omnius 1.0.774 → 1.0.776

package/dist/index.js

@@ -1,4 +1,4 @@
 [diffend] Oversized file quarantined before diffing.
 name: package/dist/index.js
-size: 38971888 bytes
-sha256: ac82236ef4f2bad8f1b8b3beac4dc64c1774bdc12f827c72714db4341bce408a
+size: 39016717 bytes
+sha256: c7caf242ceb7df7ccd5327413f4c3103dfcec2a84a84ba81c6eb6477e8f80db9

package/docs/DISCOVERY.json

@@ -35901,6 +35901,42 @@
         }
       ]
     },
+    {
+      "id": "guide.work-orders-telegram-reliability-consolidation-2026-09-28",
+      "kind": "guide",
+      "title": "Telegram Reliability Consolidation — 2026-09-28",
+      "summary": "Rule: check a box only after the named behavior is implemented and its listed verification passes. A parent box stays open until every child box is checked.",
+      "keywords": [
+        "work",
+        "orders",
+        "telegram",
+        "reliability",
+        "consolidation",
+        "2026",
+        "09",
+        "28",
+        "md"
+      ],
+      "maturity": "internal",
+      "audiences": [
+        "maintainer",
+        "large-context-agent"
+      ],
+      "layer": "documentation",
+      "interfaces": [
+        {
+          "type": "file",
+          "target": "docs/work-orders/telegram-reliability-consolidation-2026-09-28.md"
+        }
+      ],
+      "references": [
+        {
+          "type": "documentation",
+          "target": "docs/work-orders/telegram-reliability-consolidation-2026-09-28.md",
+          "relation": "canonical-artifact"
+        }
+      ]
+    },
     {
       "id": "guide.work-orders-wo-am-gaps-uppercase",
       "kind": "guide",

package/docs/DISCOVERY.md

@@ -610,6 +610,7 @@ Daemon equivalents are `GET /v1/discovery/bootstrap`, `GET /v1/discovery?q=<inte
 | `guide.work-orders-runtime-health-remediation-wo-56-evidence-driven-objective-loop-uppercase` | WO-56: Evidence-driven objective loop and nonblocking execution | Status: repository implementation complete; live Telegram acceptance remains an external gate Priority: P0 Incident: September 27, 2026, Telegram SSH-over-Cloudflare task spiral and TUI freeze Owner: Omnius orchestration, execution, and Telegram runtime Depends on: WO-05, WO-21, WO-27, WO-41, WO-44, WO-49, WO-54, WO-55 Completion rule: A checkbox may be mark |
 | `guide.work-orders-telegram-dmn-wo-22-dmn-outreach-and-learning-uppercase` | WO-22: DMN outreach, DM sharing, and outcome learning | On 2026-09-03 at 17:07 PDT the bot posted in the OMNIUS group without being addressed. The operator asked whether this was self-induced reflection. |
 | `guide.work-orders-telegram-dropbear-context-rca-workorder` | Telegram Dropbear Context Engineering RCA Work Order | Observed run: /home/roko/Documents/Projects/Adjacent/telegramtest/.omnius, run id 1782873796963-i5r7mv. |
+| `guide.work-orders-telegram-reliability-consolidation-2026-09-28` | Telegram Reliability Consolidation — 2026-09-28 | Rule: check a box only after the named behavior is implemented and its listed verification passes. A parent box stays open until every child box is checked. |
 | `guide.work-orders-wo-am-gaps-uppercase` | Associative Memory Gap Work Orders | Generated: 2026-04-13 Source: Deep audit of multimodal associative memory systems Status: READY FOR IMPLEMENTATION |
 | `guide.work-orders-world-class-memory-compiler-readme-uppercase` | World-Class Memory Compiler Program | Status: active implementation program Owner: Omnius orchestration and memory packages Last updated: 2026-07-13 |
 | `guide.work-orders-world-class-memory-compiler-tracker-uppercase` | Memory Compiler Implementation Tracker | Last reconciled: 2026-07-13 Authority: This file is the granular program tracker. The work-order documents define the contract and acceptance criteria; neither test output nor a model claim may check an item without the named evidence. |

