diff --git a/packages/agent/CHANGELOG.md b/packages/agent/CHANGELOG.md
index 0e578ff6..fdd47c8e 100644
--- a/packages/agent/CHANGELOG.md
+++ b/packages/agent/CHANGELOG.md
@@ -2,6 +2,14 @@
 
 ## [Unreleased]
 
+### Added
+
+- Added retry policy support and lifecycle events for compaction and branch-summary operations in `AgentHarness` ([#6901](https://github.com/earendil-works/pi/pull/6901) by [@davidbrai](https://github.com/davidbrai)).
+
+### Fixed
+
+- Restored the `Agent` `streamFn` option and host-configurable fallback for omitted agent-loop stream functions without reintroducing a `pi-ai/compat` dependency ([#6915](https://github.com/earendil-works/pi/issues/6915)).
+
 ## [0.81.0] - 2026-07-21
 
 ### Breaking Changes
diff --git a/packages/agent/README.md b/packages/agent/README.md
index 543f4c8f..a4d5fdd9 100644
--- a/packages/agent/README.md
+++ b/packages/agent/README.md
@@ -29,7 +29,7 @@ const agent = new Agent({
     systemPrompt: "You are a helpful assistant.",
     model,
   },
-  streamFunction: models.streamSimple.bind(models),
+  streamFn: models.streamSimple.bind(models),
 });
 
 agent.subscribe((event) => {
@@ -199,7 +199,7 @@ const agent = new Agent({
   followUpMode: "one-at-a-time",
 
   // Required stream function
-  streamFunction: models.streamSimple.bind(models),
+  streamFn: models.streamSimple.bind(models),
 
   // Session ID for provider caching
   sessionId: "session-123",
@@ -386,7 +386,7 @@ Handle custom types in `convertToLlm`:
 
 ```typescript
 const agent = new Agent({
-  streamFunction: models.streamSimple.bind(models),
+  streamFn: models.streamSimple.bind(models),
   convertToLlm: (messages) => messages.flatMap(m => {
     if (m.role === "notification") return []; // Filter out
     return [m];
@@ -457,7 +457,7 @@ For browser apps that proxy through a backend:
 import { Agent, streamProxy } from "@earendil-works/pi-agent-core";
 
 const agent = new Agent({
-  streamFunction: (model, context, options) =>
+  streamFn: (model, context, options) =>
     streamProxy(model, context, {
       ...options,
       authToken: "...",
@@ -489,13 +489,13 @@ const config: AgentLoopConfig = {
 
 const userMessage = { role: "user", content: "Hello", timestamp: Date.now() };
 
-const streamFunction = models.streamSimple.bind(models);
-for await (const event of agentLoop([userMessage], context, config, undefined, streamFunction)) {
+const streamFn = models.streamSimple.bind(models);
+for await (const event of agentLoop([userMessage], context, config, undefined, streamFn)) {
   console.log(event.type);
 }
 
 // Continue from existing context
-for await (const event of agentLoopContinue(context, config, undefined, streamFunction)) {
+for await (const event of agentLoopContinue(context, config, undefined, streamFn)) {
   console.log(event.type);
 }
 ```
diff --git a/packages/agent/src/agent-loop.ts b/packages/agent/src/agent-loop.ts
index fff06996..3a12506c 100644
--- a/packages/agent/src/agent-loop.ts
+++ b/packages/agent/src/agent-loop.ts
@@ -10,6 +10,7 @@ import {
 	type ToolResultMessage,
 	validateToolArguments,
 } from "@earendil-works/pi-ai";
+import { getDefaultStreamFn } from "./stream-fn.ts";
 import type {
 	AgentContext,
 	AgentEvent,
@@ -32,7 +33,7 @@ export function agentLoop(
 	context: AgentContext,
 	config: AgentLoopConfig,
 	signal: AbortSignal | undefined,
-	streamFunction: StreamFn,
+	streamFn: StreamFn,
 ): EventStream<AgentEvent, AgentMessage[]> {
 	const stream = createAgentStream();
 
@@ -44,7 +45,7 @@ export function agentLoop(
 			stream.push(event);
 		},
 		signal,
-		streamFunction,
+		streamFn,
 	).then((messages) => {
 		stream.end(messages);
 	});
@@ -64,7 +65,7 @@ export function agentLoopContinue(
 	context: AgentContext,
 	config: AgentLoopConfig,
 	signal: AbortSignal | undefined,
-	streamFunction: StreamFn,
+	streamFn: StreamFn,
 ): EventStream<AgentEvent, AgentMessage[]> {
 	if (context.messages.length === 0) {
 		throw new Error("Cannot continue: no messages in context");
@@ -83,7 +84,7 @@ export function agentLoopContinue(
 			stream.push(event);
 		},
 		signal,
-		streamFunction,
+		streamFn,
 	).then((messages) => {
 		stream.end(messages);
 	});
@@ -97,7 +98,7 @@ export async function runAgentLoop(
 	config: AgentLoopConfig,
 	emit: AgentEventSink,
 	signal: AbortSignal | undefined,
-	streamFunction: StreamFn,
+	streamFn: StreamFn,
 ): Promise<AgentMessage[]> {
 	const newMessages: AgentMessage[] = [...prompts];
 	const currentContext: AgentContext = {
@@ -112,7 +113,7 @@ export async function runAgentLoop(
 		await emit({ type: "message_end", message: prompt });
 	}
 
-	await runLoop(currentContext, newMessages, config, signal, emit, streamFunction);
+	await runLoop(currentContext, newMessages, config, signal, emit, streamFn ?? getDefaultStreamFn());
 	return newMessages;
 }
 
@@ -121,7 +122,7 @@ export async function runAgentLoopContinue(
