fix(coding-agent): prepare v0.3.3 release

packages/agent/CHANGELOG.md

@@ -2,6 +2,15 @@
 
 ## [Unreleased]
 
+### Added
+
+- Added explicit runtime telemetry contexts and nested run, turn, request, and tool spans ([#95](https://github.com/diffexai/diffex/pull/95)).
+
+### Changed
+
+- Updated runtime dependencies, including file-ignore matching, schema handling, YAML parsing, and diff generation ([#272](https://github.com/diffexai/diffex/pull/272), [#274](https://github.com/diffexai/diffex/pull/274)).
+- Changed package licensing to MIT ([#280](https://github.com/diffexai/diffex/pull/280)).
+
 ## [0.3.2] - 2026-09-29
 
 ### Added

packages/ai/CHANGELOG.md

@@ -2,6 +2,24 @@
 
 ## [Unreleased]
 
+### Added
+
+- Added GPT-6 Astra, Luna, Sol, and GPT-6.1 Sol to the Codex catalog for ChatGPT authentication.
+- Added `isRetryableProviderErrorMessage` for classifying transient provider and transport failures ([#161](https://github.com/diffexai/diffex/pull/161)).
+
+### Changed
+
+- Updated provider SDKs and runtime dependencies ([#272](https://github.com/diffexai/diffex/pull/272), [#273](https://github.com/diffexai/diffex/pull/273), [#274](https://github.com/diffexai/diffex/pull/274)).
+- Anthropic requests now omit unsupported sampling parameters, including the unified `temperature` option. Google Gemini and Vertex AI disable automatic SDK continuations; Anthropic context-window limits and Google continuation limits map to `length`.
+  See [Provider SDK migration notes](README.md#provider-sdk-migration-notes) ([#273](https://github.com/diffexai/diffex/pull/273)).
+- Changed package licensing to MIT ([#280](https://github.com/diffexai/diffex/pull/280)).
+
+### Fixed
+
+- Fixed cancelled authentication and lazy provider setup being reported as errors instead of `aborted` responses.
+- Preserved Kimi Coding catalog generation after the upstream source key changed.
+- Fixed Google model streaming through OpenCode Go by registering its Google API implementation ([#265](https://github.com/diffexai/diffex/pull/265)).
+
 ## [0.3.2] - 2026-09-29
 
 ## [0.3.1] - 2026-09-20

packages/ai/src/image-models.generated.ts

@@ -5,6 +5,36 @@ import type { ImagesApi, ImagesModel } from "./types.ts";
 