package/docs/work-orders/telegram-reliability-consolidation-2026-09-28.md

@@ -0,0 +1,178 @@
+# Telegram Reliability Consolidation — 2026-09-28
+
+Status: complete
+
+Rule: check a box only after the named behavior is implemented and its listed
+verification passes. A parent box stays open until every child box is checked.
+
+## Failure classification
+
+- [x] **R1 — Response identity split (correctness, P0).** A runner can retain
+  one input/task identity while visible reply ownership moves to a later input.
+  - [x] Define one response binding containing `inputId`, `messageId`,
+    `decisionId`, `taskEpoch`, `runId`, generation, reply target, and route.
+  - [x] Use that binding for social-decision commit.
+  - [x] Use that binding for visible reply ownership.
+  - [x] Use that binding for final Telegram reply targeting.
+  - [x] Use that binding for delivery-ledger attribution.
+  - [x] Reject delivery when any binding identity differs.
+  - [x] Prevent a materially new request from inheriting an older objective.
+  - [x] Cover the message-496/message-499 production shape with a regression
+    test.
+
+- [x] **R2 — Active-runner authority bypass (correctness/safety, P0).** Public
+  group steering can bypass the group reply-authority review used by new work.
+  - [x] Apply equivalent input-bound reply-authority validation to active-runner
+    routing.
+  - [x] Fail closed when active-runner authority is unavailable or invalid.
+  - [x] Test a human-to-human active-runner turn that must remain silent.
+  - [x] Test a direct active-runner address that remains reply-capable.
+
+- [x] **R3 — Conversational admin DM misclassified as mutation (correctness,
+  P0).** Metaphorical or conversational reset language enters the full action
+  pipeline and creates an unfulfillable mutation obligation.
+  - [x] Add a conversational/internal-control execution class before action
+    planning.
+  - [x] Ensure metaphorical reset/acknowledgement requests do not require a
+    mutation effect.
+  - [x] Avoid engineering workboard creation for a simple acknowledgement.
+  - [x] Avoid `tool_search`/`todo_write` for the acknowledgement fast path.
+  - [x] Preserve the natural assistant answer as the delivered reply.
+  - [x] Test `clear your slate ... mentally and proverbially` end to end.
+
+- [x] **R4 — Completion verifier replaces valid replies with bookkeeping
+  (correctness/UX, P0).** A host-side contract mismatch can discard a current,
+  evidence-bound `task_complete.user_reply`.
+  - [x] Separate conversational reply delivery from action verification status.
+  - [x] Keep a valid current user reply deliverable when only the generated
+    expected-effect contract is unsatisfied.
+  - [x] Keep genuine mutation failures fail-closed for completion claims.
+  - [x] Put verification gaps in diagnostics/evidence rather than substituting
+    them for a valid conversational reply.
+  - [x] Replace tests that encode the undesirable canned admin response.
+
+- [x] **R5 — Admin control-plane leakage (UX, P1).** Steering reconciliation is
+  rendered as `Plan update` on the primary admin response surface.
+  - [x] Remove reconciliation blocks from the admin response page.
+  - [x] Retain reconciliation evidence under Tasks/Logs/Context diagnostics.
+  - [x] Ensure final admin reply composition does not prepend plan telemetry.
+  - [x] Test that ordinary admin replies contain no `Plan update` block.
+
+- [x] **R6 — Serial inference multiplication (latency, P1).** Router,
+  authority, generation, semantic review, and repair retries stack on the
+  critical path.
+  - [x] Split minimal social authority/route output from action planning fields.
+  - [x] Run action planning only for an accepted action route.
+  - [x] Bound malformed-router repair to one compact repair attempt.
+  - [x] Avoid an unconditional semantic review for low-risk conversational chat.
+  - [x] Run the second-pass public authority review only for positive group
+    candidates; do not spend that call on already-silent decisions.
+  - [x] Add an operation-level routing deadline that includes broker admission
+    and retries.
+  - [x] Test bounded inference-call counts for no-reply, private chat, group
+    chat, and action routes.
+
+- [x] **R7 — Oversized context for trivial turns (latency/cost, P1).** A simple
+  acknowledgement consumed five turns and 110,433 cumulative tokens.
+  - [x] Define a strict context budget for conversational/admin fast paths.
+  - [x] Exclude workspace tree, work graph, mutation receipts, and engineering
+    contracts when the execution class does not need them.
+  - [x] Stop repeating no-plan reminders on conversational turns.
+  - [x] Add a regression ceiling for prompt size and runner turns.
+
+- [x] **R8 — Durable spool replay churn (durability/performance, P0).** A raw
+  update remains reclaimable after it has been durably transferred to intake,
+  producing repeated dispatch attempts.
+  - [x] Define the transport-terminal state for an update durably bound to
+    intake.
+  - [x] Mark the raw spool row transport-terminal atomically when the first
+    durable intake lifecycle record commits.
+  - [x] Let the intake lifecycle, not raw update replay, own deferred work.
+  - [x] Preserve raw-envelope restart recovery before the durable intake
+    handoff commits.