 	config: AgentLoopConfig,
 	emit: AgentEventSink,
 	signal: AbortSignal | undefined,
-	streamFunction: StreamFn,
+	streamFn: StreamFn,
 ): Promise<AgentMessage[]> {
 	if (context.messages.length === 0) {
 		throw new Error("Cannot continue: no messages in context");
@@ -137,7 +138,7 @@ export async function runAgentLoopContinue(
 	await emit({ type: "agent_start" });
 	await emit({ type: "turn_start" });
 
-	await runLoop(currentContext, newMessages, config, signal, emit, streamFunction);
+	await runLoop(currentContext, newMessages, config, signal, emit, streamFn ?? getDefaultStreamFn());
 	return newMessages;
 }
 
diff --git a/packages/agent/src/agent.ts b/packages/agent/src/agent.ts
index 653adf8e..147edc83 100644
--- a/packages/agent/src/agent.ts
+++ b/packages/agent/src/agent.ts
@@ -8,6 +8,7 @@ import type {
 	Transport,
 } from "@earendil-works/pi-ai";
 import { runAgentLoop, runAgentLoopContinue } from "./agent-loop.ts";
+import { getDefaultStreamFn } from "./stream-fn.ts";
 import type {
 	AfterToolCallContext,
 	AfterToolCallResult,
@@ -97,7 +98,7 @@ export interface AgentOptions {
 	initialState?: Partial<Omit<AgentState, "pendingToolCalls" | "isStreaming" | "streamingMessage" | "errorMessage">>;
 	convertToLlm?: (messages: AgentMessage[]) => Message[] | Promise<Message[]>;
 	transformContext?: (messages: AgentMessage[], signal?: AbortSignal) => Promise<AgentMessage[]>;
-	streamFunction: StreamFn;
+	streamFn: StreamFn;
 	getApiKey?: (provider: string) => Promise<string | undefined> | string | undefined;
 	onPayload?: SimpleStreamOptions["onPayload"];
 	onResponse?: SimpleStreamOptions["onResponse"];
@@ -207,24 +208,26 @@ export class Agent {
 	public toolExecution: ToolExecutionMode;
 
 	constructor(options: AgentOptions) {
-		this._state = createMutableAgentState(options.initialState);
-		this.convertToLlm = options.convertToLlm ?? defaultConvertToLlm;
-		this.transformContext = options.transformContext;
-		this.streamFunction = options.streamFunction;
-		this.getApiKey = options.getApiKey;
-		this.onPayload = options.onPayload;
-		this.onResponse = options.onResponse;
-		this.beforeToolCall = options.beforeToolCall;
-		this.afterToolCall = options.afterToolCall;
-		this.prepareNextTurn = options.prepareNextTurn;
-		this.prepareNextTurnWithContext = options.prepareNextTurnWithContext;
-		this.steeringQueue = new PendingMessageQueue(options.steeringMode ?? "one-at-a-time");
-		this.followUpQueue = new PendingMessageQueue(options.followUpMode ?? "one-at-a-time");
-		this.sessionId = options.sessionId;
-		this.thinkingBudgets = options.thinkingBudgets;
-		this.transport = options.transport ?? "auto";
-		this.maxRetryDelayMs = options.maxRetryDelayMs;
-		this.toolExecution = options.toolExecution ?? "parallel";
+		// Older compiled consumers may omit options or streamFn even though the current API requires them.
+		const runtimeOptions: Partial<AgentOptions> = options ?? {};
+		this._state = createMutableAgentState(runtimeOptions.initialState);
+		this.convertToLlm = runtimeOptions.convertToLlm ?? defaultConvertToLlm;
+		this.transformContext = runtimeOptions.transformContext;
+		this.streamFunction = runtimeOptions.streamFn ?? getDefaultStreamFn();
+		this.getApiKey = runtimeOptions.getApiKey;
+		this.onPayload = runtimeOptions.onPayload;
+		this.onResponse = runtimeOptions.onResponse;
+		this.beforeToolCall = runtimeOptions.beforeToolCall;
+		this.afterToolCall = runtimeOptions.afterToolCall;
+		this.prepareNextTurn = runtimeOptions.prepareNextTurn;
+		this.prepareNextTurnWithContext = runtimeOptions.prepareNextTurnWithContext;
+		this.steeringQueue = new PendingMessageQueue(runtimeOptions.steeringMode ?? "one-at-a-time");
+		this.followUpQueue = new PendingMessageQueue(runtimeOptions.followUpMode ?? "one-at-a-time");
+		this.sessionId = runtimeOptions.sessionId;
+		this.thinkingBudgets = runtimeOptions.thinkingBudgets;
+		this.transport = runtimeOptions.transport ?? "auto";
+		this.maxRetryDelayMs = runtimeOptions.maxRetryDelayMs;
+		this.toolExecution = runtimeOptions.toolExecution ?? "parallel";
 	}
 
 	/**
diff --git a/packages/agent/src/index.ts b/packages/agent/src/index.ts
index b98023b0..1cd718cc 100644
--- a/packages/agent/src/index.ts
+++ b/packages/agent/src/index.ts
@@ -43,5 +43,7 @@ export * from "./harness/utils/shell-output.ts";
 export * from "./harness/utils/truncate.ts";
 // Proxy utilities
 export * from "./proxy.ts";
+// Stream defaults
+export { setDefaultStreamFn } from "./stream-fn.ts";
 // Types
 export * from "./types.ts";
diff --git a/packages/agent/src/proxy.ts b/packages/agent/src/proxy.ts
index 5d4b7a55..5f0925c9 100644
--- a/packages/agent/src/proxy.ts
+++ b/packages/agent/src/proxy.ts
@@ -84,12 +84,12 @@ export interface ProxyStreamOptions extends ProxySerializableStreamOptions {
  * The server strips the partial field from delta events to reduce bandwidth.
