omnius 1.0.693 → 1.0.694

package/dist/index.js

@@ -662702,6 +662702,15 @@ function classifyThinkOutcome(raw) {
   }
   return null;
 }
+function logicalRequestVariantId(logicalRequestId, variant) {
+  return `ollama:${variant}:${_createHash("sha256").update(`${logicalRequestId}\0${variant}`).digest("hex").slice(0, 24)}`;
+}
+function withLogicalRequestVariant(request, variant) {
+  return request.logicalRequestId ? {
+    ...request,
+    logicalRequestId: logicalRequestVariantId(request.logicalRequestId, variant)
+  } : request;
+}
 var PRESERVE_MODEL_VISIBLE_HISTORY, TOOL_SUBSETS, TOOL_AUTO_DEMOTE_TURNS, LEGACY_ACTION_REASON_KEYS, CLAIM_TASK_STOPWORDS, TASK_COMPLETE_CLOSEOUT_DESCRIPTION, SYSTEM_PROMPT_COMMON, SYSTEM_PROMPT, SYSTEM_PROMPT_MEDIUM, SYSTEM_PROMPT_SMALL, VISUAL_TOOLS, AUDIO_TOOLS, SOCIAL_TOOLS, SPATIAL_TOOLS, CODE_TOOLS, AgenticRunner, BROKER_QUEUE_DEADLINE_ERROR, InferenceHttpError, OllamaAgenticBackend;
 var init_agenticRunner = __esm({
   "packages/orchestrator/dist/agenticRunner.js"() {
@@ -692604,7 +692613,8 @@ arguments=` : ""}` + (chunk.toolCallArgs ?? "");
           const shouldRetryThinkGuard = outcome !== null && effectiveThink === true && (justSuppressed || outcome === "empty_after_strip" || outcome === "unclosed_think");
           if (shouldProbeNativeOnEmpty) {
             try {
-              const _native = await this.nativeOllamaChatCompletion(request, {
+              const nativeRecoveryRequest = withLogicalRequestVariant(request, "native-empty-recovery");
+              const _native = await this.nativeOllamaChatCompletion(nativeRecoveryRequest, {
                 // Keep recovery on the pool endpoint that already admitted and
                 // warmed this request. Falling back to the broker/base URL can
                 // cold-load a second copy of a large model while the admitted
@@ -692647,13 +692657,8 @@ arguments=` : ""}` + (chunk.toolCallArgs ?? "");
               retryBody["num_ctx"] = reqNumCtx;
             }
             try {
-              const retryOpts = {
-                method: "POST",
-                headers: this.authHeaders(request),
-                body: JSON.stringify(retryBody)
-              };
-              if (combinedAbortSignal)
-                retryOpts.signal = combinedAbortSignal;
+              const openAiRetryRequest = withLogicalRequestVariant(request, "openai-empty-retry");
+              const retryOpts = this.brokerRequestOptions(openAiRetryRequest, retryBody, combinedAbortSignal);
               const retryResp = await fetch(this.providerRequestUrl(requestBaseUrl, "chat"), retryOpts);
               if (retryResp.ok) {
                 const retryData = await retryResp.json();
@@ -801757,6 +801762,9 @@ function telegramStreamFailureAllowsUnaryFallback(err) {
     text3
   );
 }
+function telegramUnaryCompatibilityLogicalRequestId(streamLogicalRequestId) {
+  return `telegram:unary-compat:${createHash91("sha256").update(streamLogicalRequestId).digest("hex").slice(0, 24)}`;
+}
 function telegramRouterErrorText(err) {
   return err instanceof Error ? err.message : String(err);
 }
@@ -811130,7 +811138,14 @@ ${JSON.stringify(modelFollowupProposals, null, 2)}`,
               ],
               tools: [],
               temperature: 0.2,
-              maxTokens: 300,
+              // The contract permits up to 700 characters of message text plus
+              // evidence IDs and decision metadata. A 300-token ceiling can end
+              // at `finish=length` before a complete JSON object exists.
+              maxTokens: 1024,
+              // Ollama's OpenAI-compatible renderer can spend the full token
+              // budget in hidden reasoning even with think:false. Native chat
+              // enforces direct-answer mode and avoids an empty-stream fallback.
+              preferNativeOllamaChat: this.agentConfig.backendType === "ollama",
               timeoutMs: Math.min(
                 Math.max(this.agentConfig.timeoutMs ?? 3e4, 5e3),
                 2e4
@@ -813819,7 +813834,15 @@ ${this.quoteTelegramContextBlock(msg.text, 1200)}`
                     id2,
                     `stream protocol did not return a usable contract; using one unary compatibility request: ${streamErr instanceof Error ? streamErr.message : String(streamErr)}`
                   );
-                  result = await backend.chatCompletion(requestWithCtx);
+                  const unaryCompatibilityRequest = {
+                    ...requestWithCtx,
+                    logicalRequestId: telegramUnaryCompatibilityLogicalRequestId(
+                      logicalRequestId
+                    )
+                  };
+                  result = await backend.chatCompletion(
+                    unaryCompatibilityRequest
+                  );
                   this.updateTelegramInferenceFinal(id2, result);
                   this.tuiWrite(
                     () => renderTelegramSubAgentEvent(

