docs/USAGE.md CHANGED
@@ -379,6 +379,57 @@ See [Frontier Agent Workflows](FRONTIER_AGENT_FEATURES.md) for the complete
379
379
  managed-worker, steering, learned-playbook, cited-memory, Agentic CI,
380
380
  trajectory, desktop-QA, side-chat, multi-repository, and arena trust model.
381
381
 
382
+ ## Large prompts, parallel work, and compact reasoning
383
+
384
+ UR turns a multi-part prompt into a bounded task graph instead of guessing
385
+ future task IDs. The normal target is 2–8 concrete tasks (hard limit 12): tasks
386
+ are created first, real IDs are captured, and dependencies are then connected
387
+ in dependency order. `ur exec` materializes this graph deterministically; the
388
+ interactive agent follows the same lifecycle and its task/agent tools enforce
389
+ the concurrency boundary.
390
+
391
+ Every real user work turn is seeded into the live task board immediately. If
392
+ the model creates explicit subtasks, its first `TaskCreate` atomically replaces
393
+ that seed instead of duplicating it. A successful simple turn completes the
394
+ seed automatically. If the user interrupts, the active seed remains pending;
395
+ pending and in-progress tasks stay visible, the new instruction is added to
396
+ the same board, and the agent reconciles the list rather than starting empty.
397
+ Only a list with no unfinished work is archived into readable history.
398
+
399
+ Independent read-only tasks can run in parallel. A task that may write to the
400
+ shared checkout is serialized with other possible writers, even when it comes
401
+ from another top-level prompt or crew worker. Parallel writers require explicit
402
+ worktree isolation. Agent worktrees use the exact current commit and require a
403
+ clean source checkout, so uncommitted work is never silently omitted or based
404
+ on an unrelated remote revision.
405
+
406
+ While the agent works, the `◭ Mashoofing…` row stays visible and its
407
+ parenthesized phase changes with activity: `thinking`, `requesting`,
408
+ `responding`, `preparing tool`, or `working`. Elapsed thinking time and token
409
+ activity continue to appear when available.
410
+
411
+ If a model sends a long internal planning monologue as ordinary answer text,
412
+ UR leaves the useful summary visible and replaces the noisy middle with a slim
413
+ `Reasoning condensed` rail. Press `ctrl+o` to expand the full trace. The
414
+ compaction is presentation-only: session history and the model's next-turn
415
+ context retain the original text. Detection is cached per completed message
416
+ and uses one linear pass; it does not make another model/API call or reduce the
417
+ agent's reasoning budget. During generation, UR buffers the undecided trailing
418
+ paragraph instead of briefly painting it and hiding it later. If that paragraph
419
+ is internal planning, the live preview advances to one stable `Reasoning in
420
+ progress` rail; it never exposes the raw self-talk on the normal screen.
421
+ Provider-native thinking follows the same rule and is visible in full only when
422
+ the transcript is explicitly opened or verbose diagnostics are enabled. The
423
+ persistent `Mashoofing` row continues to report live work on the normal screen.
424
+
425
+ When a provider reports that the context limit was reached despite the normal
426
+ proactive threshold, UR withholds that transient error, runs one emergency
427
+ compaction, and retries the interrupted turn automatically. Oversized images or
428
+ documents are replaced with markers only in the emergency summary request so a
429
+ large Computer screenshot cannot block recovery; the original transcript is
430
+ not rewritten. If that single bounded recovery fails—or automatic compaction
431
+ was explicitly disabled—UR shows the manual `/compact` or `/clear` action.
432
+
382
433
  ## Status bar
383
434
 
384
435
  Interactive sessions include a compact bottom status bar when stdout is a real
@@ -19,7 +19,7 @@ You need:
19
19
 
20
20
  ```sh
21
21
  ur --version
22
- # expected for this release: "1.78.7 (UR-Nexus)"
22
+ # expected for this release: "1.78.9 (UR-Nexus)"
23
23
  ```
24
24
 
25
25
  ## 0.1 First-workspace model selection (1.45.4)
@@ -45,7 +45,7 @@
45
45
  <main id="content" class="content">
46
46
  <header class="topbar">
47
47
  <div>
48
- <p class="eyebrow">Version 1.78.7</p>
48
+ <p class="eyebrow">Version 1.78.9</p>
49
49
  <h1>UR-Nexus Documentation</h1>
50
50
  <p class="lead">A practical, tutorial-style reference for installing, configuring, automating, extending, and operating UR-Nexus.</p>
51
51
  </div>
@@ -7,7 +7,7 @@ plugins {
7
7
  }
8
8
 
9
9
  group = "dev.urnexus"
10
- version = "1.78.7"
10
+ version = "1.78.9"
11
11
 
12
12
  repositories {
13
13
  mavenCentral()
@@ -2,7 +2,7 @@
2
2
  "name": "ur-inline-diffs",
3
3
  "displayName": "UR Inline Diffs",
4
4
  "description": "Review, apply, and reject UR inline diff bundles from .ur/ide/diffs inside VS Code.",
5
- "version": "1.78.7",
5
+ "version": "1.78.9",
6
6
  "publisher": "ur-nexus",
7
7
  "engines": {
8
8
  "vscode": "^1.92.0"
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "ur-agent",
3
- "version": "1.78.7",
3
+ "version": "1.78.9",
4
4
  "description": "UR-Nexus — autonomous engineering workflow engine (plan, execute, test, verify, document, benchmark, reproduce)",
5
5
  "type": "module",
6
6
  "packageManager": "bun@1.3.14",
@@ -49,6 +49,7 @@
49
49
  "bundle": "node scripts/bundle.mjs",
50
50
  "build": "node scripts/bundle.mjs",
51
51
  "release:check": "node scripts/release-check.mjs",
52
+ "release:tag": "node scripts/release-tag.mjs",
52
53
  "release:create-source-zip": "bun scripts/source-zip-create.mjs",
53
54
  "release:check-source-zip": "node scripts/source-zip-check.mjs",
54
55
  "typecheck": "bun run tsc --noEmit && bun run typecheck:strict-core",