@jsonstudio/rcc 0.89.1960 → 0.89.2195

package/docs/daemon-admin-ui.html

@@ -881,6 +881,35 @@
                   <tbody id="controlQuotaProvidersTbody"></tbody>
                 </table>
               </div>
+
+              <p class="section-title" style="margin-top: 14px;">ServerTool runtime</p>
+              <p class="section-sub">
+                Toggle servertool orchestration on/off instantly for this server process.
+              </p>
+              <div class="row" style="margin-bottom: 10px;">
+                <span id="controlServerToolStatus" class="pill">state: —</span>
+                <span id="controlServerToolUpdatedAt" class="mono muted">updated: —</span>
+                <button id="controlServerToolToggleBtn" class="primary">Toggle</button>
+              </div>
+              <div class="notice mono" id="controlServerToolCountsBox" style="white-space: pre-wrap; margin-bottom: 10px;">
+                executions: —
+                success: —
+                failure: —
+              </div>
+              <div class="table-wrap">
+                <table class="table">
+                  <thead>
+                    <tr>
+                      <th>tool</th>
+                      <th>executions</th>
+                      <th>success</th>
+                      <th>failure</th>
+                      <th>details</th>
+                    </tr>
+                  </thead>
+                  <tbody id="controlServerToolToolsTbody"></tbody>
+                </table>
+              </div>
             </div>
           </div>
         </section>
@@ -1453,7 +1482,7 @@
 	            <div class="card" style="box-shadow:none;">
 	              <p class="section-title">Routing editor</p>
 	              <p class="section-sub">
-	                Edits routing in a selected config file (auto-detects <span class="mono">virtualrouter.routing</span> or <span class="mono">routing</span>).
+	                Edits routing policy in a selected config file (auto-detects <span class="mono">virtualrouter.routing + virtualrouter.loadBalancing</span> or <span class="mono">routing + loadBalancing</span>).
 	              </p>
 	              <div class="row" style="margin-bottom: 10px;">
 	                <label for="routingSourceSelect">source</label>
@@ -1481,6 +1510,8 @@
 	              <div class="row" style="margin-bottom: 10px;">
 	                <button id="loadRoutingBtn" class="primary">Load</button>
 	                <button id="saveRoutingBtn" class="primary">Save</button>
+                  <button id="saveRoutingRestartLocalBtn">Save + Restart local</button>
+                  <button id="saveRoutingRestartAllBtn">Save + Restart all</button>
                   <button id="refreshRoutingPoolBtn">Refresh pool status</button>
 	                <button id="routingRegExpandBtn">Expand all</button>
 	                <button id="routingRegCollapseBtn">Collapse all</button>
@@ -1521,6 +1552,7 @@
               routingSources: [],
               routingSourcesUpdatedAt: 0,
               routingLocation: "virtualrouter.routing",
+              routingHasLoadBalancing: false,
               credentials: [],
               oauthMode: "manual"
             };
@@ -2782,13 +2814,24 @@
         const serversBody = $("controlServersTbody");
         const hitsBody = $("controlHitsTbody");
         const quotaBody = $("controlQuotaProvidersTbody");
+        const serverToolBody = $("controlServerToolToolsTbody");
+        const serverToolStatus = $("controlServerToolStatus");
+        const serverToolUpdatedAt = $("controlServerToolUpdatedAt");
+        const serverToolCountsBox = $("controlServerToolCountsBox");
+        const serverToolToggleBtn = $("controlServerToolToggleBtn");
         if (serversBody) serversBody.replaceChildren();
         if (hitsBody) hitsBody.replaceChildren();
         if (quotaBody) quotaBody.replaceChildren();
+        if (serverToolBody) serverToolBody.replaceChildren();
         if (!snap || typeof snap !== "object") {
           if (serversBody) serversBody.appendChild(createErrorRow(4, "No snapshot"));
           if (hitsBody) hitsBody.appendChild(createErrorRow(2, "No snapshot"));
           if (quotaBody) quotaBody.appendChild(createErrorRow(5, "No snapshot"));
+          if (serverToolBody) serverToolBody.appendChild(createErrorRow(5, "No snapshot"));
+          if (serverToolStatus) serverToolStatus.textContent = "state: —";
+          if (serverToolUpdatedAt) serverToolUpdatedAt.textContent = "updated: —";
+          if (serverToolCountsBox) serverToolCountsBox.textContent = "executions: —\nsuccess: —\nfailure: —";
+          if (serverToolToggleBtn) serverToolToggleBtn.textContent = "Toggle";
           return;
         }
         const servers = Array.isArray(snap.servers) ? snap.servers : [];
@@ -2865,9 +2908,108 @@
           }
         }
 