 export const IMAGE_MODELS = {
 	openrouter: {
+		"black-forest-labs/flux-3-image": {
+			id: "black-forest-labs/flux-3-image",
+			name: "Black Forest Labs: FLUX 3 Image",
+			api: "openrouter-images",
+			provider: "openrouter",
+			baseUrl: "https://openrouter.ai/api/v1",
+			input: ["text", "image"],
+			output: ["image"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+		} satisfies ImagesModel<"openrouter-images">,
+		"black-forest-labs/flux-3-image-klein": {
+			id: "black-forest-labs/flux-3-image-klein",
+			name: "Black Forest Labs: FLUX 3 Image Klein",
+			api: "openrouter-images",
+			provider: "openrouter",
+			baseUrl: "https://openrouter.ai/api/v1",
+			input: ["text", "image"],
+			output: ["image"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+		} satisfies ImagesModel<"openrouter-images">,
 		"black-forest-labs/flux.2-flex": {
 			id: "black-forest-labs/flux.2-flex",
 			name: "Black Forest Labs: FLUX.2 Flex",
@@ -80,6 +110,21 @@ export const IMAGE_MODELS = {
 				cacheWrite: 0,
 			},
 		} satisfies ImagesModel<"openrouter-images">,
+		"bytedance-seed/seedream-5-0-flash": {
+			id: "bytedance-seed/seedream-5-0-flash",
+			name: "ByteDance Seed: Seedream 5.0 Flash",
+			api: "openrouter-images",
+			provider: "openrouter",
+			baseUrl: "https://openrouter.ai/api/v1",
+			input: ["text", "image"],
+			output: ["image"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+		} satisfies ImagesModel<"openrouter-images">,
 		"bytedance-seed/seedream-5-0-lite": {
 			id: "bytedance-seed/seedream-5-0-lite",
 			name: "ByteDance Seed: Seedream 5.0 Lite",
@@ -200,6 +245,36 @@ export const IMAGE_MODELS = {
 				cacheWrite: 0,
 			},
 		} satisfies ImagesModel<"openrouter-images">,
+		"google/gemini-nano-banana-2.1": {
+			id: "google/gemini-nano-banana-2.1",
+			name: "Google: Nano Banana 2.1",
+			api: "openrouter-images",
+			provider: "openrouter",
+			baseUrl: "https://openrouter.ai/api/v1",
+			input: ["image", "text"],
+			output: ["image", "text"],
+			cost: {
+				input: 1.5,
+				output: 7.5,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+		} satisfies ImagesModel<"openrouter-images">,
+		"krea/k3-edit": {
+			id: "krea/k3-edit",
+			name: "Krea: K3 Edit",
+			api: "openrouter-images",
+			provider: "openrouter",
+			baseUrl: "https://openrouter.ai/api/v1",
+			input: ["text", "image"],
+			output: ["image"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+		} satisfies ImagesModel<"openrouter-images">,
 		"krea/krea-2-large": {
 			id: "krea/krea-2-large",
 			name: "Krea: Krea 2 Large",
@@ -470,6 +545,36 @@ export const IMAGE_MODELS = {
 				cacheWrite: 0,
 			},
 		} satisfies ImagesModel<"openrouter-images">,
+		"qwen/qwen-image-2.1-pro": {
+			id: "qwen/qwen-image-2.1-pro",
+			name: "Qwen: Qwen Image 2.1 Pro",
+			api: "openrouter-images",
+			provider: "openrouter",
+			baseUrl: "https://openrouter.ai/api/v1",
+			input: ["text", "image"],
+			output: ["image"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+		} satisfies ImagesModel<"openrouter-images">,
+		"qwen/qwen-image-2.1-turbo": {
+			id: "qwen/qwen-image-2.1-turbo",
+			name: "Qwen: Qwen Image 2.1 Turbo",
+			api: "openrouter-images",
+			provider: "openrouter",
+			baseUrl: "https://openrouter.ai/api/v1",
+			input: ["text", "image"],
+			output: ["image"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+		} satisfies ImagesModel<"openrouter-images">,
 		"qwen/qwen-image-3": {
 			id: "qwen/qwen-image-3",
 			name: "Qwen: Qwen Image 3",
@@ -650,6 +755,21 @@ export const IMAGE_MODELS = {
 				cacheWrite: 0,
 			},
 		} satisfies ImagesModel<"openrouter-images">,
+		"recraft/recraft-v4.1-flash": {
+			id: "recraft/recraft-v4.1-flash",
+			name: "Recraft: Recraft V4.1 Flash",
+			api: "openrouter-images",
+			provider: "openrouter",
+			baseUrl: "https://openrouter.ai/api/v1",
+			input: ["text"],
+			output: ["image"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+		} satisfies ImagesModel<"openrouter-images">,
 		"recraft/recraft-v4.1-pro": {
 			id: "recraft/recraft-v4.1-pro",
 			name: "Recraft: Recraft V4.1 Pro",
@@ -785,6 +905,21 @@ export const IMAGE_MODELS = {
 				cacheWrite: 0,
 			},
 		} satisfies ImagesModel<"openrouter-images">,
+		"tencent/hy-image-v3.5-preview": {
+			id: "tencent/hy-image-v3.5-preview",
+			name: "Tencent: Hy Image 3.5 Preview",
+			api: "openrouter-images",
+			provider: "openrouter",
+			baseUrl: "https://openrouter.ai/api/v1",
+			input: ["text", "image"],
+			output: ["image"],
+			cost: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+		} satisfies ImagesModel<"openrouter-images">,
 		"x-ai/grok-imagine-image-2.0": {
 			id: "x-ai/grok-imagine-image-2.0",
 			name: "xAI: Grok Imagine Image 2.0",