  * We reconstruct the partial message client-side.
  *
- * Use this as the `streamFunction` option when creating an Agent that needs to go through a proxy.
+ * Use this as the `streamFn` option when creating an Agent that needs to go through a proxy.
  *
  * @example
  * ```typescript
  * const agent = new Agent({
- *   streamFunction: (model, context, options) =>
+ *   streamFn: (model, context, options) =>
  *     streamProxy(model, context, {
  *       ...options,
  *       authToken: await getAuthToken(),
diff --git a/packages/agent/src/stream-fn.ts b/packages/agent/src/stream-fn.ts
new file mode 100644
index 00000000..337d16b3
--- /dev/null
+++ b/packages/agent/src/stream-fn.ts
@@ -0,0 +1,20 @@
+import type { StreamFn } from "./types.ts";
+
+let defaultStreamFn: StreamFn | undefined;
+
+/**
+ * Configure the fallback used by Agent and low-level loops when callers omit streamFn.
+ *
+ * Hosts that provide a default model runtime can install its stream function here
+ * without making pi-agent-core depend on a provider catalog or compatibility layer.
+ */
+export function setDefaultStreamFn(streamFn: StreamFn | undefined): void {
+	defaultStreamFn = streamFn;
+}
+
+export function getDefaultStreamFn(): StreamFn {
+	if (!defaultStreamFn) {
+		throw new Error("No default stream function configured. Pass streamFn explicitly or call setDefaultStreamFn().");
+	}
+	return defaultStreamFn;
+}
diff --git a/packages/agent/test/agent-loop.test.ts b/packages/agent/test/agent-loop.test.ts
index 47380c42..8a2a7388 100644
--- a/packages/agent/test/agent-loop.test.ts
+++ b/packages/agent/test/agent-loop.test.ts
@@ -9,6 +9,7 @@ import {
 import { Type } from "typebox";
 import { describe, expect, it } from "vitest";
 import { agentLoop, agentLoopContinue } from "../src/agent-loop.ts";
+import { setDefaultStreamFn } from "../src/index.ts";
 import type { AgentContext, AgentEvent, AgentLoopConfig, AgentMessage, AgentTool } from "../src/types.ts";
 
 // Mock stream for testing - mimics MockAssistantStream
@@ -80,6 +81,40 @@ function identityConverter(messages: AgentMessage[]): Message[] {
 	return messages.filter((m) => m.role === "user" || m.role === "assistant" || m.role === "toolResult") as Message[];
 }
 
+describe("default stream function compatibility", () => {
+	it("uses the configured default when a legacy caller omits streamFn", async () => {
+		let calls = 0;
+		setDefaultStreamFn(() => {
+			calls++;
+			const stream = new MockAssistantStream();
+			queueMicrotask(() => {
+				stream.push({
+					type: "done",
+					reason: "stop",
+					message: createAssistantMessage([{ type: "text", text: "fallback" }]),
+				});
+			});
+			return stream;
+		});
+
+		try {
+			const context: AgentContext = { systemPrompt: "", messages: [], tools: [] };
+			const config: AgentLoopConfig = { model: createModel(), convertToLlm: identityConverter };
+			const stream = Reflect.apply(agentLoop, undefined, [
+				[createUserMessage("Hello")],
+				context,
+				config,
+				undefined,
+			]) as ReturnType<typeof agentLoop>;
+
+			await stream.result();
+			expect(calls).toBe(1);
+		} finally {
+			setDefaultStreamFn(undefined);
+		}
+	});
+});
+
 describe("agentLoop with AgentMessage", () => {
 	it("should emit events with AgentMessage types", async () => {
 		const context: AgentContext = {
diff --git a/packages/agent/test/agent.test.ts b/packages/agent/test/agent.test.ts
index 6cd77f11..50d20480 100644
--- a/packages/agent/test/agent.test.ts
+++ b/packages/agent/test/agent.test.ts
@@ -1,7 +1,14 @@
 import { type AssistantMessage, type AssistantMessageEvent, EventStream, getModel } from "@earendil-works/pi-ai/compat";
 import { Type } from "typebox";
 import { describe, expect, it } from "vitest";
-import { Agent, type AgentEvent, type AgentTool, type AgentToolUpdateCallback, type StreamFn } from "../src/index.ts";
+import {
+	Agent,
+	type AgentEvent,
+	type AgentTool,
+	type AgentToolUpdateCallback,
+	type StreamFn,
+	setDefaultStreamFn,
+} from "../src/index.ts";
 
 // Mock stream that mimics AssistantMessageEventStream
 class MockAssistantStream extends EventStream<AssistantMessageEvent, AssistantMessage> {
@@ -75,8 +82,29 @@ function createDeferred(): {
 }
 
 describe("Agent", () => {
+	it("uses the configured default when a legacy caller omits streamFn", async () => {
+		let calls = 0;
+		setDefaultStreamFn(() => {
+			calls++;
+			const stream = new MockAssistantStream();
+			queueMicrotask(() => {
+				const message = createAssistantMessage("fallback");
+				stream.push({ type: "done", reason: "stop", message });
+			});
+			return stream;
+		});
+
+		try {
+			const agent = Reflect.construct(Agent, [{}]) as Agent;
+			await agent.prompt("Hello");
+			expect(calls).toBe(1);
+		} finally {
+			setDefaultStreamFn(undefined);
+		}
+	});
+
 	it("should create an agent instance with default state", () => {
-		const agent = new Agent({ streamFunction: unusedStreamFunction });
+		const agent = new Agent({ streamFn: unusedStreamFunction });
 
 		expect(agent.state).toBeDefined();
 		expect(agent.state.systemPrompt).toBe("");
@@ -93,7 +121,7 @@ describe("Agent", () => {
