fix(agents): keep large read tool results visible

CHANGELOG.md

@@ -129,6 +129,7 @@ Docs: https://docs.openclaw.ai
 - Discord: keep REST, webhook, and monitor traffic on the configured proxy, preserve component-only media sends, honor `@everyone` and `@here` mention gates, keep ACK reactions on the active account, and split voice connect/playback timeouts so auto-join is more reliable. (#57465, #60361, #60345) Thanks @geekhuashan.
 - WhatsApp: restore `channels.whatsapp.blockStreaming` and reset watchdog timeouts after reconnect so quiet chats stop falling into reconnect loops. (#60007, #60069) Thanks @MonkeyLeeT and @mcaxtr.
 - Memory: keep `memory-core` builtin embedding registration on the already-registered path so selecting `memory-core` no longer recurses through plugin discovery and crashes during startup. (#61402) Thanks @ngutman.
+- Agents/tool results: keep larger `read` outputs visible on big-window models by raising the live tool-result ceiling instead of compacting normal file reads right after the first section. Thanks @vincentkoc.
 - Memory/QMD: prefer modern `qmd collection add --glob`, accept newer single-line JSON hit metadata while keeping legacy line fields, refresh QMD docs/doctor install guidance and model-override guidance, and keep older QMD releases working. Thanks @vincentkoc.
 - MS Teams: download inline DM images via Graph API and preserve channel reply threading in proactive fallback. (#52212, #55198) Thanks @Ted-developer and @hyojin.
 - MS Teams: replace the deprecated Teams SDK HttpPlugin stub with `httpServerAdapter` so recurring gateway deprecation warnings stop firing and the Express 5 compatibility workaround stays on the supported SDK path. (#60939) Thanks @coolramukaka-sys.

src/agents/pi-embedded-runner/tool-result-truncation.test.ts

@@ -199,7 +199,7 @@ describe("calculateMaxToolResultChars", () => {
   });
 
   it("exports the live cap through both constant names", () => {
-    expect(DEFAULT_MAX_LIVE_TOOL_RESULT_CHARS).toBe(40_000);
+    expect(DEFAULT_MAX_LIVE_TOOL_RESULT_CHARS).toBe(120_000);
     expect(HARD_MAX_TOOL_RESULT_CHARS).toBe(DEFAULT_MAX_LIVE_TOOL_RESULT_CHARS);
   });
 
@@ -212,6 +212,20 @@ describe("calculateMaxToolResultChars", () => {
     const result = calculateMaxToolResultChars(128_000);
     expect(result).toBe(DEFAULT_MAX_LIVE_TOOL_RESULT_CHARS);
   });
+
+  it("keeps moderately large reads intact on 128K contexts", () => {
+    const messages: AgentMessage[] = [
+      makeUserMessage("hello"),
+      makeAssistantMessage("reading changelog"),
+      makeToolResult("x".repeat(60_000)),
+    ];
+    const { messages: result, truncatedCount } = truncateOversizedToolResultsInMessages(
+      messages,
+      128_000,
+    );
+    expect(truncatedCount).toBe(0);
+    expect(result).toEqual(messages);
+  });
 });
 
 describe("isOversizedToolResult", () => {

src/agents/pi-embedded-runner/tool-result-truncation.ts

@@ -18,10 +18,11 @@ const MAX_TOOL_RESULT_CONTEXT_SHARE = 0.3;
  *
  * Pi already truncates tool results aggressively when serializing old history
  * for compaction summaries. For the live request path we keep a larger slice so
- * the model can still act on recent tool output, but we still want a bounded
- * request-local ceiling that cannot dominate the next turn.
+ * the model can still act on recent tool output, especially large read results
+ * on modern 128K+ context models, while still keeping a bounded request-local
+ * ceiling that cannot dominate the next turn.
  */
-export const DEFAULT_MAX_LIVE_TOOL_RESULT_CHARS = 40_000;
+export const DEFAULT_MAX_LIVE_TOOL_RESULT_CHARS = 120_000;
 
 /**
  * Backwards-compatible alias for older call sites/tests.