Fix API key priority and compaction bugs

- getEnvApiKey: ANTHROPIC_OAUTH_TOKEN now takes precedence over ANTHROPIC_API_KEY
- findCutPoint: Stop scan-backwards loop at session header (was decrementing past it causing null preparation)
- generateSummary/generateTurnPrefixSummary: Throw on stopReason=error instead of returning empty string
- Test files: Fix API key priority order, use keepRecentTokens=1 for small test conversations

packages/ai/src/models.generated.ts

@@ -6359,6 +6359,23 @@ export const MODELS = {
 			contextWindow: 128000,
 			maxTokens: 16384,
 		} satisfies Model<"openai-completions">,
+		"meta-llama/llama-3.1-70b-instruct": {
+			id: "meta-llama/llama-3.1-70b-instruct",
+			name: "Meta: Llama 3.1 70B Instruct",
+			api: "openai-completions",
+			provider: "openrouter",
+			baseUrl: "https://openrouter.ai/api/v1",
+			reasoning: false,
+			input: ["text"],
+			cost: {
+				input: 0.39999999999999997,
+				output: 0.39999999999999997,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 131072,
+			maxTokens: 4096,
+		} satisfies Model<"openai-completions">,
 		"meta-llama/llama-3.1-8b-instruct": {
 			id: "meta-llama/llama-3.1-8b-instruct",
 			name: "Meta: Llama 3.1 8B Instruct",
@@ -6393,23 +6410,6 @@ export const MODELS = {
 			contextWindow: 10000,
 			maxTokens: 4096,
 		} satisfies Model<"openai-completions">,
-		"meta-llama/llama-3.1-70b-instruct": {
-			id: "meta-llama/llama-3.1-70b-instruct",
-			name: "Meta: Llama 3.1 70B Instruct",
-			api: "openai-completions",
-			provider: "openrouter",
-			baseUrl: "https://openrouter.ai/api/v1",
-			reasoning: false,
-			input: ["text"],
-			cost: {
-				input: 0.39999999999999997,
-				output: 0.39999999999999997,
-				cacheRead: 0,
-				cacheWrite: 0,
-			},
-			contextWindow: 131072,
-			maxTokens: 4096,
-		} satisfies Model<"openai-completions">,
 		"mistralai/mistral-nemo": {
 			id: "mistralai/mistral-nemo",
 			name: "Mistral: Mistral Nemo",
@@ -6546,23 +6546,6 @@ export const MODELS = {
 			contextWindow: 128000,
 			maxTokens: 4096,
 		} satisfies Model<"openai-completions">,
-		"openai/gpt-4o-2024-05-13": {
-			id: "openai/gpt-4o-2024-05-13",
-			name: "OpenAI: GPT-4o (2024-05-13)",
-			api: "openai-completions",
-			provider: "openrouter",
-			baseUrl: "https://openrouter.ai/api/v1",
-			reasoning: false,
-			input: ["text", "image"],
-			cost: {
-				input: 5,
-				output: 15,
-				cacheRead: 0,
-				cacheWrite: 0,
-			},
-			contextWindow: 128000,
-			maxTokens: 4096,
-		} satisfies Model<"openai-completions">,
 		"openai/gpt-4o": {
 			id: "openai/gpt-4o",
 			name: "OpenAI: GPT-4o",
@@ -6597,6 +6580,23 @@ export const MODELS = {
 			contextWindow: 128000,
 			maxTokens: 64000,
 		} satisfies Model<"openai-completions">,
+		"openai/gpt-4o-2024-05-13": {
+			id: "openai/gpt-4o-2024-05-13",
+			name: "OpenAI: GPT-4o (2024-05-13)",
+			api: "openai-completions",
+			provider: "openrouter",
+			baseUrl: "https://openrouter.ai/api/v1",
+			reasoning: false,
+			input: ["text", "image"],
+			cost: {
+				input: 5,
+				output: 15,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 128000,
+			maxTokens: 4096,
+		} satisfies Model<"openai-completions">,
 		"meta-llama/llama-3-70b-instruct": {
 			id: "meta-llama/llama-3-70b-instruct",
 			name: "Meta: Llama 3 70B Instruct",
@@ -6716,23 +6716,6 @@ export const MODELS = {
 			contextWindow: 128000,
 			maxTokens: 4096,
 		} satisfies Model<"openai-completions">,
-		"openai/gpt-3.5-turbo-0613": {
-			id: "openai/gpt-3.5-turbo-0613",
-			name: "OpenAI: GPT-3.5 Turbo (older v0613)",
-			api: "openai-completions",
-			provider: "openrouter",
-			baseUrl: "https://openrouter.ai/api/v1",
-			reasoning: false,
-			input: ["text"],
-			cost: {
-				input: 1,
-				output: 2,
-				cacheRead: 0,
-				cacheWrite: 0,
-			},
-			contextWindow: 4095,
-			maxTokens: 4096,
-		} satisfies Model<"openai-completions">,
 		"openai/gpt-4-turbo-preview": {
 			id: "openai/gpt-4-turbo-preview",
 			name: "OpenAI: GPT-4 Turbo Preview",
@@ -6750,6 +6733,23 @@ export const MODELS = {
 			contextWindow: 128000,
 			maxTokens: 4096,
 		} satisfies Model<"openai-completions">,
+		"openai/gpt-3.5-turbo-0613": {
+			id: "openai/gpt-3.5-turbo-0613",
+			name: "OpenAI: GPT-3.5 Turbo (older v0613)",
+			api: "openai-completions",
+			provider: "openrouter",
+			baseUrl: "https://openrouter.ai/api/v1",
+			reasoning: false,
+			input: ["text"],
+			cost: {
+				input: 1,
+				output: 2,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 4095,
+			maxTokens: 4096,
+		} satisfies Model<"openai-completions">,
 		"mistralai/mistral-tiny": {
 			id: "mistralai/mistral-tiny",
 			name: "Mistral Tiny",