 	it("should create an agent instance with custom initial state", () => {
 		const customModel = getModel("openai", "gpt-4o-mini");
 		const agent = new Agent({
-			streamFunction: unusedStreamFunction,
+			streamFn: unusedStreamFunction,
 			initialState: {
 				systemPrompt: "You are a helpful assistant.",
 				model: customModel,
@@ -107,7 +135,7 @@ describe("Agent", () => {
 	});
 
 	it("should subscribe to events", () => {
-		const agent = new Agent({ streamFunction: unusedStreamFunction });
+		const agent = new Agent({ streamFn: unusedStreamFunction });
 
 		let eventCount = 0;
 		const unsubscribe = agent.subscribe((_event) => {
@@ -130,7 +158,7 @@ describe("Agent", () => {
 
 	it("emits full lifecycle events for thrown run failures", async () => {
 		const agent = new Agent({
-			streamFunction: () => {
+			streamFn: () => {
 				throw new Error("provider exploded");
 			},
 		});
@@ -162,7 +190,7 @@ describe("Agent", () => {
 	it("should await async subscribers before prompt resolves", async () => {
 		const barrier = createDeferred();
 		const agent = new Agent({
-			streamFunction: () => {
+			streamFn: () => {
 				const stream = new MockAssistantStream();
 				queueMicrotask(() => {
 					stream.push({ type: "done", reason: "stop", message: createAssistantMessage("ok") });
@@ -200,7 +228,7 @@ describe("Agent", () => {
 	it("waitForIdle should wait for async subscribers", async () => {
 		const barrier = createDeferred();
 		const agent = new Agent({
-			streamFunction: () => {
+			streamFn: () => {
 				const stream = new MockAssistantStream();
 				queueMicrotask(() => {
 					stream.push({ type: "done", reason: "stop", message: createAssistantMessage("ok") });
@@ -235,7 +263,7 @@ describe("Agent", () => {
 	it("should pass the active abort signal to subscribers", async () => {
 		let receivedSignal: AbortSignal | undefined;
 		const agent = new Agent({
-			streamFunction: (_model, _context, options) => {
+			streamFn: (_model, _context, options) => {
 				const stream = new MockAssistantStream();
 				queueMicrotask(() => {
 					stream.push({ type: "start", partial: createAssistantMessage("") });
@@ -298,7 +326,7 @@ describe("Agent", () => {
 		};
 		const agent = new Agent({
 			initialState: { tools: [tool] },
-			streamFunction: () => {
+			streamFn: () => {
 				const stream = new MockAssistantStream();
 				queueMicrotask(() => {
 					stream.push({
@@ -373,7 +401,7 @@ describe("Agent", () => {
 		};
 		const agent = new Agent({
 			initialState: { tools: [settledTool, slowTool] },
-			streamFunction: () => {
+			streamFn: () => {
 				const stream = new MockAssistantStream();
 				queueMicrotask(() => {
 					stream.push({
@@ -412,7 +440,7 @@ describe("Agent", () => {
 	});
 
 	it("should update state with mutators", () => {
-		const agent = new Agent({ streamFunction: unusedStreamFunction });
+		const agent = new Agent({ streamFn: unusedStreamFunction });
 
 		// Test setSystemPrompt
 		agent.state.systemPrompt = "Custom prompt";
@@ -451,7 +479,7 @@ describe("Agent", () => {
 	});
 
 	it("should support steering message queue", async () => {
-		const agent = new Agent({ streamFunction: unusedStreamFunction });
+		const agent = new Agent({ streamFn: unusedStreamFunction });
 
 		const message = { role: "user" as const, content: "Steering message", timestamp: Date.now() };
 		agent.steer(message);
@@ -461,7 +489,7 @@ describe("Agent", () => {
 	});
 
 	it("should support follow-up message queue", async () => {
-		const agent = new Agent({ streamFunction: unusedStreamFunction });
+		const agent = new Agent({ streamFn: unusedStreamFunction });
 
 		const message = { role: "user" as const, content: "Follow-up message", timestamp: Date.now() };
 		agent.followUp(message);
@@ -471,7 +499,7 @@ describe("Agent", () => {
 	});
 
 	it("should handle abort controller", () => {
-		const agent = new Agent({ streamFunction: unusedStreamFunction });
+		const agent = new Agent({ streamFn: unusedStreamFunction });
 
 		// Should not throw even if nothing is running
 		expect(() => agent.abort()).not.toThrow();
@@ -481,7 +509,7 @@ describe("Agent", () => {
 		let abortSignal: AbortSignal | undefined;
 		const agent = new Agent({
 			// Use a stream function that responds to abort
-			streamFunction: (_model, _context, options) => {
+			streamFn: (_model, _context, options) => {
 				abortSignal = options?.signal;
 				const stream = new MockAssistantStream();
 				queueMicrotask(() => {
@@ -520,7 +548,7 @@ describe("Agent", () => {
 	it("should throw when continue() called while streaming", async () => {
 		let abortSignal: AbortSignal | undefined;
 		const agent = new Agent({
-			streamFunction: (_model, _context, options) => {
+			streamFn: (_model, _context, options) => {
 				abortSignal = options?.signal;
 				const stream = new MockAssistantStream();
 				queueMicrotask(() => {
@@ -555,7 +583,7 @@ describe("Agent", () => {
 
 	it("continue() should process queued follow-up messages after an assistant turn", async () => {
 		const agent = new Agent({
-			streamFunction: () => {
+			streamFn: () => {
 				const stream = new MockAssistantStream();