+  - [x] Rehydrate nonterminal logical intake after restart without reclaiming
+    the raw transport envelope.
+  - [x] Recover the newest edited-message revision while preserving the
+    original logical intake owner.
+  - [x] Test that deferred active-runner intake is not replay-claimed.
+  - [x] Test production restart recovery through the real replay entry point,
+    with one logical scheduling and one raw dispatch attempt.
+
+- [x] **R9 — Head-of-line route blocking (latency, P1).** One global active
+  route slot lets a slow/retried chat delay unrelated chats.
+  - [x] Replace the global single route slot with bounded fair concurrency.
+  - [x] Preserve per-chat inference serialization and exact input binding.
+  - [x] Prioritize direct messages, direct mentions, and known reply-capable
+    chats.
+  - [x] Test that a stalled route does not block an unrelated direct address.
+
+- [x] **R10 — Terminal review duplication (latency, P1).** Identical unchanged
+  evidence can receive repeated inline reviews and recovery cycles.
+  - [x] Key terminal review results by input, objective-frontier hash, and
+    evidence digest.
+  - [x] Reuse a review when its evidence identity is unchanged.
+  - [x] Do not immediately retry an identical malformed review inline.
+  - [x] Keep simple conversation replies out of action recovery; reuse an
+    unchanged evidence review for read-only research instead of reviewing it
+    again at each stalled/terminal phase.
+  - [x] Test bounded review/recovery counts.
+
+- [x] **R11 — Normal-mode event noise (observability, P2).** Sequential router
+  and reply-phase cards obscure the user-visible critical path.
+  - [x] Bind every inference phase to the exact input and reuse one TUI block
+    for sequential phases of that input only.
+  - [x] Label group inference with the group name rather than a generic title.
+  - [x] Collapse completed control-plane JSON while keeping it available in
+    the block's explicit read mode.
+  - [x] Keep periodic queue diagnostics behind an explicit debug flag.
+  - [x] Preserve the underlying segments and durable forensic evidence.
+  - [x] Test that normal mode omits raw completed router JSON and repeated
+    queue chatter.
+
+- [x] **R12 — Todo-state path override ignored (testability/configuration,
+  P2).** The todo store wrote only to `~/.omnius`, so restricted environments
+  reported a false lock timeout instead of exercising Telegram context tests.
+  - [x] Make the todo store honor the existing `OMNIUS_HOME` state-root
+    override used by the other execution stores.
+  - [x] Make its disk-layout assertion derive the path from `todoDir()`.
+  - [x] Run the todo-store and task-completion guard suites under an isolated
+    `OMNIUS_HOME` (42/42 passing).
+
+## Verification and delivery
+
+- [x] Run focused Telegram routing and authority tests.
+- [x] Run active-runner steering tests.
+- [x] Run admin live-panel and terminal-delivery tests.
+- [x] Run completion-readiness/orchestrator tests.
+- [x] Run durable update spool and restart-replay tests.
+- [x] Run Telegram queue/priority tests.
+- [x] Run the complete Telegram CLI test subset affected by these changes.
+- [x] Build every changed workspace package.
+- [x] Inspect the diff for unrelated files and generated artifacts.
+- [x] Commit each coherent verified repair with scoped paths only.
+- [x] Fetch the tracked branch and push without rewriting shared history.
+- [x] Record commit hashes, branch, and verification results here.
+
+## Evidence log
+
+- Live admin-DM run:
+  `telegram-64ac9937dc7647a8-1790575905415-1`.
+- Live Case Theory mismatch: source task message 496, transferred visible reply
+  owner message 499, delivered message 500.
+- No completion boxes were pre-checked when this work order was created.
+- Focused Telegram routing/authority/UI verification: 106/106 passing.
+- Integrated affected Telegram CLI verification: 426/426 passing across 21
+  files, including `telegram-bot-api-10.test.ts` at 155/155.
+- Orchestrator completion authority verification: 30/30 passing.
+- Execution todo-store override and completion-guard verification: 42/42
+  passing under an isolated `OMNIUS_HOME`.
+- Workspace builds: `@omnius/execution`, `@omnius/orchestrator`, and `omnius`
+  all passing.
+- `git diff --check`: passing. Unrelated pre-existing changes in
+  `docs/DISCOVERY.json`, `docs/DISCOVERY.md`, `publish/npm-shrinkwrap.json`,
+  and `publish/package.json` are intentionally excluded from delivery.
+- Scoped delivery commits: `21134195` (todo-state override), `d489e4ca`
+  (Telegram/orchestrator reliability), and `8d594887` (verification record).
+- Fetched `origin/main` with zero remote-only commits, then pushed without
+  rewriting history. Remote `main` was verified at
+  `8d594887e87364ecf3e6d7475ca1cda24cde9852`.

package/npm-shrinkwrap.json

@@ -1,12 +1,12 @@
 {
   "name": "omnius",
-  "version": "1.0.774",
+  "version": "1.0.776",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "omnius",
-      "version": "1.0.774",
+      "version": "1.0.776",
       "bundleDependencies": [
         "image-to-ascii"
       ],

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "omnius",
-  "version": "1.0.774",
+  "version": "1.0.776",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/library.js",