+        try {
+          const serverTool = snap.serverTool && typeof snap.serverTool === "object" ? snap.serverTool : null;
+          const state = serverTool && serverTool.state && typeof serverTool.state === "object" ? serverTool.state : null;
+          const stats = serverTool && serverTool.stats && typeof serverTool.stats === "object" ? serverTool.stats : null;
+          const enabled = state && state.enabled === true;
+          if (serverToolStatus) {
+            serverToolStatus.textContent = enabled ? "state: enabled" : "state: disabled";
+            serverToolStatus.className = enabled ? "pill ok" : "pill bad";
+          }
+          if (serverToolUpdatedAt) {
+            const updatedBy = state && state.updatedBy ? String(state.updatedBy) : "—";
+            const updatedAt = state && state.updatedAtMs ? formatTs(state.updatedAtMs) : "—";
+            serverToolUpdatedAt.textContent = `updated: ${updatedAt} by ${updatedBy}`;
+          }
+          if (serverToolToggleBtn) {
+            serverToolToggleBtn.textContent = enabled ? "Disable now" : "Enable now";
+            serverToolToggleBtn.className = enabled ? "danger" : "primary";
+          }
+
+          const executions = stats && Number.isFinite(Number(stats.executions)) ? Number(stats.executions) : 0;
+          const successCount = stats && Number.isFinite(Number(stats.success)) ? Number(stats.success) : 0;
+          const failureCount = stats && Number.isFinite(Number(stats.failure)) ? Number(stats.failure) : 0;
+          const scannedLines = stats && Number.isFinite(Number(stats.scannedLines)) ? Number(stats.scannedLines) : 0;
+          if (serverToolCountsBox) {
+            serverToolCountsBox.textContent =
+              `executions: ${formatInt(executions)}\n` +
+              `success: ${formatInt(successCount)}\n` +
+              `failure: ${formatInt(failureCount)}\n` +
+              `log lines scanned: ${formatInt(scannedLines)}`;
+          }
+
+          const byTool = stats && Array.isArray(stats.byTool) ? stats.byTool : [];
+          const recent = stats && Array.isArray(stats.recent) ? stats.recent : [];
+          const recentByTool = new Map();
+          for (const ev of recent) {
+            const tool = textOf(ev && ev.tool ? ev.tool : "unknown");
+            if (!recentByTool.has(tool)) recentByTool.set(tool, []);
+            recentByTool.get(tool).push(ev);
+          }
+
+          if (!byTool.length) {
+            if (serverToolBody) serverToolBody.appendChild(createErrorRow(5, "No servertool execution records"));
+          } else if (serverToolBody) {
+            for (const row of byTool.slice(0, 30)) {
+              const tool = textOf(row.tool);
+              const tr = document.createElement("tr");
+              tr.appendChild(createCell("td", tool || "—", "mono"));
+              tr.appendChild(createCell("td", formatInt(Number(row.executions || 0)), "mono"));
+              tr.appendChild(createCell("td", formatInt(Number(row.success || 0)), "mono"));
+              tr.appendChild(createCell("td", formatInt(Number(row.failure || 0)), "mono"));
+
+              const detailsTd = document.createElement("td");
+              const detailList = Array.isArray(recentByTool.get(tool)) ? recentByTool.get(tool).slice(0, 8) : [];
+              if (!detailList.length) {
+                detailsTd.textContent = "—";
+              } else {
+                const details = document.createElement("details");
+                const summary = document.createElement("summary");
+                summary.textContent = `${detailList.length} recent`;
+                details.appendChild(summary);
+                const pre = document.createElement("pre");
+                pre.className = "mono";
+                pre.style.margin = "8px 0 0";
+                pre.style.whiteSpace = "pre-wrap";
+                pre.style.wordBreak = "break-word";
+                pre.textContent = detailList
+                  .map((ev) => {
+                    const ts = textOf(ev.ts || "—");
+                    const stage = textOf(ev.stage || "—");
+                    const result = textOf(ev.result || "—");
+                    const req = textOf(ev.requestId || "—");
+                    const msg = textOf(ev.message || "");
+                    return `${ts} | ${req} | ${stage}/${result}${msg ? ` | ${msg}` : ""}`;
+                  })
+                  .join("\n");
+                details.appendChild(pre);
+                detailsTd.appendChild(details);
+              }
+              tr.appendChild(detailsTd);
+              serverToolBody.appendChild(tr);
+            }
+            if (byTool.length > 30) {
+              const tr = document.createElement("tr");
+              const td = document.createElement("td");
+              td.colSpan = 5;
+              td.className = "muted";
+              td.textContent = `… ${byTool.length - 30} more hidden`;
+              tr.appendChild(td);
+              serverToolBody.appendChild(tr);
+            }
+          }
+        } catch {
+          if (serverToolBody) serverToolBody.appendChild(createErrorRow(5, "Failed to parse servertool stats"));
+        }
+
         const hint = $("controlHint");
         if (hint) {
-          hint.textContent = `servers: ${servers.length} · quota keys: ${list.length} · updated: ${formatTs(snap.nowMs)}`;
+          const st = snap.serverTool && snap.serverTool.stats && typeof snap.serverTool.stats === "object"
+            ? snap.serverTool.stats
+            : null;
+          const stCount = st && Number.isFinite(Number(st.executions)) ? Number(st.executions) : 0;
+          hint.textContent = `servers: ${servers.length} · quota keys: ${list.length} · servertool exec: ${stCount} · updated: ${formatTs(snap.nowMs)}`;
         }
       }
 
