fix(coding-agent): raise branch summary output cap

closes #8845

packages/coding-agent/CHANGELOG.md

@@ -4,6 +4,7 @@
 
 ### Fixed
 
+- Fixed branch summaries failing when reasoning consumes the previous 2048-token output cap ([#8845](https://github.com/earendil-works/pi/issues/8845)).
 - Fixed the write tool reporting UTF-16 code-unit counts as byte counts by removing the misleading count ([#8979](https://github.com/earendil-works/pi/issues/8979)).
 - Fixed proxied plain-HTTP provider requests hanging after a tool call by tunneling them with CONNECT ([#8134](https://github.com/earendil-works/pi/issues/8134)).
 - Fixed RPC `abort` reporting success without cancelling an in-progress manual compaction ([#8920](https://github.com/earendil-works/pi/issues/8920)).

packages/coding-agent/docs/settings.md

@@ -133,7 +133,7 @@ Set `PI_SKIP_VERSION_CHECK=1` to disable the Pi version update check. Use `--off
 
 | Setting | Type | Default | Description |
 |---------|------|---------|-------------|
-| `branchSummary.reserveTokens` | number | `16384` | Tokens reserved for branch summarization |
+| `branchSummary.reserveTokens` | number | `16384` | Tokens reserved when selecting branch history; output is capped at 4096 tokens |
 | `branchSummary.skipPrompt` | boolean | `false` | Skip "Summarize branch?" prompt on `/tree` navigation (defaults to no summary) |
 
 ### Retry

packages/coding-agent/src/core/compaction/branch-summarization.ts

@@ -79,7 +79,7 @@ export interface GenerateBranchSummaryOptions {
 	customInstructions?: string;
 	/** If true, customInstructions replaces the default prompt instead of being appended */
 	replaceInstructions?: boolean;
-	/** Tokens reserved for prompt + LLM response (default 16384) */
+	/** Tokens reserved when selecting branch history (default 16384) */
 	reserveTokens?: number;
 	/** Optional session stream function. Used to preserve SDK request behavior without mutating agent state. */
 	streamFn?: StreamFn;
@@ -342,12 +342,14 @@ export async function generateBranchSummary(
 		},
 	];
 
+	const maxTokens = Math.min(4096, model.maxTokens > 0 ? model.maxTokens : Number.POSITIVE_INFINITY);
+
 	// Call LLM for summarization. Prefer the session stream function so SDK
 	// request behavior (timeouts, retries, attribution headers) stays consistent
 	// without running through agent state/events. Retried via completeSummarization
 	// so transient stream drops reuse the configured retry policy.
 	const context = { systemPrompt: SUMMARIZATION_SYSTEM_PROMPT, messages: summarizationMessages };
-	const requestOptions: SimpleStreamOptions = { apiKey, headers, env, signal, maxTokens: 2048 };
+	const requestOptions: SimpleStreamOptions = { apiKey, headers, env, signal, maxTokens };
 	const response = await completeSummarization(model, context, requestOptions, streamFn, retry, callbacks);
 
 	// Check if aborted or errored

packages/coding-agent/test/branch-summarization.test.ts

@@ -61,9 +61,30 @@ describe("branch summarization", () => {
 			streamFn,
 		});
 
+		expect(requestOptions?.maxTokens).toBe(4096);
 		expect(requestOptions?.toolChoice).toBeUndefined();
 	});
 
+	it("clamps the branch summary output cap to the model limit", async () => {
+		let requestOptions: SimpleStreamOptions | undefined;
+		const streamFn: StreamFn = (_model, _context, options) => {
+			requestOptions = options;
+			const stream = createAssistantMessageEventStream();
+			queueMicrotask(() =>
+				stream.push({ type: "done", reason: "stop", message: response([{ type: "text", text: "summary" }]) }),
+			);
+			return stream;
+		};
+
+		await generateBranchSummary(entries, {
+			model: { ...model, maxTokens: 1024 },
+			signal: new AbortController().signal,
+			streamFn,
+		});
+
+		expect(requestOptions?.maxTokens).toBe(1024);
+	});
+
 	it("rejects tool calls from branch summaries", async () => {
 		const streamFn: StreamFn = () => {
 			const stream = createAssistantMessageEventStream();