fix(sse): merge purify_history compression notice into the leading system message (#11113)

Validated on the combined batch board: purify-system-first suite 4/4, typecheck clean. Pre-merge: file-size baseline gained a frozen entry for contextManager.ts at 1001 (+1, this PR's merge-into-leading-system branch) with a dated annotation — the gate caps unlisted files at 1000. Producer side of the live-confirmed TokenRouter 400 class: no internal path emits a mid-array system message anymore. Thank you @ggdayup — the call-log evidence made this airtight!

config/quality/file-size-baseline.json

@@ -388,6 +388,8 @@
     "open-sse/services/claudeCodeCompatible.ts": 1563,
     "open-sse/services/combo.ts": 4742,
     "open-sse/services/compression/strategySelector.ts": 1379,
+    "open-sse/services/contextManager.ts": 1001,
+    "_rebaseline_2026_08_22_11113_purify_system_first": "PR #11113 (ggdayup) own growth: open-sse/services/contextManager.ts 1000->1001 (+1, purifyHistory merges the compression notice into the leading system message instead of splicing a second one mid-array — live-confirmed TokenRouter 400s; the +1 is the merge-into-leading branch, not extractable). Covered by tests/unit/context-manager-purify-system-first.test.ts. Owner pre-authorized baseline bumps 2026-08-22.",
     "open-sse/services/rateLimitManager.ts": 1517,
     "open-sse/translator/response/openai-responses.ts": 1652,
     "open-sse/utils/cursorAgentProtobuf.ts": 1956,

open-sse/services/contextManager.ts

@@ -669,13 +669,35 @@ function purifyHistory(messages: Record<string, unknown>[], targetTokens: number
   result = fixToolPairs(result);
   result = stripTrailingAssistantOrphanToolUse(result);
 
-  // Add summary of dropped messages
+  // Add summary of dropped messages. Merge the notice INTO the leading
+  // system/developer message instead of splicing a second system-role message
+  // mid-array: strict gateways (TokenRouter confirmed live 2026-08-22, see the
+  // PROVIDERS_SYSTEM_MUST_BE_FIRST list in src/lib/memory/injection.ts) reject
+  // any system message at index > 0 with HTTP 400 "System message must be at
+  // the beginning". When there is no leading system message, prepend one --
+  // index 0 is accepted by every provider (same slot the old splice used when
+  // system[] was empty).
   if (keep < nonSystem.length) {
     const dropped = nonSystem.length - keep;
-    result.splice(system.length, 0, {
-      role: "system",
-      content: `[Context compressed: ${dropped} earlier messages removed to fit context window]`,
-    });
+    const droppedNotice = `[Context compressed: ${dropped} earlier messages removed to fit context window]`;
+    const first = result[0];
+    if (first && (first.role === "system" || first.role === "developer")) {
+      if (typeof first.content === "string") {
+        result[0] = {
+          ...first,
+          content: first.content ? `${droppedNotice}\n${first.content}` : droppedNotice,
+        };
+      } else if (Array.isArray(first.content)) {
+        result[0] = {
+          ...first,
+          content: [{ type: "text", text: droppedNotice }, ...(first.content as unknown[])],
+        };
+      } else {
+        result[0] = { ...first, content: droppedNotice };
+      }
+    } else {
+      result.unshift({ role: "system", content: droppedNotice });
+    }
   }
 
   return result;

tests/unit/context-manager-purify-system-first.test.ts

@@ -0,0 +1,91 @@
+import { test } from "node:test";
+import assert from "node:assert/strict";
+import { compressContext } from "../../open-sse/services/contextManager.ts";
+
+/**
+ * Plan-A root fix for the 2026-08-22 tokenrouter 400s. purifyHistory() used to
+ * splice the `[Context compressed: …]` notice as a SECOND system-role message at
+ * index system.length; strict gateways (TokenRouter, xiaomi-mimo/mimo) reject any
+ * system message at index > 0 with HTTP 400 "System message must be at the
+ * beginning". The notice must now merge into the leading system/developer
+ * message — or prepend a single system message when none exists — so the output
+ * never contains a system role after index 0, for ANY provider.
+ */
+
+function bigTurn(n: number) {
+  return { role: "user", content: `turn ${n}: ${"x".repeat(4_000)}` };
+}
+
+function run(body: Record<string, unknown>) {
+  // ~30k tokens of history vs a small target forces Layer-3 purify_history.
+  return compressContext(body, { maxTokens: 5_000, reserveTokens: 0 });
+}
+
+function systemIndices(messages: Array<{ role: string }>) {
+  return messages.map((m, i) => (m.role === "system" ? i : -1)).filter((i) => i >= 0);
+}
+
+test("purify_history merges dropped-notice into existing leading system message", () => {
+  const body = {
+    model: "any-model",
+    messages: [
+      { role: "system", content: "You are a helpful assistant." },
+      ...Array.from({ length: 12 }, (_, i) => bigTurn(i)),
+    ],
+  };
+  const result = run(body);
+  assert.equal(result.compressed, true);
+  const messages = (result.body as { messages: Array<Record<string, unknown>> }).messages;
+  assert.deepEqual(systemIndices(messages as Array<{ role: string }>).slice(1), []);
+  const first = messages[0];
+  assert.equal(first.role, "system");
+  const text = String(first.content);
+  assert.match(text, /Context compressed: \d+ earlier messages removed/);
+  assert.match(text, /You are a helpful assistant\./);
+});
+
+test("purify_history prepends a single system notice when no system message exists", () => {
+  const body = {
+    model: "any-model",
+    messages: Array.from({ length: 12 }, (_, i) => bigTurn(i)),
+  };
+  const result = run(body);
+  assert.equal(result.compressed, true);
+  const messages = (result.body as { messages: Array<Record<string, unknown>> }).messages;
+  assert.deepEqual(systemIndices(messages as Array<{ role: string }>), [0]);
+  assert.match(String(messages[0].content), /Context compressed: \d+ earlier messages removed/);
+});
+
+test("purify_history merges into leading developer message without adding a second one", () => {
+  const body = {
+    model: "any-model",
+    messages: [
+      { role: "developer", content: "dev instructions" },
+      ...Array.from({ length: 12 }, (_, i) => bigTurn(i)),
+    ],
+  };
+  const result = run(body);
+  assert.equal(result.compressed, true);
+  const messages = (result.body as { messages: Array<Record<string, unknown>> }).messages;
+  assert.deepEqual(
+    messages.filter((m) => m.role === "developer").length,
+    1,
+    "exactly one developer message"
+  );
+  assert.match(String(messages[0].content), /Context compressed: \d+ earlier messages removed/);
+  assert.match(String(messages[0].content), /dev instructions/);
+});
+
+test("no compression means no notice and untouched history", () => {
+  const body = {
+    model: "any-model",
+    messages: [
+      { role: "system", content: "sys" },
+      { role: "user", content: "hi" },
+    ],
+  };
+  const result = run(body);
+  assert.equal(result.compressed, false);
+  const messages = (result.body as { messages: unknown[] }).messages;
+  assert.equal(messages.length, 2);
+});