@@ -3868,7 +4010,7 @@
 	          try { container.innerHTML = ""; } catch {}
 	        }
 	        try {
-	          container.appendChild(routingEditorRenderNode("routing", routingEditorState.value || {}, [], 0, true));
+	          container.appendChild(routingEditorRenderNode("routingPolicy", routingEditorState.value || {}, [], 0, true));
 	        } catch (e) {
 	          const msg = e && e.message ? e.message : String(e);
 	          try { container.textContent = `Render failed: ${msg}`; } catch {}
@@ -4097,6 +4239,43 @@
 	        return details;
 	      }
 
+      function toPlainObjectOrNull(value) {
+        if (!value || typeof value !== "object" || Array.isArray(value)) return null;
+        return value;
+      }
+
+      function buildRoutingPolicyEditorValue(snapshot) {
+        const routingObj = toPlainObjectOrNull(snapshot && snapshot.routing) || {};
+        const hasLoadBalancing = Boolean(snapshot && snapshot.hasLoadBalancing === true);
+        const loadBalancingObj = toPlainObjectOrNull(snapshot && snapshot.loadBalancing) || {};
+        UI.routingHasLoadBalancing = hasLoadBalancing;
+        return {
+          routing: routingObj,
+          loadBalancing: loadBalancingObj
+        };
+      }
+
+      function parseRoutingPolicyEditorValue(value) {
+        const policy = toPlainObjectOrNull(value) || {};
+        const routing = toPlainObjectOrNull(policy.routing) || {};
+        const hasLoadBalancingField = Object.prototype.hasOwnProperty.call(policy, "loadBalancing");
+        const rawLoadBalancing = hasLoadBalancingField ? policy.loadBalancing : undefined;
+        if (
+          rawLoadBalancing !== undefined
+          && rawLoadBalancing !== null
+          && (typeof rawLoadBalancing !== "object" || Array.isArray(rawLoadBalancing))
+        ) {
+          throw new Error("routingPolicy.loadBalancing must be an object");
+        }
+        const loadBalancingNode = toPlainObjectOrNull(rawLoadBalancing);
+        const hasLoadBalancingData = Boolean(loadBalancingNode && Object.keys(loadBalancingNode).length > 0);
+        return {
+          routing,
+          loadBalancing: loadBalancingNode,
+          shouldWriteLoadBalancing: hasLoadBalancingData || UI.routingHasLoadBalancing === true
+        };
+      }
+
 	      async function loadRouting() {
 	        setLog("routingOpLog", "");
 	        const auth = UI.adminAuth ? UI.adminAuth : await refreshAdminAuthStatus();
@@ -4109,9 +4288,12 @@
 	          const query = selectedPath ? `?path=${encodeURIComponent(selectedPath)}` : "";
 	          const out = await apiFetch(`/config/routing${query}`);
 	          UI.routingLocation = out.location || "virtualrouter.routing";
-	          routingEditorSetValue(out.routing || {});
-	          setLog("routingOpLog", `Loaded. Path: ${out.path || "—"}\nLocation: ${UI.routingLocation}`);
-	          toast("Routing loaded.", "ok");
+	          routingEditorSetValue(buildRoutingPolicyEditorValue(out));
+	          setLog(
+              "routingOpLog",
+              `Loaded. Path: ${out.path || "—"}\nLocation: ${UI.routingLocation}\nloadBalancing: ${UI.routingHasLoadBalancing ? "present" : "absent"}`
+            );
+	          toast("Routing policy loaded.", "ok");
 	        } catch (e) {
 	          if (e && e.status === 401) notifyUnauthorizedOnce("routing");
 	          setLog("routingOpLog", `Load failed: ${e.message}`);
@@ -4124,27 +4306,84 @@
 	        const auth = UI.adminAuth ? UI.adminAuth : await refreshAdminAuthStatus();
 	        if (!auth || !auth.authenticated) {
 	          notifyUnauthorizedOnce("routing save");
-	          return;
+	          return null;
 	        }
 	        try {
 	          const selectedPath = textOf($("routingSourceSelect").value || "").trim();
 	          const query = selectedPath ? `?path=${encodeURIComponent(selectedPath)}` : "";
-	          const routing = routingEditorGetValue() || {};
+	          const policy = parseRoutingPolicyEditorValue(routingEditorGetValue() || {});
+            const payload = {
+              routing: policy.routing,
+              location: UI.routingLocation,
+              path: selectedPath || undefined
+            };
+            if (policy.shouldWriteLoadBalancing) {
+              payload.loadBalancing = policy.loadBalancing || {};
+            }
 	          const out = await apiFetch(`/config/routing${query}`, {
 	            method: "PUT",
-	            body: JSON.stringify({ routing, location: UI.routingLocation, path: selectedPath || undefined })
+	            body: JSON.stringify(payload)
 	          });
 	          UI.routingLocation = out.location || UI.routingLocation;
-	          setLog("routingOpLog", `Saved. Path: ${out.path || "—"}\nLocation: ${UI.routingLocation}\nRestart required to apply.`);
-	          toast("Routing saved.", "ok");
+            if (typeof out.hasLoadBalancing === "boolean") {
+              UI.routingHasLoadBalancing = out.hasLoadBalancing;
+            }
+	          setLog(
+              "routingOpLog",
+              `Saved. Path: ${out.path || "—"}\nLocation: ${UI.routingLocation}\nloadBalancing: ${UI.routingHasLoadBalancing ? "present" : "absent"}\nRestart required to apply.`
+            );
+	          toast("Routing policy saved.", "ok");
 	          routingEditorSetDirty(false);
+            return out;
 	        } catch (e) {
 	          if (e && e.status === 401) notifyUnauthorizedOnce("routing save");
 	          setLog("routingOpLog", `Save failed: ${e.message}`);
 	          toast(`Save failed: ${e.message}`);
+            return null;
 	        }
 	      }
 