 				queueMicrotask(() => {
 					stream.push({ type: "done", reason: "stop", message: createAssistantMessage("Processed") });
@@ -594,7 +622,7 @@ describe("Agent", () => {
 	it("continue() should keep one-at-a-time steering semantics from assistant tail", async () => {
 		let responseCount = 0;
 		const agent = new Agent({
-			streamFunction: () => {
+			streamFn: () => {
 				const stream = new MockAssistantStream();
 				responseCount++;
 				queueMicrotask(() => {
@@ -652,7 +680,7 @@ describe("Agent", () => {
 				sawAbortSignal = signal instanceof AbortSignal;
 				return undefined;
 			},
-			streamFunction: () => {
+			streamFn: () => {
 				requestCount++;
 				const stream = new MockAssistantStream();
 				queueMicrotask(() => {
@@ -680,7 +708,7 @@ describe("Agent", () => {
 		let receivedSessionId: string | undefined;
 		const agent = new Agent({
 			sessionId: "session-abc",
-			streamFunction: (_model, _context, options) => {
+			streamFn: (_model, _context, options) => {
 				receivedSessionId = options?.sessionId;
 				const stream = new MockAssistantStream();
 				queueMicrotask(() => {
diff --git a/packages/agent/test/e2e.test.ts b/packages/agent/test/e2e.test.ts
index 484ef079..6b27c744 100644
--- a/packages/agent/test/e2e.test.ts
+++ b/packages/agent/test/e2e.test.ts
@@ -38,7 +38,7 @@ afterEach(() => {
 
 async function basicPrompt(model: Model<string>) {
 	const agent = new Agent({
-		streamFunction: streamSimple,
+		streamFn: streamSimple,
 		initialState: {
 			systemPrompt: "You are a helpful assistant. Keep your responses concise.",
 			model,
@@ -61,7 +61,7 @@ async function basicPrompt(model: Model<string>) {
 
 async function toolExecution(model: Model<string>) {
 	const agent = new Agent({
-		streamFunction: streamSimple,
+		streamFn: streamSimple,
 		initialState: {
 			systemPrompt: "You are a helpful assistant. Always use the calculator tool for math.",
 			model,
@@ -101,7 +101,7 @@ async function toolExecution(model: Model<string>) {
 
 async function abortExecution(model: Model<string>) {
 	const agent = new Agent({
-		streamFunction: streamSimple,
+		streamFn: streamSimple,
 		initialState: {
 			systemPrompt: "You are a helpful assistant.",
 			model,
@@ -129,7 +129,7 @@ async function abortExecution(model: Model<string>) {
 
 async function stateUpdates(model: Model<string>) {
 	const agent = new Agent({
-		streamFunction: streamSimple,
+		streamFn: streamSimple,
 		initialState: {
 			systemPrompt: "You are a helpful assistant.",
 			model,
@@ -162,7 +162,7 @@ async function stateUpdates(model: Model<string>) {
 
 async function multiTurnConversation(model: Model<string>) {
 	const agent = new Agent({
-		streamFunction: streamSimple,
+		streamFn: streamSimple,
 		initialState: {
 			systemPrompt: "You are a helpful assistant.",
 			model,
@@ -244,7 +244,7 @@ describe("Agent integration with faux provider", () => {
 		faux.setResponses([fauxAssistantMessage([fauxThinking("step by step"), fauxText("4")])]);
 
 		const agent = new Agent({
-			streamFunction: streamSimple,
+			streamFn: streamSimple,
 			initialState: {
 				systemPrompt: "You are a helpful assistant.",
 				model: faux.getModel(),
@@ -269,7 +269,7 @@ describe("Agent.continue() with faux provider", () => {
 		it("throws when no messages in context", async () => {
 			const faux = createFauxRegistration();
 			const agent = new Agent({
-				streamFunction: streamSimple,
+				streamFn: streamSimple,
 				initialState: {
 					systemPrompt: "Test",
 					model: faux.getModel(),
@@ -283,7 +283,7 @@ describe("Agent.continue() with faux provider", () => {
 			const faux = createFauxRegistration();
 			const model = faux.getModel();
 			const agent = new Agent({
-				streamFunction: streamSimple,
+				streamFn: streamSimple,
 				initialState: {
 					systemPrompt: "Test",
 					model,
@@ -318,7 +318,7 @@ describe("Agent.continue() with faux provider", () => {
 			const faux = createFauxRegistration();
 			faux.setResponses([fauxAssistantMessage("HELLO WORLD")]);
 			const agent = new Agent({
-				streamFunction: streamSimple,
+				streamFn: streamSimple,
 				initialState: {
 					systemPrompt: "You are a helpful assistant. Follow instructions exactly.",
 					model: faux.getModel(),
@@ -353,7 +353,7 @@ describe("Agent.continue() with faux provider", () => {
 			const model = faux.getModel();
 			faux.setResponses([fauxAssistantMessage("The answer is 8.")]);
 			const agent = new Agent({
-				streamFunction: streamSimple,
+				streamFn: streamSimple,
 				initialState: {
 					systemPrompt:
 						"You are a helpful assistant. After getting a calculation result, state the answer clearly.",
diff --git a/packages/ai/CHANGELOG.md b/packages/ai/CHANGELOG.md
index c559aed0..1867078b 100644
--- a/packages/ai/CHANGELOG.md
+++ b/packages/ai/CHANGELOG.md
@@ -2,6 +2,14 @@
 
 ## [Unreleased]
 
+### Added
+
+- Added `retryAssistantCall()` for bounded retries of transient assistant failures with lifecycle callbacks and abort handling ([#6901](https://github.com/earendil-works/pi/pull/6901) by [@davidbrai](https://github.com/davidbrai)).