packages/ai/src/stream.ts

@@ -30,9 +30,13 @@ export function getEnvApiKey(provider: any): string | undefined {
 		return process.env.COPILOT_GITHUB_TOKEN || process.env.GH_TOKEN || process.env.GITHUB_TOKEN;
 	}
 
+	// ANTHROPIC_OAUTH_TOKEN takes precedence over ANTHROPIC_API_KEY
+	if (provider === "anthropic") {
+		return process.env.ANTHROPIC_OAUTH_TOKEN || process.env.ANTHROPIC_API_KEY;
+	}
+
 	const envMap: Record<string, string> = {
 		openai: "OPENAI_API_KEY",
-		anthropic: "ANTHROPIC_API_KEY",
 		google: "GEMINI_API_KEY",
 		groq: "GROQ_API_KEY",
 		cerebras: "CEREBRAS_API_KEY",

packages/coding-agent/examples/hooks/custom-compaction.ts

@@ -23,8 +23,15 @@ export default function (pi: HookAPI) {
 
 		ctx.ui.notify("Custom compaction hook triggered", "info");
 
-		const { messagesToSummarize, messagesToKeep, previousSummary, tokensBefore, resolveApiKey, entries, signal } =
-			event;
+		const {
+			messagesToSummarize,
+			messagesToKeep,
+			previousSummary,
+			tokensBefore,
+			resolveApiKey,
+			entries: _,
+			signal,
+		} = event;
 
 		// Use Gemini Flash for summarization (cheaper/faster than most conversation models)
 		const model = getModel("google", "gemini-2.5-flash");

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

@@ -224,7 +224,7 @@ export function findCutPoint(
 
 	// Walk backwards from newest, accumulating estimated message sizes
 	let accumulatedTokens = 0;
-	let cutIndex = startIndex; // Default: keep everything in range
+	let cutIndex = cutPoints[0]; // Default: keep from first message (not header)
 
 	for (let i = endIndex - 1; i >= startIndex; i--) {
 		const entry = entries[i];
@@ -250,8 +250,8 @@ export function findCutPoint(
 	// Scan backwards from cutIndex to include any non-message entries (bash, settings, etc.)
 	while (cutIndex > startIndex) {
 		const prevEntry = entries[cutIndex - 1];
-		// Stop at compaction boundaries
-		if (prevEntry.type === "compaction") {
+		// Stop at session header or compaction boundaries
+		if (prevEntry.type === "session" || prevEntry.type === "compaction") {
 			break;
 		}
 		if (prevEntry.type === "message") {
@@ -320,6 +320,10 @@ export async function generateSummary(
 
 	const response = await complete(model, { messages: summarizationMessages }, { maxTokens, signal, apiKey });
 
+	if (response.stopReason === "error") {
+		throw new Error(`Summarization failed: ${response.errorMessage || "Unknown error"}`);
+	}
+
 	const textContent = response.content
 		.filter((c): c is { type: "text"; text: string } => c.type === "text")
 		.map((c) => c.text)
@@ -550,6 +554,10 @@ async function generateTurnPrefixSummary(
 
 	const response = await complete(model, { messages: summarizationMessages }, { maxTokens, signal, apiKey });
 
+	if (response.stopReason === "error") {
+		throw new Error(`Turn prefix summarization failed: ${response.errorMessage || "Unknown error"}`);
+	}
+
 	return response.content
 		.filter((c): c is { type: "text"; text: string } => c.type === "text")
 		.map((c) => c.text)

packages/coding-agent/test/agent-session-branching.test.ts

@@ -20,7 +20,7 @@ import { SessionManager } from "../src/core/session-manager.js";
 import { SettingsManager } from "../src/core/settings-manager.js";
 import { codingTools } from "../src/core/tools/index.js";
 
-const API_KEY = process.env.ANTHROPIC_API_KEY || process.env.ANTHROPIC_OAUTH_TOKEN;
+const API_KEY = process.env.ANTHROPIC_OAUTH_TOKEN || process.env.ANTHROPIC_API_KEY;
 
 describe.skipIf(!API_KEY)("AgentSession branching", () => {
 	let session: AgentSession;

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

@@ -20,7 +20,7 @@ import { SessionManager } from "../src/core/session-manager.js";
 import { SettingsManager } from "../src/core/settings-manager.js";
 import { codingTools } from "../src/core/tools/index.js";
 