+      async function saveRoutingAndRestartLocal() {
+        const saved = await saveRouting();
+        if (!saved) return;
+        if (!confirm("Restart this RouteCodex runtime now to apply routing policy changes?")) return;
+        try {
+          const out = await apiFetch("/daemon/restart", { method: "POST" });
+          const warnings = Array.isArray(out.warnings) && out.warnings.length ? `\nWarnings:\n- ${out.warnings.join("\n- ")}` : "";
+          setLog(
+            "routingOpLog",
+            `Saved + restarted local.\nconfigPath: ${out.configPath || "—"}\nreloadedAt: ${out.reloadedAt || "—"}${warnings}`
+          );
+          toast("Routing saved and local runtime restarted.", "ok");
+          await refreshStatus();
+          await refreshProviders();
+          await refreshCredentials();
+          await refreshQuota();
+          await refreshQuotaSnapshot();
+          await refreshRuntimes();
+          await refreshControl();
+        } catch (e) {
+          setLog("routingOpLog", `Local restart failed: ${e && e.message ? e.message : String(e)}`);
+          toast(`Local restart failed: ${e && e.message ? e.message : String(e)}`);
+        }
+      }
+
+      async function saveRoutingAndRestartAll() {
+        const saved = await saveRouting();
+        if (!saved) return;
+        if (!confirm("Broadcast restart all local RouteCodex servers now?")) return;
+        try {
+          const out = await controlMutate("servers.restart", {});
+          setLog("routingOpLog", `Saved + restart-all requested.\n${JSON.stringify(out, null, 2)}`);
+          toast("Routing saved and restart-all requested.", "ok");
+          await refreshControl();
+          await refreshStatus();
+        } catch (e) {
+          setLog("routingOpLog", `Restart-all failed: ${e && e.message ? e.message : String(e)}`);
+          toast(`Restart-all failed: ${e && e.message ? e.message : String(e)}`);
+        }
+      }
+
 	      async function refreshRoutingSources() {
 	        const select = $("routingSourceSelect");
 	        if (!select) return;
@@ -5212,6 +5451,22 @@
           setLog("controlOpLog", `Quota refresh failed: ${e.message || e}`);
         }
       });
+      $("controlServerToolToggleBtn").addEventListener("click", async () => {
+        setLog("controlOpLog", "");
+        try {
+          const snap = UI.controlSnapshot && typeof UI.controlSnapshot === "object" ? UI.controlSnapshot : null;
+          const state = snap && snap.serverTool && typeof snap.serverTool === "object" && snap.serverTool.state
+            ? snap.serverTool.state
+            : null;
+          const enabled = state && state.enabled === true;
+          const nextEnabled = !enabled;
+          const out = await controlMutate("servertool.set_enabled", { enabled: nextEnabled });
+          setLog("controlOpLog", `ServerTool ${nextEnabled ? "enabled" : "disabled"}.\n${JSON.stringify(out, null, 2)}`);
+          await refreshControl();
+        } catch (e) {
+          setLog("controlOpLog", `ServerTool toggle failed: ${e.message || e}`);
+        }
+      });
       $("controlQuotaOfflineBtn").addEventListener("click", async () => {
         setLog("controlOpLog", "");
         const providerKey = ($("controlQuotaKeyInput").value || "").trim();
@@ -5385,6 +5640,8 @@
 