+
+### Fixed
+
+- Fixed Kimi K3 models from Moonshot AI and Moonshot AI China to use the OpenAI thinking format and expose reasoning effort support.
+
 ## [0.81.0] - 2026-07-21
 
 ### Added
diff --git a/packages/coding-agent/CHANGELOG.md b/packages/coding-agent/CHANGELOG.md
index fe7d5cf0..1572d6e4 100644
--- a/packages/coding-agent/CHANGELOG.md
+++ b/packages/coding-agent/CHANGELOG.md
@@ -2,9 +2,21 @@
 
 ## [Unreleased]
 
+### New Features
+
+- **Verifiable release source archives** — GitHub releases now include deterministic, checksummed source archives with instructions for rebuilding standalone binaries. See [Building standalone binaries from release source](../../README.md#building-standalone-binaries-from-release-source).
+- **Resilient compaction and branch summaries** — Transient provider failures now follow the configured retry policy, with retry lifecycle events available to interactive, JSON, RPC, and SDK consumers. See [Compaction & Branch Summarization](docs/compaction.md) and [RPC retry events](docs/rpc.md#summarization_retry_scheduled--summarization_retry_attempt_start--summarization_retry_finished).
+
+### Added
+
+- Added deterministic, checksummed source archives to GitHub releases with documented standalone binary rebuild instructions ([#6913](https://github.com/earendil-works/pi/pull/6913) by [@christianklotz](https://github.com/christianklotz)).
+
 ### Fixed
 
+- Fixed compaction and branch summarization to retry transient provider failures using the configured retry policy, with retry lifecycle events exposed to interactive, JSON, RPC, and SDK consumers ([#6901](https://github.com/earendil-works/pi/pull/6901) by [@davidbrai](https://github.com/davidbrai)).
 - Fixed interactive startup waiting for background model catalog refresh while computing the footer provider count.
+- Restored the default stream fallback for extensions using the pre-0.81 agent-core API ([#6915](https://github.com/earendil-works/pi/issues/6915)).
+- Fixed inherited Kimi K3 models from Moonshot AI and Moonshot AI China to use the OpenAI thinking format and expose reasoning effort support.
 
 ## [0.81.0] - 2026-07-21
 
diff --git a/packages/coding-agent/src/core/sdk.ts b/packages/coding-agent/src/core/sdk.ts
index 09138f9f..7f662d33 100644
--- a/packages/coding-agent/src/core/sdk.ts
+++ b/packages/coding-agent/src/core/sdk.ts
@@ -1,6 +1,6 @@
 import { join } from "node:path";
-import { Agent, type AgentMessage, type ThinkingLevel } from "@earendil-works/pi-agent-core";
-import { clampThinkingLevel, type Message, type Model } from "@earendil-works/pi-ai/compat";
+import { Agent, type AgentMessage, setDefaultStreamFn, type ThinkingLevel } from "@earendil-works/pi-agent-core";
+import { clampThinkingLevel, type Message, type Model, streamSimple } from "@earendil-works/pi-ai/compat";
 import { getAgentDir } from "../config.ts";
 import { resolvePath } from "../utils/paths.ts";
 import { AgentSession } from "./agent-session.ts";
@@ -30,6 +30,11 @@ import {
 	withFileMutationQueue,
 } from "./tools/index.ts";
 
+// Preserve the pre-0.81 fallback for extensions that construct Agent instances
+// or invoke low-level agent loops without supplying streamFn. Agent core remains
+// provider-agnostic and does not import pi-ai/compat itself.
+setDefaultStreamFn(streamSimple);
+
 export interface CreateAgentSessionOptions {
 	/** Working directory for project-local discovery. Default: process.cwd() */
 	cwd?: string;
@@ -294,7 +299,7 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 			tools: [],
 		},
 		convertToLlm: convertToLlmWithBlockImages,
-		streamFunction: async (model, context, options) => {
+		streamFn: async (model, context, options) => {
 			const providerRetrySettings = settingsManager.getProviderRetrySettings();
 			const httpIdleTimeoutMs = settingsManager.getHttpIdleTimeoutMs();
 			// SDKs treat timeout=0 as 0ms (immediate timeout), not "no timeout".
