fix(coding-agent): truncate tool results in compaction summarization to prevent overflow, fixes #1796

packages/coding-agent/CHANGELOG.md

@@ -12,6 +12,7 @@
 - Fixed custom editors having their `onEscape`/`onCtrlD` handlers unconditionally overwritten by app-level defaults, making vim-style escape handling impossible ([#1838](https://github.com/badlogic/pi-mono/issues/1838))
 - Fixed auto-compaction retriggering on the first prompt after compaction due to stale pre-compaction assistant usage ([#1860](https://github.com/badlogic/pi-mono/issues/1860) by [@joelhooks](https://github.com/joelhooks))
 - Fixed sessions never auto-compacting when hitting persistent API errors (e.g. 529 overloaded) by estimating context size from the last successful response ([#1834](https://github.com/badlogic/pi-mono/issues/1834))
+- Fixed compaction summarization requests exceeding context limits by truncating tool results to 2k chars ([#1796](https://github.com/badlogic/pi-mono/issues/1796))
 - Fixed `/new` leaving startup header content, including the changelog, visible after starting a fresh session ([#1880](https://github.com/badlogic/pi-mono/issues/1880))
 - Fixed misleading docs and example implying that returning `{ isError: true }` from a tool's `execute` function marks the execution as failed; errors must be signaled by throwing ([#1881](https://github.com/badlogic/pi-mono/issues/1881))
 - Fixed model switches through non-reasoning models to preserve the saved default thinking level instead of persisting a capability-forced `off` clamp ([#1864](https://github.com/badlogic/pi-mono/issues/1864))

packages/coding-agent/src/core/compaction/utils.ts

@@ -85,10 +85,26 @@ export function formatFileOperations(readFiles: string[], modifiedFiles: string[
 // Message Serialization
 // ============================================================================
 
+/** Maximum characters for a tool result in serialized summaries. */
+const TOOL_RESULT_MAX_CHARS = 2000;
+
+/**
+ * Truncate text to a maximum character length for summarization.
+ * Keeps the beginning and appends a truncation marker.
+ */
+function truncateForSummary(text: string, maxChars: number): string {
+	if (text.length <= maxChars) return text;
+	const truncatedChars = text.length - maxChars;
+	return `${text.slice(0, maxChars)}\n\n[... ${truncatedChars} more characters truncated]`;
+}
+
 /**
  * Serialize LLM messages to text for summarization.
  * This prevents the model from treating it as a conversation to continue.
  * Call convertToLlm() first to handle custom message types.
+ *
+ * Tool results are truncated to keep the summarization request within
+ * reasonable token budgets. Full content is not needed for summarization.
  */
 export function serializeConversation(messages: Message[]): string {
 	const parts: string[] = [];
@@ -137,7 +153,7 @@ export function serializeConversation(messages: Message[]): string {
 				.map((c) => c.text)
 				.join("");
 			if (content) {
-				parts.push(`[Tool result]: ${content}`);
+				parts.push(`[Tool result]: ${truncateForSummary(content, TOOL_RESULT_MAX_CHARS)}`);
 			}
 		}
 	}

packages/coding-agent/test/compaction-serialization.test.ts

@@ -0,0 +1,79 @@
+import type { Message } from "@mariozechner/pi-ai";
+import { describe, expect, it } from "vitest";
+import { serializeConversation } from "../src/core/compaction/utils.js";
+
+describe("serializeConversation", () => {
+	it("should truncate long tool results", () => {
+		const longContent = "x".repeat(5000);
+		const messages: Message[] = [
+			{
+				role: "toolResult",
+				toolCallId: "tc1",
+				toolName: "read",
+				content: [{ type: "text", text: longContent }],
+				isError: false,
+				timestamp: Date.now(),
+			},
+		];
+
+		const result = serializeConversation(messages);
+
+		expect(result).toContain("[Tool result]:");
+		expect(result).toContain("[... 3000 more characters truncated]");
+		expect(result).not.toContain("x".repeat(3000));
+		// First 2000 chars should be present
+		expect(result).toContain("x".repeat(2000));
+	});
+
+	it("should not truncate short tool results", () => {
+		const shortContent = "x".repeat(1500);
+		const messages: Message[] = [
+			{
+				role: "toolResult",
+				toolCallId: "tc1",
+				toolName: "read",
+				content: [{ type: "text", text: shortContent }],
+				isError: false,
+				timestamp: Date.now(),
+			},
+		];
+
+		const result = serializeConversation(messages);
+
+		expect(result).toBe(`[Tool result]: ${shortContent}`);
+		expect(result).not.toContain("truncated");
+	});
+
+	it("should not truncate assistant or user messages", () => {
+		const longText = "y".repeat(5000);
+		const messages: Message[] = [
+			{
+				role: "user",
+				content: [{ type: "text", text: longText }],
+				timestamp: Date.now(),
+			},
+			{
+				role: "assistant",
+				content: [{ type: "text", text: longText }],
+				api: "anthropic",
+				provider: "anthropic",
+				model: "test",
+				usage: {
+					input: 0,
+					output: 0,
+					cacheRead: 0,
+					cacheWrite: 0,
+					totalTokens: 0,
+					cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
+				},
+				stopReason: "stop",
+				timestamp: Date.now(),
+			},
+		];
+
+		const result = serializeConversation(messages);
+
+		expect(result).not.toContain("truncated");
+		expect(result).toContain(longText);
+	});
+});