       $("loadRoutingBtn").addEventListener("click", loadRouting);
       $("saveRoutingBtn").addEventListener("click", saveRouting);
+      $("saveRoutingRestartLocalBtn").addEventListener("click", () => void saveRoutingAndRestartLocal());
+      $("saveRoutingRestartAllBtn").addEventListener("click", () => void saveRoutingAndRestartAll());
       $("refreshRoutingSourcesBtn").addEventListener("click", refreshRoutingSources);
       $("routingSourceSelect").addEventListener("change", loadRouting);
       $("routingRegExpandBtn").addEventListener("click", () => setAllDetailsOpen($("routingKvEditor"), true));

package/docs/file-line-limit-gate.md

@@ -0,0 +1,30 @@
+# File Line-Limit Gate (Host)
+
+Host (`routecodex`) enforces a hard line limit for changed code files.
+
+## Rule
+
+- Limit: `< 500` lines per file.
+- Scope: changed files in current diff range.
+- Trigger: CI `tests` workflow (`verify:file-line-limit`).
+
+## Policy File
+
+- `config/file-line-limit-policy.json`
+- Keys:
+  - `limit`: max lines
+  - `extensions`: checked code suffixes
+  - `excludeDirs`: skipped directory prefixes
+  - `allowList`: temporary exemptions (exact paths)
+
+## Exemption Process
+
+1. Only use `allowList` for short-lived exceptions.
+2. Add a linked `bd --no-db` issue for removal/splitting.
+3. Remove exemption in the follow-up PR after refactor lands.
+
+## Local Run
+
+```bash
+npm run verify:file-line-limit
+```

package/docs/refactoring/host-164.3-responsibility-migration.md

@@ -0,0 +1,62 @@
+# Host Refactor 164.3 Responsibility Migration
+
+This document records the responsibility split for `routecodex-164.3`:
+- `src/server/runtime/http-server/request-executor.ts`
+- `src/providers/core/runtime/http-transport-provider.ts`
+
+## Current Line-Count Snapshot
+
+| File | Lines |
+|---|---:|
+| `src/server/runtime/http-server/request-executor.ts` | 497 |
+| `src/providers/core/runtime/http-transport-provider.ts` | 485 |
+| `src/providers/core/runtime/http-request-executor.ts` | 474 |
+| `src/providers/core/runtime/gemini-cli-http-provider.ts` | 479 |
+
+## Request Executor Split Map
+
+| Old responsibility (in `request-executor.ts`) | New module |
+|---|---|
+| Retry decisions, max attempts, backoff | `src/server/runtime/http-server/executor/retry-engine.ts` |
+| Antigravity retry signal, status/error extraction, clock binding | `src/server/runtime/http-server/executor/request-retry-helpers.ts` |
+| Provider response normalization, status/model extraction, request semantics extraction | `src/server/runtime/http-server/executor/provider-response-utils.ts` |
+| Runtime resolution + provider error emission path | `src/server/runtime/http-server/executor/provider-runtime-resolver.ts` |
+| Provider request identity (`requestId`/`providerLabel`/model/protocol) | `src/server/runtime/http-server/executor/provider-request-context.ts` |
+| Provider response conversion bridge | `src/server/runtime/http-server/executor/provider-response-converter.ts` |
+| Usage logging output | `src/server/runtime/http-server/executor/usage-logger.ts` |
+| Inbound snapshot + pool-exhausted classifier | `src/server/runtime/http-server/executor/request-executor-core-utils.ts` |
+
+## Provider Runtime / Transport Split Map
+
+| Old responsibility (`http-transport-provider.ts` / related runtime hot files) | New module |
+|---|---|
+| Auth provider + HTTP client bootstrap wiring | `src/providers/core/runtime/provider-bootstrap-utils.ts` |
+| Runtime profile/runtimeKey/codex mode context shaping | `src/providers/core/runtime/provider-runtime-utils.ts` |
+| HTTP executor retry + OAuth replay helpers | `src/providers/core/runtime/provider-http-executor-utils.ts` |
+| Service-profile resolution | `src/providers/core/runtime/service-profile-resolver.ts` |
+| Endpoint/baseURL resolution | `src/providers/core/runtime/runtime-endpoint-resolver.ts` |
+| Provider request preprocess + metadata normalization | `src/providers/core/runtime/provider-request-preprocessor.ts` |
+| Request header orchestration | `src/providers/core/runtime/provider-request-header-orchestrator.ts` |
+| Family/profile specific shaping helpers | `src/providers/core/runtime/provider-family-profile-utils.ts` |
+| Request shaping helpers | `src/providers/core/runtime/provider-request-shaping-utils.ts` |
+| Response postprocessing hooks | `src/providers/core/runtime/provider-response-postprocessor.ts` |
+| Gemini CLI response normalization | `src/providers/core/runtime/gemini-cli-response-postprocessor.ts` |
+| Gemini SSE normalization | `src/providers/core/runtime/gemini-sse-normalizer.ts` |
+| IFlow business-error classifier helpers | `src/providers/core/runtime/provider-iflow-business-error-utils.ts` |
+| Session/Codex header injection helpers | `src/providers/core/runtime/transport/session-header-utils.ts` |
+| OAuth header preflight helpers | `src/providers/core/runtime/transport/oauth-header-preflight.ts` |
+| Request header builder | `src/providers/core/runtime/transport/request-header-builder.ts` |
+| Auth mode utility | `src/providers/core/runtime/transport/auth-mode-utils.ts` |
+
+## Regression Verification Commands
+
+```bash
+npm run jest:run -- --runTestsByPath \
+  tests/server/runtime/request-executor.single-attempt.spec.ts \
+  tests/providers/core/runtime/gemini-cli-http-provider.unit.test.ts \
+  tests/providers/core/runtime/deepseek-http-provider.unit.test.ts
+```
+
+```bash
+npm run build:dev
+```