packages/coding-agent/CHANGELOG.md

@@ -2,9 +2,36 @@
 
 ## [Unreleased]
 
+### New Features
+
+- Added GPT-6 Astra, Luna, Sol, and GPT-6.1 Sol to the model list for ChatGPT login.
+  See [Providers](docs/providers.md).
+- Added cache-preserving harness updates: background memory and evolved-skill changes apply after successful compaction or in a new session, with pending-update notices in the TUI.
+  See [Request boundaries](docs/evolution.md#request-boundaries) ([#260](https://github.com/diffexai/diffex/pull/260) by [@Issacboy123](https://github.com/Issacboy123)).
+- Retained lessons and evolved skills longer before score-based grooming removes them, while keeping existing capacity limits.
+  See [Lesson feedback and grooming](docs/evolution.md#lesson-feedback-and-grooming) and [Private skill feedback](docs/evolution.md#private-skill-feedback).
+
+### Added
+
+- Added GPT-6 Astra, Luna, Sol, and GPT-6.1 Sol to the Codex model list for ChatGPT login.
+
 ### Changed
 
-- Regular root sessions now reconcile selected harness memory and evolved skills at construction, idle reload, and eligible manual compaction boundaries.
+- Deferred background harness memory and evolved-skill updates until successful manual or automatic compaction or a new session, preserving the active conversation's input cache. The TUI announces pending updates; `/clear` starts with the latest revision and `/version` still switches immediately.
+  See [Request boundaries](docs/evolution.md#request-boundaries) ([#260](https://github.com/diffexai/diffex/pull/260) by [@Issacboy123](https://github.com/Issacboy123)).
+- Retained lessons and evolved skills longer by lowering grooming removal thresholds from -5 to -7 for lessons and from -8 to -11 for skills, without changing capacity limits.
+  See [Lesson feedback and grooming](docs/evolution.md#lesson-feedback-and-grooming) and [Private skill feedback](docs/evolution.md#private-skill-feedback).
+- Updated provider SDKs and runtime dependencies, including terminal rendering, syntax highlighting, credential storage, and HTTP libraries ([#272](https://github.com/diffexai/diffex/pull/272), [#273](https://github.com/diffexai/diffex/pull/273), [#274](https://github.com/diffexai/diffex/pull/274)).
+- Anthropic requests now omit unsupported sampling parameters, including `temperature`. Google Gemini and Vertex AI requests no longer perform automatic SDK continuations; continuation and context-window limits are reported as `length`.
+  See [Provider SDK migration notes](../ai/README.md#provider-sdk-migration-notes) ([#273](https://github.com/diffexai/diffex/pull/273)).
+- Changed project and package licensing to MIT ([#280](https://github.com/diffexai/diffex/pull/280)).
+
+### Fixed
+
+- Fixed Google model streaming through OpenCode Go by registering its Google API implementation ([#265](https://github.com/diffexai/diffex/pull/265)).
+- Fixed failed evolution evidence scans being skipped permanently; bounded retries preserve the failed cursor and expose retry status in `/evolve` ([#258](https://github.com/diffexai/diffex/pull/258) by [@Issacboy123](https://github.com/Issacboy123)).
+- Enabled short prompt-cache retention for background evolution model calls ([#259](https://github.com/diffexai/diffex/pull/259) by [@Issacboy123](https://github.com/Issacboy123)).
+- Updated vulnerable CLI dependencies, including Undici and Zod, and patched shell-quote in the sandbox extension example ([#262](https://github.com/diffexai/diffex/pull/262), [#272](https://github.com/diffexai/diffex/pull/272), [#274](https://github.com/diffexai/diffex/pull/274), [#281](https://github.com/diffexai/diffex/pull/281)).
 
 ## [0.3.2] - 2026-09-29
 

packages/tui/CHANGELOG.md

@@ -2,6 +2,16 @@
 
 ## [Unreleased]
 
+### Added
+
+- Added atomic editor tokens with compact display labels that expand on submission, including `EditorTextPart`, `EditorTextToken`, and optional `EditorComponent.setTextWithTokens` ([#132](https://github.com/diffexai/diffex/pull/132)).
+- Added configurable placeholder-cursor styling and shared selected-list prefix exports ([#58](https://github.com/diffexai/diffex/pull/58)).
+
+### Changed
+
+- Updated Markdown parsing and character-width dependencies ([#272](https://github.com/diffexai/diffex/pull/272)).
+- Changed package licensing to MIT ([#280](https://github.com/diffexai/diffex/pull/280)).
+
 ## [0.3.2] - 2026-09-29
 
