pi-orchestrator-config 4.7.2 → 4.8.0

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.
@@ -1,7 +1,7 @@
1
1
  {
2
2
  "name": "pidiff-ui",
3
3
  "private": true,
4
- "version": "4.7.2",
4
+ "version": "4.8.0",
5
5
  "type": "module",
6
6
  "scripts": {
7
7
  "dev": "vite",
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "pi-orchestrator-config",
3
- "version": "4.7.2",
3
+ "version": "4.8.0",
4
4
  "description": "Orchestrator pattern for pi: specialist subagents, enforcement, code review loop, workflow commands",
5
5
  "keywords": [
6
6
  "pi-package"
@@ -80,6 +80,7 @@
80
80
  "@earendil-works/pi-coding-agent": "^0.99.1",
81
81
  "@earendil-works/pi-tui": "^0.99.1",
82
82
  "@types/node": "^22.20.1",
83
+ "jsdom": "^30.1.1",
83
84
  "tsx": "^4.0.0"
84
85
  },
85
86
  "scripts": {
@@ -504,6 +504,16 @@ Qodo if autoqodo, both if both flags active).
504
504
 
505
505
  #### 9a+9b: Wait and Fetch (combined async)
506
506
 
507
+ > **NEVER set `fireAndForget: true` on these workers.** The poll's entire purpose is
508
+ > to hand you its output. A fireAndForget agent is silent — it finishes and you are
509
+ > never told, so the loop stalls until the user comes back and tells you it was done.
510
+ > Leave `fireAndForget` unset. Also pass a `taskId` (`-1` if unlinked); a call without
511
+ > one is rejected outright.
512
+ >
513
+ > **After spawning, end your turn.** Do not poll, sleep, or re-spawn. The result arrives
514
+ > as a follow-up message. If you find yourself "checking" on the worker, you have
515
+ > already lost the result.
516
+
507
517
  **Always pass `--source` explicitly. No defaults.**
508
518
 
509
519
  **If autorabbit is ON (only):** Spawn ONE async worker:
@@ -537,7 +547,10 @@ When EITHER returns with new comments, process them (Phases 2-8). Then re-spawn
537
547
  The other agent keeps running independently.
538
548
 
539
549
  **While waiting for the async result**, the session remains interactive — the user
540
- can continue working. When the result surfaces, process it:
550
+ can continue working. You do NOT need to track it: the result surfaces on its own as a
551
+ follow-up message, and it is addressed to you, not to the user. Do not tell the user the
552
+ loop is running and then drop it; do not report "still waiting" as if that were an
553
+ update. When the result surfaces, process it:
541
554
 
542
555
  Check the poll JSON output — it always contains an `approved` key:
543
556
 
@@ -14,6 +14,7 @@
14
14
  | Kubernetes/OpenShift | `kubernetes-expert` |
15
15
  | Jenkins/CI/Groovy | `jenkins-expert` |
16
16
  | Git operations (local) | `git-expert` |
17
+ | Git merge/rebase/cherry-pick conflicts | `conflict-resolver` (git-expert is blocked from resolving them) |
17
18
  | GitHub (PRs, issues, releases, workflows) | `github-expert` |
18
19
  | Running existing tests / analyzing failures | `test-runner` |
19
20
  | Creating or modifying tests / test strategy | `test-automator` |
@@ -32,6 +33,7 @@ Route based on the task intent, not just the tool being used.
32
33
  - Creating or changing Python tests? → `test-automator`
33
34
  - Editing Python production files? → `python-expert` (even with sed/awk)
34
35
  - Creating a PR? → `github-expert` (not git-expert)
36
+ - A merge/rebase/cherry-pick left unmerged paths? → `conflict-resolver` (git-expert is mechanically blocked from resolving them; report the conflict and delegate)
35
37
  - External library docs? → `docs-fetcher` (not direct fetch)
36
38
 
37
39
  ## Documentation Routing (MANDATORY)
@@ -188,7 +188,12 @@ uv run myk-pi-tools memory <command> # Commands: add, forget, show, migrate, p
188
188
 
189
189
  Background async agent that reads the session, extracts things worth remembering, adds/deduplicates/reorganizes topic files, and removes stale entries. **NEVER removes Pinned entries.**
190
190
  Triggered by `/dream` (manual) or session shutdown (automatic).
191
- **ALWAYS run as async + fireAndForget** — never block the session.
191
+ **Dreaming ONLY: run as async + fireAndForget** — never block the session. Dream's
192
+ output is not consumed by anything, so it must stay silent.
193
+ **This is the sole fireAndForget agent in pi-config.** Every other async agent
194
+ returns a result you act on and must leave `fireAndForget` unset — a silent agent
195
+ cannot tell you anything. Never copy this pattern to a watcher, poll, or review
196
+ agent.
192
197
 
193
198
  **Dreaming follows the same NEVER memorize rules as per-turn self-improvement:**
194
199
  do NOT extract rejected/abandoned ideas, unfinished conversations, or unimplemented requirements.
@@ -28,7 +28,14 @@ Before every response, identify operations that can run in parallel and execute
28
28
  Use `async: true` for independent tasks (reviews, research, polling, monitoring, CI checks) — any task where you don't need the result immediately.
29
29
  Only use sync when the very next step depends on this agent's output.
30
30
  After spawning async agents, end your turn — results arrive automatically as a follow-up message.
31
- (`fireAndForget: true` agents are silent — no follow-up.)
31
+
32
+ **`fireAndForget: true` delivers nothing.** The agent runs silently and you never see
33
+ its output. Use it ONLY for work whose result nobody consumes — in pi-config, that is
34
+ dreaming and nothing else. Every watcher, poll, and review agent exists to report
35
+ something back, so leave `fireAndForget` unset for them. If you are about to set it,
36
+ ask whether you will ever read the result; if yes, do not set it.
37
+ **`taskId` is required on every async call** (use `-1` when unlinked) — a call without
38
+ it is rejected, and the agent does not run.
32
39
 
33
40
  Do NOT write bash loops, sleep commands, or poll for results after spawning async agents.
34
41
 
@@ -26,6 +26,7 @@ This rule applies ONLY to agents defined in this repository (`agents/` directory
26
26
  - docker-expert
27
27
  - ts-expert
28
28
  - git-expert
29
+ - conflict-resolver
29
30
  - github-expert
30
31
  - go-expert
31
32
  - java-expert
@@ -158,15 +158,18 @@ uv run pytest -q # or npm test, go test, etc.
158
158
  ## Phase 5: Complete
159
159
 
160
160
  ```bash
161
- # For merge
161
+ # For merge: stage, then report for git-expert to commit
162
162
  git add <all-resolved-files>
163
- git commit # uses merge commit message
164
163
 
165
- # For rebase
164
+ # For rebase: stage, then report for git-expert to continue
166
165
  git add <all-resolved-files>
167
- git rebase --continue
168
166
  ```
169
167
 
168
+ Stage the resolved files, then stop. **Committing is `git-expert`'s job** — a
169
+ `git commit` here is rejected by the harness, so following the old sequence
170
+ leaves the operation unfinished. Report what you resolved and that it is staged,
171
+ and let the caller delegate the commit or `rebase --continue` to `git-expert`.
172
+
170
173
  ## Common Patterns
171
174
 
172
175
  ### Import Conflicts