package/docs/release-iflow-400-gate.md

@@ -0,0 +1,58 @@
+# iflow 400 Fix Release Gate (llms -> rcc)
+
+Date: 2026-02-15
+Issue: routecodex-75
+Prerequisite bug: routecodex-70 (CLOSED)
+
+## 1) Preconditions
+- routecodex-70 is closed with replay evidence.
+- Matching tags exist on both repos:
+  - `rc-v0.6.1733-v0.89.1798`
+  - `rc-v0.6.1733-v0.89.1799`
+- Artifacts prepared:
+  - `/tmp/jsonstudio-llms-0.6.1733.tgz`
+  - `jsonstudio-rcc-0.89.1799.tgz`
+
+## 2) Mandatory verification gate
+1. Same-shape replay (affected provider: iflow)
+- sample: `req_1770377342669_6d3e332c`
+- expected: HTTP 200, no 514 business error
+
+2. Control replay (unaffected provider)
+- sample: `req_1770381056958_a35ee9ff` (tab)
+- expected: HTTP 200, behavior unchanged
+
+3. Runtime/build checks
+- `npm run jest:run -- --runTestsByPath tests/server/runtime/http-server/executor-provider.spec.ts tests/server/runtime/executor-provider.retryable.spec.ts --runInBand`
+- `npm run jest:run -- --runTestsByPath tests/server/runtime/request-executor.single-attempt.spec.ts tests/server/http-server/execute-pipeline-failover.spec.ts --runInBand`
+- `npm run build:dev`
+- `npm run install:global`
+
+## 3) Publish order (manual)
+1. Publish llms first
+- `cd sharedmodule/llmswitch-core`
+- `npm publish /tmp/jsonstudio-llms-0.6.1733.tgz`
+
+2. Verify rcc package depends on published llms version
+- `cd /Users/fanzhang/Documents/github/routecodex`
+- check `package.json` has `@jsonstudio/llms: 0.6.1733`
+
+3. Publish rcc
+- `npm publish jsonstudio-rcc-0.89.1799.tgz`
+
+## 4) Rollback checklist
+- If llms publish breaks replay gate:
+  1. stop rcc publish
+  2. rollback to previous llms version in routecodex package.json
+  3. rebuild + rerun same-shape/control replays
+
+- If rcc publish breaks runtime:
+  1. pin users back to previous rcc version
+  2. collect failing requestId/providerKey/model
+  3. preserve same-shape/control replay outputs
+
+## 5) Evidence references
+- routecodex-70 notes include request-level replay outputs:
+  - `/tmp/replay-iflow-base-req_1770377342669_6d3e332c.txt`
+  - `/tmp/replay-iflow-stop-followup-20260207T100636992-003.withkey.txt`
+  - `/tmp/replay-control-tab-key1.req_1770381056958_a35ee9ff.txt`

package/docs/replay-evidence-iflow-400.txt

@@ -0,0 +1,33 @@
+# iflow 400 Fix Same-shape & Control Replay Evidence
+
+## Affected Provider (iflow.kimi-k2.5) - Same-shape Replay
+
+### Sample 1: Base request (req_1770377342669_6d3e332c)
+- Before fix: HTTP 400 business error 514
+- After fix: HTTP 200, successful response
+- Output saved: /tmp/replay-iflow-base-req_1770377342669_6d3e332c.txt (no longer available)
+- Verification command: `npm run replay:codex-sample -- --sample ~/.routecodex/codex-samples/openai-responses/iflow.1-186.kimi-k2.5/req_1770377342669_6d3e332c/client-request.json`
+
+### Sample 2: Stop_followup request (openai-responses-iflow.1-186...20260207T100636992-003_stop_followup)
+- Before fix: HTTP 400 business error 514
+- After fix: HTTP 200, successful response
+- Output saved: /tmp/replay-iflow-stop-followup-20260207T100636992-003.withkey.txt (no longer available)
+- Verification command: `npm run replay:codex-sample -- --sample ~/.routecodex/codex-samples/openai-responses/iflow.1-186.kimi-k2.5/.../client-request.json`
+
+## Unaffected Provider (tab.gpt-5.3-codex) - Control Replay
+
+### Sample: req_1770381056958_a35ee9ff
+- Before fix: HTTP 200, unchanged
+- After fix: HTTP 200, unchanged
+- Output saved: /tmp/replay-control-tab-key1.req_1770381056958_a35ee9ff.txt (no longer available)
+- Verification command: `npm run replay:codex-sample -- --sample ~/.routecodex/codex-samples/openai-chat/tab.key1.gpt-5.3-codex/req_1770381056958_a35ee9ff/client-request.json`
+
+## Summary
+All same-shape and control replays pass. The fix ensures:
+- iflow kimi-k2.5 requests no longer return 514 for large payloads (compat layer adjustments, followup trimming policy).
+- Control provider (tab) behavior remains unchanged.
+- Retry policy now treats 514 as retryable, allowing failover to next provider.
+
+## References
+- routecodex-70 notes contain detailed analysis.
+- Unit tests: `tests/server/runtime/http-server/executor-provider.spec.ts`, `tests/server/runtime/executor-provider.retryable.spec.ts`.