package/docs/DISCOVERY.json

@@ -32842,7 +32842,7 @@
       "id": "guide.work-orders-runtime-health-remediation-tracker-uppercase",
       "kind": "guide",
       "title": "Runtime Health Remediation Master Tracker",
-      "summary": "Authority: canonical granular tracker for RHR-2026-09-02 Checked-item rule: code, focused tests, and named evidence must all exist Last reconciled: 2026-09-02",
+      "summary": "Authority: canonical granular tracker for RHR-2026-09-02 Checked-item rule: code, focused tests, and named evidence must all exist Last reconciled: 2026-09-04",
       "keywords": [
         "work",
         "orders",
@@ -33942,6 +33942,46 @@
         }
       ]
     },
+    {
+      "id": "guide.work-orders-runtime-health-remediation-wo-24-telegram-logical-request-variant-identity-uppercase",
+      "kind": "guide",
+      "title": "WO-24: Telegram logical request variant identity",
+      "summary": "Telegram follow-up requests could finish a streamed inference with no visible contract and then fail their one unary compatibility request with HTTP 409: logicalrequestconflict. The observed stream used all 300 completion tokens and ended with finish=length.",
+      "keywords": [
+        "work",
+        "orders",
+        "runtime",
+        "health",
+        "remediation",
+        "WO",
+        "24",
+        "telegram",
+        "logical",
+        "request",
+        "variant",
+        "identity",
+        "md"
+      ],
+      "maturity": "internal",
+      "audiences": [
+        "maintainer",
+        "large-context-agent"
+      ],
+      "layer": "documentation",
+      "interfaces": [
+        {
+          "type": "file",
+          "target": "docs/work-orders/runtime-health-remediation/WO-24-telegram-logical-request-variant-identity.md"
+        }
+      ],
+      "references": [
+        {
+          "type": "documentation",
+          "target": "docs/work-orders/runtime-health-remediation/WO-24-telegram-logical-request-variant-identity.md",
+          "relation": "canonical-artifact"
+        }
+      ]
+    },
     {
       "id": "guide.work-orders-telegram-dmn-wo-22-dmn-outreach-and-learning-uppercase",
       "kind": "guide",