 ## [0.3.1] - 2026-09-20

scripts/artifact-acceptance.mjs

@@ -183,7 +183,7 @@ export function installConsumer(
 		manager === "npm"
 			? ["install", "--omit=dev", "--ignore-scripts", "--no-audit", "--no-fund"]
 			: ["install", "--production", "--ignore-scripts", "--network-concurrency=8"],
-		{ cwd: directory, env, stdio: "inherit", timeout: 180_000 },
+		{ cwd: directory, env, stdio: "inherit", timeout: 600_000 },
 	);
 	for (const pkg of selected) {
 		const installed = join(directory, "node_modules", pkg.name);
@@ -329,7 +329,7 @@ export function acceptArtifacts(artifactDirectory, outputDirectory, { managers =
 				outputDirectory,
 				...(managers.includes("bun") ? [] : ["--skip-bun"]),
 			],
-			{ cwd: root, stdio: "inherit", timeout: 600000 },
+			{ cwd: root, stdio: "inherit", timeout: 3_600_000 },
 		);
 		verifyTarballs(tarballDirectory, records, packages);
 		writeFileSync(

scripts/npm-lifecycle.mjs

@@ -104,7 +104,13 @@ await new Promise((done) => server.listen(0, "127.0.0.1", done));
 async function command(command, args, extra = {}) {
 	try {
 		console.log(`Lifecycle: ${command} ${args.join(" ")}`);
-		const result = await run(command, args, { ...options, ...extra });
+		const packageManagerOperation =
+			(["npm", "bun"].includes(command) && args.includes("install")) || args.includes("--self");
+		const result = await run(command, args, {
+			...options,
+			timeout: packageManagerOperation ? 600_000 : options.timeout,
+			...extra,
+		});
 		evidence.commands.push({ command, args, exitCode: 0 });
 		return result.stdout;
 	} catch (error) {

scripts/release-source.mjs

@@ -102,7 +102,7 @@ export function pushPreparedRelease(tag) {
 	const commit = git("rev-parse", `refs/tags/${tag}^{commit}`);
 	if (
 		git("rev-parse", "HEAD^") !== commit ||
-		git("log", "-1", "--format=%s") !== "Add [Unreleased] section for next cycle"
+		git("log", "-1", "--format=%s") !== "docs(coding-agent): start next release cycle"
 	)
 		throw new Error("Prepared main must contain exactly the next-cycle commit after the release tag");
 	const changed = git("diff-tree", "--no-commit-id", "--name-only", "-r", "HEAD").split("\n");

scripts/release-source.test.mjs

@@ -220,7 +220,7 @@ for (const failure of ["rejected transaction", "lost acknowledgement"])
 				"# Changelog\n\n## [Unreleased]\n\n## [0.0.3]\n\n- Fixture\n",
 			);
 			f.git("add", "packages/ai/CHANGELOG.md");
-			f.git("commit", "-m", "Add [Unreleased] section for next cycle");
+			f.git("commit", "-m", "docs(coding-agent): start next release cycle");
 			const final = f.git("rev-parse", "HEAD");
 			const hook = join(f.remote, "hooks/pre-receive");
 			if (failure === "rejected transaction") writeFileSync(hook, "#!/bin/sh\nexit 1\n", { mode: 0o755 });

scripts/release.mjs

@@ -224,7 +224,7 @@ if (
 )
 	throw new Error("origin/main changed during preparation; review the new main before releasing");
 stageReleaseOutputs(intendedOutputs);
-run(`git commit -m "Release v${version}"`);
+run(`git commit -m "fix(coding-agent): release v${version}"`);
 run(`git tag v${version}`);
 console.log();
 
@@ -236,7 +236,7 @@ console.log();
 // Commit
 console.log("Committing changelog updates...");
 stageReleaseOutputs(new Set(getChangelogs()));
-run(`git commit -m "Add [Unreleased] section for next cycle"`);
+run(`git commit -m "docs(coding-agent): start next release cycle"`);
 console.log();
 
 // Push