diff --git a/packages/coding-agent/test/agent-session-auto-compaction-queue.test.ts b/packages/coding-agent/test/agent-session-auto-compaction-queue.test.ts
index 3b12b1ef..8c1769dc 100644
--- a/packages/coding-agent/test/agent-session-auto-compaction-queue.test.ts
+++ b/packages/coding-agent/test/agent-session-auto-compaction-queue.test.ts
@@ -24,7 +24,7 @@ describe("AgentSession auto-compaction queue resume", () => {
 
 		const model = getModel("anthropic", "claude-sonnet-4-5")!;
 		const agent = new Agent({
-			streamFunction: streamSimple,
+			streamFn: streamSimple,
 			initialState: {
 				model,
 				systemPrompt: "Test",
diff --git a/packages/coding-agent/test/agent-session-compaction.test.ts b/packages/coding-agent/test/agent-session-compaction.test.ts
index e35158f8..6484aea0 100644
--- a/packages/coding-agent/test/agent-session-compaction.test.ts
+++ b/packages/coding-agent/test/agent-session-compaction.test.ts
@@ -49,7 +49,7 @@ describe.skipIf(!API_KEY)("AgentSession compaction e2e", () => {
 		const model = getModel("anthropic", "claude-sonnet-4-5")!;
 		const agent = new Agent({
 			getApiKey: () => API_KEY,
-			streamFunction: streamSimple,
+			streamFn: streamSimple,
 			initialState: {
 				model,
 				systemPrompt: "You are a helpful assistant. Be concise.",
diff --git a/packages/coding-agent/test/agent-session-concurrent.test.ts b/packages/coding-agent/test/agent-session-concurrent.test.ts
index bd3ef463..bd825ec7 100644
--- a/packages/coding-agent/test/agent-session-concurrent.test.ts
+++ b/packages/coding-agent/test/agent-session-concurrent.test.ts
@@ -90,7 +90,7 @@ describe("AgentSession concurrent prompt guard", () => {
 				systemPrompt: "Test",
 				tools: [],
 			},
-			streamFunction: (_model, _context, options) => {
+			streamFn: (_model, _context, options) => {
 				abortSignal = options?.signal;
 				const stream = new MockAssistantStream();
 				queueMicrotask(() => {
@@ -195,7 +195,7 @@ describe("AgentSession concurrent prompt guard", () => {
 				systemPrompt: "Test",
 				tools: [],
 			},
-			streamFunction: (_model, context, options) => {
+			streamFn: (_model, context, options) => {
 				abortSignal = options?.signal;
 				const stream = new MockAssistantStream();
 				queueMicrotask(() => {
@@ -301,7 +301,7 @@ describe("AgentSession concurrent prompt guard", () => {
 				systemPrompt: "Test",
 				tools: [],
 			},
-			streamFunction: () => {
+			streamFn: () => {
 				const stream = new MockAssistantStream();
 				queueMicrotask(() => {
 					stream.push({ type: "start", partial: createAssistantMessage("") });
@@ -362,7 +362,7 @@ describe("AgentSession concurrent prompt guard", () => {
 				systemPrompt: "Test",
 				tools: [tool],
 			},
-			streamFunction: async (_model, context) => {
+			streamFn: async (_model, context) => {
 				const stream = new MockAssistantStream();
 				queueMicrotask(() => {
 					const toolResultCount = context.messages.filter((message) => message.role === "toolResult").length;
@@ -508,7 +508,7 @@ describe("AgentSession concurrent prompt guard", () => {
 				systemPrompt: "Test",
 				tools: [tool],
 			},
-			streamFunction: async (_model, context) => {
+			streamFn: async (_model, context) => {
 				const stream = new MockAssistantStream();
 				queueMicrotask(() => {
 					const hasToolResult = context.messages.some((message) => message.role === "toolResult");
diff --git a/packages/coding-agent/test/agent-session-retry.test.ts b/packages/coding-agent/test/agent-session-retry.test.ts
index f61e71cb..25a996ca 100644
--- a/packages/coding-agent/test/agent-session-retry.test.ts
+++ b/packages/coding-agent/test/agent-session-retry.test.ts
@@ -82,7 +82,7 @@ describe("AgentSession retry", () => {
 		const agent = new Agent({
 			getApiKey: () => "test-key",
 			initialState: { model, systemPrompt: "Test", tools: [] },
-			streamFunction: () => {
+			streamFn: () => {
 				callCount++;
 				const stream = new MockAssistantStream();
 				queueMicrotask(() => {
@@ -203,7 +203,7 @@ describe("AgentSession retry", () => {
 		const agent = new Agent({
 			getApiKey: () => "test-key",
 			initialState: { model, systemPrompt: "Test", tools: [] },
-			streamFunction: streamFn,
+			streamFn: streamFn,
 		});
 		const sessionManager = SessionManager.inMemory();
 		const settingsManager = SettingsManager.create(tempDir, tempDir);
@@ -255,7 +255,7 @@ describe("AgentSession retry", () => {
 		const agent = new Agent({
 			getApiKey: () => "test-key",
 			initialState: { model, systemPrompt: "Test", tools: [] },
-			streamFunction: () => {
+			streamFn: () => {
 				callCount++;
 				const stream = new MockAssistantStream();
 				queueMicrotask(() => {
diff --git a/packages/coding-agent/test/agent-session-stats.test.ts b/packages/coding-agent/test/agent-session-stats.test.ts
index 37706c56..434524d4 100644
--- a/packages/coding-agent/test/agent-session-stats.test.ts
+++ b/packages/coding-agent/test/agent-session-stats.test.ts
@@ -75,7 +75,7 @@ async function createSession() {
 	const session = new AgentSession({
 		agent: new Agent({