package/docs/stop-message-auto.md

@@ -7,6 +7,7 @@
 - 语法与现有 `<**!glm**>` / `<**#glm**>` 一致：只在路由层解析，provider 不看到标签。
 - 所有自动续写逻辑通过 **servertool** 实现，与 `vision_auto` / `gemini_empty_reply_continue` 保持一致。
 - 状态通过 sticky session 持久化；支持清理与覆盖。
+- stopMessage 文案建议使用“执行导向”措辞（例如“立即执行待处理任务”），避免模型只做任务查看不执行。
 - 客户端真实断开（HTTP aborted/close 未写完）后不会继续自动请求，避免后台无限回环；正常完成不会被误判。
 
 ## 1. 语法与状态
@@ -15,10 +16,10 @@
 
 支持三种形式：
 
-- 启用 / 更新：`<**stopMessage:"继续",3**>`
-  - `"继续"`：自动补发的用户消息内容，内部如需引号，用 `\"` 转义。
+- 启用 / 更新：`<**stopMessage:"立即执行待处理任务",3**>`
+  - `"立即执行待处理任务"`：自动补发的用户消息内容，要求模型直接执行而非查看状态。内部如需引号，用 `\"` 转义。
   - `3`：本会话最多自动续写 3 轮；省略时默认 `10`。
-- 仅设置文案（默认 10 次）：`<**stopMessage:"继续"**>` → `maxRepeats = 10`。
+- 仅设置文案（默认 10 次）：`<**stopMessage:"立即执行待处理任务，如有未完成任务直接完成并提交结果"**>` → `maxRepeats = 10`。
 - 清理：`<**stopMessage:clear**>` → 清空本会话 stopMessage 状态。
 