-const API_KEY = process.env.ANTHROPIC_API_KEY || process.env.ANTHROPIC_OAUTH_TOKEN;
+const API_KEY = process.env.ANTHROPIC_OAUTH_TOKEN || process.env.ANTHROPIC_API_KEY;
 
 describe.skipIf(!API_KEY)("AgentSession compaction e2e", () => {
 	let session: AgentSession;
@@ -46,7 +46,7 @@ describe.skipIf(!API_KEY)("AgentSession compaction e2e", () => {
 		}
 	});
 
-	function createSession() {
+	function createSession(inMemory = false) {
 		const model = getModel("anthropic", "claude-sonnet-4-5")!;
 
 		const transport = new ProviderTransport({
@@ -62,8 +62,10 @@ describe.skipIf(!API_KEY)("AgentSession compaction e2e", () => {
 			},
 		});
 
-		sessionManager = SessionManager.create(tempDir);
+		sessionManager = inMemory ? SessionManager.inMemory() : SessionManager.create(tempDir);
 		const settingsManager = SettingsManager.create(tempDir, tempDir);
+		// Use minimal keepRecentTokens so small test conversations have something to summarize
+		settingsManager.applyOverrides({ compaction: { keepRecentTokens: 1 } });
 		const authStorage = new AuthStorage(join(tempDir, "auth.json"));
 		const modelRegistry = new ModelRegistry(authStorage);
 
@@ -156,64 +158,31 @@ describe.skipIf(!API_KEY)("AgentSession compaction e2e", () => {
 		expect(compaction.type).toBe("compaction");
 		if (compaction.type === "compaction") {
 			expect(compaction.summary.length).toBeGreaterThan(0);
-			// firstKeptEntryId can be 0 if all messages fit within keepRecentTokens
-			// (which is the case for small conversations)
-			expect(compaction.firstKeptEntryId).toBeGreaterThanOrEqual(0);
+			expect(typeof compaction.firstKeptEntryId).toBe("string");
 			expect(compaction.tokensBefore).toBeGreaterThan(0);
 		}
 	}, 120000);
 
 	it("should work with --no-session mode (in-memory only)", async () => {
-		const model = getModel("anthropic", "claude-sonnet-4-5")!;
+		createSession(true); // in-memory mode
 
-		const transport = new ProviderTransport({
-			getApiKey: () => API_KEY,
-		});
+		// Send prompts
+		await session.prompt("What is 2+2? Reply with just the number.");
+		await session.agent.waitForIdle();
 
-		const agent = new Agent({
-			transport,
-			initialState: {
-				model,
-				systemPrompt: "You are a helpful assistant. Be concise.",
-				tools: codingTools,
-			},
-		});
+		await session.prompt("What is 3+3? Reply with just the number.");
+		await session.agent.waitForIdle();
 
-		// Create in-memory session manager
-		const noSessionManager = SessionManager.inMemory();
+		// Compact should work even without file persistence
+		const result = await session.compact();
 
-		const settingsManager = SettingsManager.create(tempDir, tempDir);
-		const authStorage = new AuthStorage(join(tempDir, "auth.json"));
-		const modelRegistry = new ModelRegistry(authStorage);
+		expect(result.summary).toBeDefined();
+		expect(result.summary.length).toBeGreaterThan(0);
 
-		const noSessionSession = new AgentSession({
-			agent,
-			sessionManager: noSessionManager,
-			settingsManager,
-			modelRegistry,
-		});
-
-		try {
-			// Send prompts
-			await noSessionSession.prompt("What is 2+2? Reply with just the number.");
-			await noSessionSession.agent.waitForIdle();
-
-			await noSessionSession.prompt("What is 3+3? Reply with just the number.");
-			await noSessionSession.agent.waitForIdle();
-
-			// Compact should work even without file persistence
-			const result = await noSessionSession.compact();
-
-			expect(result.summary).toBeDefined();
-			expect(result.summary.length).toBeGreaterThan(0);
-
-			// In-memory entries should have the compaction
-			const entries = noSessionManager.getEntries();
-			const compactionEntries = entries.filter((e) => e.type === "compaction");
-			expect(compactionEntries.length).toBe(1);
-		} finally {
-			noSessionSession.dispose();
-		}
+		// In-memory entries should have the compaction
+		const entries = sessionManager.getEntries();
+		const compactionEntries = entries.filter((e) => e.type === "compaction");
+		expect(compactionEntries.length).toBe(1);
 	}, 120000);
 
 	it("should emit correct events during auto-compaction", async () => {

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

@@ -16,7 +16,7 @@ import { SessionManager } from "../src/core/session-manager.js";
 import { SettingsManager } from "../src/core/settings-manager.js";
 import { codingTools } from "../src/core/tools/index.js";
 
-const API_KEY = process.env.ANTHROPIC_API_KEY || process.env.ANTHROPIC_OAUTH_TOKEN;
+const API_KEY = process.env.ANTHROPIC_OAUTH_TOKEN || process.env.ANTHROPIC_API_KEY;
 
 describe.skipIf(!API_KEY)("Compaction hooks", () => {
 	let session: AgentSession;