 			getApiKey: () => "test-key",
-			streamFunction: streamSimple,
+			streamFn: streamSimple,
 			initialState: {
 				model,
 				systemPrompt: "You are a helpful assistant.",
diff --git a/packages/coding-agent/test/compaction-extensions.test.ts b/packages/coding-agent/test/compaction-extensions.test.ts
index 053be534..7b8f8e62 100644
--- a/packages/coding-agent/test/compaction-extensions.test.ts
+++ b/packages/coding-agent/test/compaction-extensions.test.ts
@@ -89,7 +89,7 @@ describe.skipIf(!API_KEY)("Compaction extensions", () => {
 		const model = getModel("anthropic", "claude-sonnet-4-5")!;
 		const agent = new Agent({
 			getApiKey: () => API_KEY,
-			streamFunction: streamSimple,
+			streamFn: streamSimple,
 			initialState: {
 				model,
 				systemPrompt: "You are a helpful assistant. Be concise.",
diff --git a/packages/coding-agent/test/rpc-prompt-response-semantics.test.ts b/packages/coding-agent/test/rpc-prompt-response-semantics.test.ts
index 8279d57c..155d2d38 100644
--- a/packages/coding-agent/test/rpc-prompt-response-semantics.test.ts
+++ b/packages/coding-agent/test/rpc-prompt-response-semantics.test.ts
@@ -114,7 +114,7 @@ async function createRuntimeHost(options: { withAuth: boolean; responseDelayMs:
 			systemPrompt: "Test",
 			tools: [],
 		},
-		streamFunction: (_model, _context, _options) => {
+		streamFn: (_model, _context, _options) => {
 			const stream = new MockAssistantStream();
 			queueMicrotask(() => {
 				stream.push({ type: "start", partial: createAssistantMessage("") });
diff --git a/packages/coding-agent/test/suite/harness.ts b/packages/coding-agent/test/suite/harness.ts
index 95c5dbba..5f003241 100644
--- a/packages/coding-agent/test/suite/harness.ts
+++ b/packages/coding-agent/test/suite/harness.ts
@@ -137,7 +137,7 @@ export async function createHarness(options: HarnessOptions = {}): Promise<Harne
 
 	const agent = new Agent({
 		getApiKey: () => (withConfiguredAuth ? "faux-key" : undefined),
-		streamFunction: streamSimple,
+		streamFn: streamSimple,
 		initialState: {
 			model,
 			systemPrompt: options.systemPrompt ?? "You are a test assistant.",
diff --git a/packages/coding-agent/test/suite/regressions/5596-missing-theme-export.test.ts b/packages/coding-agent/test/suite/regressions/5596-missing-theme-export.test.ts
index b6efaa64..3e78c034 100644
--- a/packages/coding-agent/test/suite/regressions/5596-missing-theme-export.test.ts
+++ b/packages/coding-agent/test/suite/regressions/5596-missing-theme-export.test.ts
@@ -61,7 +61,7 @@ describe("regression #5596: missing configured theme export", () => {
 				tools: [],
 			},
 			convertToLlm,
-			streamFunction: streamSimple,
+			streamFn: streamSimple,
 		});
 		const session = new AgentSession({
 			agent,
diff --git a/packages/coding-agent/test/test-harness.ts b/packages/coding-agent/test/test-harness.ts
index 70fbf30f..89a94411 100644
--- a/packages/coding-agent/test/test-harness.ts
+++ b/packages/coding-agent/test/test-harness.ts
@@ -378,7 +378,7 @@ async function createHarnessWithResourceLoader(
 			systemPrompt: options.systemPrompt ?? "You are a test assistant.",
 			tools: options.tools ?? [],
 		},
-		streamFunction: streamFn,
+		streamFn: streamFn,
 	});
 
 	const sessionManager = SessionManager.inMemory();
diff --git a/packages/coding-agent/test/utilities.ts b/packages/coding-agent/test/utilities.ts
index e2484505..b14fed1e 100644
--- a/packages/coding-agent/test/utilities.ts
+++ b/packages/coding-agent/test/utilities.ts
@@ -246,7 +246,7 @@ export async function createTestSession(options: TestSessionOptions = {}): Promi
 			systemPrompt: options.systemPrompt ?? "You are a helpful assistant. Be extremely concise.",
 			tools: createCodingTools(process.cwd()),
 		},
-		streamFunction: streamSimple,
+		streamFn: streamSimple,
 	});
 
 	const sessionManager = options.inMemory ? SessionManager.inMemory() : SessionManager.create(tempDir);
diff --git a/scripts/agent-treeshake-smoke-entry.ts b/scripts/agent-treeshake-smoke-entry.ts
index 8a8206ec..654a0336 100644
--- a/scripts/agent-treeshake-smoke-entry.ts
+++ b/scripts/agent-treeshake-smoke-entry.ts
@@ -9,5 +9,5 @@ if (!model) throw new Error("Anthropic smoke-test model not found");
 
 export const agent = new Agent({
 	initialState: { model },
-	streamFunction: models.streamSimple.bind(models),
+	streamFn: models.streamSimple.bind(models),
 });
diff --git a/scripts/browser-smoke-entry.ts b/scripts/browser-smoke-entry.ts
index 2a7ccfbe..31dada99 100644
--- a/scripts/browser-smoke-entry.ts
+++ b/scripts/browser-smoke-entry.ts
@@ -24,7 +24,7 @@ const model = getModel("google", "gemini-2.5-flash");
 const schema = Type.Object({ prompt: Type.String() });
 const stream = createAssistantMessageEventStream();
 
-const agent = new Agent({ initialState: { model }, streamFunction: streamSimple });
+const agent = new Agent({ initialState: { model }, streamFn: streamSimple });
 agent.steer({ role: "user", content: [{ type: "text", text: "queued" }], timestamp: 0 });
 const repo = new InMemorySessionRepo();
 const result = getOrThrow(ok({ value: 1 }));