 > 注意：这些标签仅用于路由与 servertool，不应出现在发给 provider 的文本中。

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jsonstudio/rcc",
-  "version": "0.89.1960",
+  "version": "0.89.2195",
   "description": "Multi-provider OpenAI proxy server with anthropic/responses/chat support (release)",
   "type": "module",
   "main": "dist/index.js",
@@ -45,8 +45,9 @@
     "test:ci": "npm run test:ci:jest && npm run mock:regressions && npm run test:blackbox-antigravity-parity",
     "test:ci:coverage": "npm run test:ci:jest:coverage",
     "verify:repo-sanity": "node scripts/ci/repo-sanity.mjs",
+    "verify:file-line-limit": "node scripts/ci/check-file-line-limit.mjs",
     "vr:dryrun": "node scripts/virtual-router-dryrun.mjs",
-    "test:routing-instructions": "npm run jest:run -- --runTestsByPath tests/token-daemon/token-daemon.auto-refresh-noninteractive.spec.ts tests/token-daemon/token-daemon.refresh-ahead.spec.ts tests/commands/oauth-command.qwen-auto.spec.ts tests/token-daemon/interactive-refresh.spec.ts tests/server/runtime/request-executor.single-attempt.spec.ts tests/server/runtime/executor-provider.retryable.spec.ts tests/providers/auth/tokenfile-auth.iflow.spec.ts tests/providers/auth/antigravity-user-agent.unit.test.ts tests/providers/auth/antigravity-fingerprint.unit.test.ts tests/providers/auth/antigravity-warmup.unit.test.ts tests/providers/core/runtime/gemini-cli-http-provider.unit.test.ts tests/providers/core/runtime/antigravity-quota-client.unit.test.ts tests/manager/quota/provider-quota-center.spec.ts tests/manager/quota/provider-quota-store.spec.ts tests/manager/quota/quota-manager-refresh.spec.ts tests/manager/quota/provider-quota-daemon-module.spec.ts tests/manager/quota/provider-key-normalization.spec.ts tests/server/http-server/daemon-admin.e2e.spec.ts tests/server/http-server/clock-client-routes.spec.ts tests/server/http-server/quota-view-injection.spec.ts tests/server/http-server/quota-refresh-triggers.e2e.spec.ts tests/server/http-server/hub-policy-injection.spec.ts tests/server/http-server/session-header-injection.spec.ts tests/server/http-server/session-dir.spec.ts tests/server/handlers/sse-timeout.spec.ts tests/utils/is-direct-execution.test.ts tests/utils/windows-netstat.test.ts tests/servertool/virtual-router-context-fallback.spec.ts tests/servertool/virtual-router-longcontext-fallback.spec.ts tests/servertool/virtual-router-series-cooldown.spec.ts tests/servertool/recursive-detection-guard.spec.ts tests/servertool/routing-instructions.spec.ts tests/servertool/stop-message-auto.spec.ts tests/servertool/stopmessage-session-scope.spec.ts tests/servertool/stopmessage-anthropic-stop-sequence.spec.ts tests/servertool/stopmessage-compaction-false-positive.spec.ts tests/servertool/servertool-progress-logging.spec.ts tests/servertool/servertool-clock.spec.ts tests/servertool/antigravity-thought-signature-bootstrap.spec.ts tests/compat/tabglm-claude-code-profile.spec.ts tests/compat/antigravity-thought-signature.spec.ts tests/sharedmodule/mcp-tool-descriptions.spec.ts tests/sharedmodule/virtual-router-routing-model-validation.spec.ts tests/unified-hub/hub-v1-single-path-imports.spec.ts",
+    "test:routing-instructions": "npm run jest:run -- --runTestsByPath tests/token-daemon/token-daemon.auto-refresh-noninteractive.spec.ts tests/token-daemon/token-daemon.refresh-ahead.spec.ts tests/commands/oauth-command.qwen-auto.spec.ts tests/token-daemon/interactive-refresh.spec.ts tests/server/runtime/request-executor.single-attempt.spec.ts tests/server/runtime/executor-provider.retryable.spec.ts tests/server/runtime/http-server/executor/usage-aggregator.spec.ts tests/providers/auth/tokenfile-auth.iflow.spec.ts tests/providers/auth/antigravity-user-agent.unit.test.ts tests/providers/auth/antigravity-fingerprint.unit.test.ts tests/providers/auth/antigravity-warmup.unit.test.ts tests/providers/core/runtime/gemini-cli-http-provider.unit.test.ts tests/providers/core/runtime/deepseek-http-provider.unit.test.ts tests/providers/core/runtime/antigravity-quota-client.unit.test.ts tests/manager/quota/provider-quota-center.spec.ts tests/manager/quota/provider-quota-store.spec.ts tests/manager/quota/quota-manager-refresh.spec.ts tests/manager/quota/provider-quota-daemon-module.spec.ts tests/manager/quota/provider-key-normalization.spec.ts tests/server/http-server/daemon-admin.e2e.spec.ts tests/server/http-server/clock-client-routes.spec.ts tests/server/http-server/quota-view-injection.spec.ts tests/server/http-server/quota-refresh-triggers.e2e.spec.ts tests/server/http-server/hub-policy-injection.spec.ts tests/server/http-server/session-header-injection.spec.ts tests/server/http-server/session-dir.spec.ts tests/server/handlers/sse-timeout.spec.ts tests/utils/is-direct-execution.test.ts tests/utils/windows-netstat.test.ts tests/servertool/virtual-router-context-fallback.spec.ts tests/servertool/virtual-router-longcontext-fallback.spec.ts tests/servertool/virtual-router-series-cooldown.spec.ts tests/servertool/recursive-detection-guard.spec.ts tests/servertool/routing-instructions.spec.ts tests/servertool/stop-message-auto.spec.ts tests/servertool/stopmessage-session-scope.spec.ts tests/servertool/stopmessage-anthropic-stop-sequence.spec.ts tests/servertool/stopmessage-compaction-false-positive.spec.ts tests/servertool/servertool-progress-logging.spec.ts tests/servertool/servertool-clock.spec.ts tests/servertool/antigravity-thought-signature-bootstrap.spec.ts tests/compat/tabglm-claude-code-profile.spec.ts tests/compat/antigravity-thought-signature.spec.ts tests/sharedmodule/mcp-tool-descriptions.spec.ts tests/sharedmodule/virtual-router-hit-log.spec.ts tests/sharedmodule/virtual-router-routing-model-validation.spec.ts tests/unified-hub/hub-v1-single-path-imports.spec.ts",
     "test:cli": "npm run jest:run -- --runTestsByPath tests/cli/camoufox-command.spec.ts tests/cli/clean-command.spec.ts tests/cli/code-command.spec.ts tests/cli/config-command.spec.ts tests/cli/env-command.spec.ts tests/cli/env-output.spec.ts tests/cli/examples-command.spec.ts tests/cli/port-command.spec.ts tests/cli/port-utils.spec.ts tests/cli/restart-command.spec.ts tests/cli/smoke.spec.ts tests/cli/start-command.spec.ts tests/cli/status-command.spec.ts tests/cli/stop-command.spec.ts",
     "test:watch": "npm run jest:run -- --watch",
     "test:coverage": "npm run jest:run -- --coverage",
@@ -149,7 +150,7 @@
   },
   "dependencies": {
     "@anthropic-ai/sdk": "^0.65.0",
-    "@jsonstudio/llms": "^0.6.1892",
+    "@jsonstudio/llms": "0.6.2125",
     "@lmstudio/sdk": "^1.5.0",
     "@radix-ui/react-switch": "^1.2.6",
     "@types/socket.io": "^3.0.1",