package/docs/DISCOVERY.md

@@ -533,7 +533,7 @@ Daemon equivalents are `GET /v1/discovery/bootstrap`, `GET /v1/discovery?q=<inte
 | `guide.work-orders-runtime-health-remediation-live-telegram-log-review-2026-09-03-uppercase` | Live Telegram Log Review, 2026-09-03 | This review is read-only. It examines the persisted Telegram conversation and intake lifecycle records in /home/roko/Documents/Projects/Adjacent/telegramtest/.omnius. The running global Omnius package reports version 1.0.687. This means the observations describe the deployed pre-publication runtime, not the newer source commits on this repository's main bran |
 | `guide.work-orders-runtime-health-remediation-readme-uppercase` | Runtime Health Remediation Program | Program ID: RHR-2026-09-02 Status: active, implementation authorized Owner: Omnius runtime, orchestration, execution, memory, and Telegram packages External dependency: /home/roko/Desktop/ollama-unify Last reconciled: 2026-09-03 |
 | `guide.work-orders-runtime-health-remediation-traceability-uppercase` | Runtime Health Remediation Traceability Matrix | Every completed row links to a deterministic test or operational receipt. A commit hash alone is not proof. A model statement is not proof. Live logs may support a canary only after the hermetic and deterministic gates pass. A status that names pending rollout or live work is intentionally not a completion claim. |
-| `guide.work-orders-runtime-health-remediation-tracker-uppercase` | Runtime Health Remediation Master Tracker | Authority: canonical granular tracker for RHR-2026-09-02 Checked-item rule: code, focused tests, and named evidence must all exist Last reconciled: 2026-09-02 |
+| `guide.work-orders-runtime-health-remediation-tracker-uppercase` | Runtime Health Remediation Master Tracker | Authority: canonical granular tracker for RHR-2026-09-02 Checked-item rule: code, focused tests, and named evidence must all exist Last reconciled: 2026-09-04 |
 | `guide.work-orders-runtime-health-remediation-wo-00-hermetic-test-boundary-uppercase` | WO-00: Hermetic Test Boundary | Status: deterministic acceptance complete Evidence: inference network inventory and audit receipt for commit 8708e449 Risk: high, because current unit tests can contact the live broker Depends on: none |
 | `guide.work-orders-runtime-health-remediation-wo-00-inference-network-inventory-uppercase` | WO-00 Inference Network Inventory | This document describes the machine-checked boundary for production inference clients. The canonical inventory is test-support/inference-network-registry.ts. |
 | `guide.work-orders-runtime-health-remediation-wo-01-inference-pressure-scheduler-uppercase` | WO-01: Foreground and Background Inference Pressure Scheduler | Status: deterministic acceptance complete Evidence: production Telegram cognition integration and controlled-load foreground exclusion Risk: critical, because background work currently competes with user work Depends on: WO-00 |
@@ -562,6 +562,7 @@ Daemon equivalents are `GET /v1/discovery/bootstrap`, `GET /v1/discovery?q=<inte
 | `guide.work-orders-runtime-health-remediation-wo-21-task-convergence-and-todo-scope-uppercase` | WO-21: Task convergence and todo scope | On 2026-09-03, a user reported that Omnius remained at task 2/8 for about 40 turns. The report did not include a screenshot or run artifact, so the exact label remains unverified. Omnius has two similar counters: |
 | `guide.work-orders-runtime-health-remediation-wo-22-autocomplete-footer-geometry-uppercase` | WO-22: Autocomplete Footer Geometry Integrity | Scope: CLI TUI footer, DirectInput redraw scheduling, active content writes |
 | `guide.work-orders-runtime-health-remediation-wo-23-long-horizon-control-plane-reconciliation-uppercase` | WO-23: Long-horizon control-plane reconciliation | The restored telegramtest run tui-3869250-mtlvp324-1788468819792-1 continued a large pentest-tooling frontend task across a process restart. It made real source mutations and advanced from the first implementation group to todo-p2a. The run was not dead and its model context had ample nominal capacity. |
+| `guide.work-orders-runtime-health-remediation-wo-24-telegram-logical-request-variant-identity-uppercase` | WO-24: Telegram logical request variant identity | Telegram follow-up requests could finish a streamed inference with no visible contract and then fail their one unary compatibility request with HTTP 409: logicalrequestconflict. The observed stream used all 300 completion tokens and ended with finish=length. |
 | `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-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 |

package/docs/work-orders/runtime-health-remediation/TRACKER.md

@@ -2,7 +2,7 @@
 
 **Authority:** canonical granular tracker for RHR-2026-09-02
 **Checked-item rule:** code, focused tests, and named evidence must all exist
-**Last reconciled:** 2026-09-02
+**Last reconciled:** 2026-09-04
 
 ## Phase 0: durable program boundary
 
@@ -271,6 +271,23 @@ deterministically repaired P0 or P1 defect.
   workspace build.
 - [x] Commit and push the scoped repair to `main`.
 
+## WO-24: Telegram logical request variant identity
+
+- [x] Correlate empty Telegram streams with broker conflict receipts.
+- [x] Confirm the broker correctly binds one logical ID to one exact wire
+  contract.
+- [x] Give stream-to-unary compatibility fallback a distinct stable ID.
+- [x] Give `/v1`-to-native empty recovery a distinct stable ID.
+- [x] Give changed `/v1` recovery bodies a distinct stable ID.
+- [x] Preserve the original ID across byte-identical queue retries.
+- [x] Route Ollama follow-up decisions through native chat with `think:false`.
+- [x] Raise the follow-up contract budget from 300 to 1,024 tokens.
+- [x] Add a broker-faithful 409 conflict regression harness.
+- [x] Pass backend, Telegram transport, observability, and Bot API suites.
+- [x] Pass affected package typechecks and the workspace build.
+- [ ] Run a token-generating live canary only after the mandatory accelerator
+  preflight identifies the requested endpoint, model, and GPU placement.
+
 ## Program completion gate
 
 - [x] All repository-controlled deterministic WO acceptance sections pass.

package/docs/work-orders/runtime-health-remediation/WO-24-telegram-logical-request-variant-identity.md

@@ -0,0 +1,65 @@
+# WO-24: Telegram logical request variant identity
+
+## Incident
+
+Telegram follow-up requests could finish a streamed inference with no visible
+contract and then fail their one unary compatibility request with HTTP 409:
+`logical_request_conflict`. The observed stream used all 300 completion tokens
+and ended with `finish=length`.
+
+The broker response was correct. Omnius reused one logical request ID after it
+changed the request body from streaming to unary. The same contract violation
+also existed in the backend's empty-output recovery path when it changed from
+OpenAI-compatible `/v1/chat/completions` to native `/api/chat`, or when it
+constructed a changed `/v1` retry body.
+
+## Root cause
+
+ollama-unify binds a logical request ID to the exact request method, path,
+content type, and body. It uses that binding for idempotent queue admission and
+completed-response replay. Omnius treated the ID as a broad correlation ID.
+It therefore reused the ID across different transport contracts. The broker
+rejected the second contract instead of replaying an unrelated response.
+
+The initial empty output had a second cause. Telegram follow-up inference used
+the OpenAI-compatible Ollama renderer without native transport preference. A
+Qwen-derived model can consume its entire budget in renderer-managed reasoning
+even when the caller requests `think:false`. The stream adapter then has no
+visible contract to parse.
+
+## Invariants
+
+- [x] One logical request ID identifies one exact wire contract.
+- [x] Byte-identical queue retries retain their original ID.
+- [x] A stream-to-unary transition derives a new stable ID.
+- [x] A `/v1`-to-native recovery transition derives a new stable ID.
+- [x] A changed retry body derives a new stable ID.
+- [x] A Telegram Ollama follow-up uses native chat and explicit
+  direct-answer mode.
+- [x] Follow-up output capacity fits the declared JSON contract.
+- [x] A deterministic test broker rejects any future cross-contract ID reuse.
+
+## Implementation
+
+- [x] Add `telegramUnaryCompatibilityLogicalRequestId`.
+- [x] Apply the derived ID to the unary compatibility request only.
+- [x] Add backend request-variant identity derivation.
+- [x] Apply a native-recovery identity before `/api/chat`.
+- [x] Apply a changed-retry identity before the second `/v1` request.
+- [x] Use the normal broker request builder for the changed retry.
+- [x] Set `preferNativeOllamaChat` for Ollama follow-up decisions.
+- [x] Raise follow-up `maxTokens` from 300 to 1,024.
+- [x] Document logical ID semantics on the backend request contract.
+
+## Verification
+
+- `pnpm --filter @omnius/orchestrator typecheck`: passed.
+- `pnpm --filter omnius typecheck`: passed.
+- `ollama-backend-pool.test.ts`: 27 tests passed.
+- `telegram-inference-contract.test.ts`: 10 tests passed.
+- `telegram-observability.test.ts`: 22 tests passed.
+- `telegram-bot-api-10.test.ts`: 135 tests passed.
+- The mocked broker stores the first exact body for every logical ID and emits
+  the production-style non-retryable 409 response if later content differs.
+- No token-generating live inference was used. A live canary remains subject to
+  the mandatory endpoint, model, and physical GPU placement preflight.

package/npm-shrinkwrap.json

@@ -1,12 +1,12 @@
 {
   "name": "omnius",
-  "version": "1.0.693",
+  "version": "1.0.694",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "omnius",
-      "version": "1.0.693",
+      "version": "1.0.694",
       "bundleDependencies": [
         "image-to-ascii"
       ],
@@ -4354,9 +4354,9 @@
       }
     },
     "node_modules/hono": {
-      "version": "4.13.5",
-      "resolved": "https://registry.npmjs.org/hono/-/hono-4.13.5.tgz",
-      "integrity": "sha512-O6+/eCYRkzzzy0rPWwKLiGBR1nFuUPZynnwjxN1MBA62NNqbT0wQEzQyK2gSO5yDIDB336sXQleAhOHrzlYyKw==",
+      "version": "4.13.7",
+      "resolved": "https://registry.npmjs.org/hono/-/hono-4.13.7.tgz",
+      "integrity": "sha512-c8/gF9ac8Y78/agExVocyLevgR+JlpNB444Py0FSX8pJoPdYUfUzRcXtYEYGwt6l19qIlVZPN5Mfsw9jFShmQQ==",
       "license": "MIT",
       "engines": {
         "node": ">=16.9.0"

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "omnius",
-  "version": "1.0.693",
+  "version": "1.0.694",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/library.js",