docs(coding-agent): audit v0.4.2 changelogs

packages/ai/CHANGELOG.md

@@ -2,6 +2,10 @@
 
 ## [Unreleased]
 
+### Added
+
+- Exported `CONTEXT_SAFETY_TOKENS` and `estimateContextTokens` so context management can share provider request estimates and safety reserves.
+
 ## [0.4.1] - 2026-10-29
 
 ### Breaking Changes

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

@@ -515,6 +515,36 @@ export const IMAGE_MODELS = {
 				cacheWrite: 0,
 			},
 		} satisfies ImagesModel<"openrouter-images">,
+		"openai/gpt-image-2.6-flare": {
+			id: "openai/gpt-image-2.6-flare",
+			name: "OpenAI: GPT Image 2.6 Flare",
+			api: "openrouter-images",
+			provider: "openrouter",
+			baseUrl: "https://openrouter.ai/api/v1",
+			input: ["text", "image"],
+			output: ["image"],
+			cost: {
+				input: 8,
+				output: 8,
+				cacheRead: 2,
+				cacheWrite: 10,
+			},
+		} satisfies ImagesModel<"openrouter-images">,
+		"openai/gpt-image-2.6-sunburst": {
+			id: "openai/gpt-image-2.6-sunburst",
+			name: "OpenAI: GPT Image 2.6 Sunburst",
+			api: "openrouter-images",
+			provider: "openrouter",
+			baseUrl: "https://openrouter.ai/api/v1",
+			input: ["text", "image"],
+			output: ["image"],
+			cost: {
+				input: 8,
+				output: 8,
+				cacheRead: 2,
+				cacheWrite: 10,
+			},
+		} satisfies ImagesModel<"openrouter-images">,
 		"openrouter/auto": {
 			id: "openrouter/auto",
 			name: "Auto Router",
@@ -545,6 +575,21 @@ export const IMAGE_MODELS = {
 				cacheWrite: 0,
 			},
 		} satisfies ImagesModel<"openrouter-images">,
+		"openrouter/auto-code": {
+			id: "openrouter/auto-code",
+			name: "Auto Code",
+			api: "openrouter-images",
+			provider: "openrouter",
+			baseUrl: "https://openrouter.ai/api/v1",
+			input: ["text", "image"],
+			output: ["text", "image"],
+			cost: {
+				input: -1000000,
+				output: -1000000,
+				cacheRead: 0,
+				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",
@@ -905,9 +950,9 @@ 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",
+		"sourceful/riverflow-v2.6-pro": {
+			id: "sourceful/riverflow-v2.6-pro",
+			name: "Sourceful: Riverflow V2.6 Pro",
 			api: "openrouter-images",
 			provider: "openrouter",
 			baseUrl: "https://openrouter.ai/api/v1",
@@ -920,9 +965,9 @@ export const IMAGE_MODELS = {
 				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",
+		"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",
@@ -935,9 +980,9 @@ export const IMAGE_MODELS = {
 				cacheWrite: 0,
 			},
 		} satisfies ImagesModel<"openrouter-images">,
-		"x-ai/grok-imagine-image-quality": {
-			id: "x-ai/grok-imagine-image-quality",
-			name: "SpaceXAI: Grok Imagine Image Quality",
+		"x-ai/grok-imagine-image-2.0": {
+			id: "x-ai/grok-imagine-image-2.0",
+			name: "xAI: Grok Imagine Image 2.0",
 			api: "openrouter-images",
 			provider: "openrouter",
 			baseUrl: "https://openrouter.ai/api/v1",

packages/coding-agent/CHANGELOG.md

@@ -2,6 +2,35 @@
 
 ## [Unreleased]
 
+### New Features
+
+- Added session-local `/goal` mode with persistent context archives and private recall of original evidence.
+  See [Goal mode](README.md#goal-mode) and [Planning and recall](docs/goal-recall.md) ([#312](https://github.com/diffexai/diffex/pull/312)).
+- Added selective compaction with historical summaries that preserve observations, actions, decisions, and verification limits without reviving old plans.
+  See [Planning and recall](docs/goal-recall.md) ([#317](https://github.com/diffexai/diffex/pull/317), [#319](https://github.com/diffexai/diffex/pull/319)).
+- Goal mode delegates to subagents only when explicitly requested by the user ([#322](https://github.com/diffexai/diffex/pull/322)).
+
+### Added
+
+- Added `/goal` controls, footer status, and SDK session configuration with branch-local persistence across resume, forks, and tree navigation ([#312](https://github.com/diffexai/diffex/pull/312)).
+- Added `compact_block` and `query_block` tools with durable original evidence, branch-scoped archive discovery, bounded private recall, and cancellation-safe recovery.
+  See [Planning and recall](docs/goal-recall.md) ([#312](https://github.com/diffexai/diffex/pull/312)).
+- Exported shared context estimation and provider safety-reserve helpers from the AI toolkit for consistent request budgets.
+
+### Changed
+
+- Goal mode suspends harness evolution and hides `update_plan` without changing global preferences; leaving goal mode restores configured planning and evolution behavior ([#312](https://github.com/diffexai/diffex/pull/312)).
+- Reduced repeated goal context-usage notes to every other completed tool batch while reporting higher pressure tiers immediately ([#312](https://github.com/diffexai/diffex/pull/312)).
+- Goal compaction now uses a four-section historical evidence outline and can summarize user messages and the latest complete tool exchange while preserving custom instructions and malformed exchanges ([#317](https://github.com/diffexai/diffex/pull/317), [#319](https://github.com/diffexai/diffex/pull/319)).
+- Required explicit user direction before goal mode spawns subagents, while allowing omitted counts and purposes to be inferred within that request ([#322](https://github.com/diffexai/diffex/pull/322)).
+
+### Fixed
+
+- Preserved protected context and archive provenance during emergency goal recovery, and allowed summary generation to use the model's available context and output capacity ([#316](https://github.com/diffexai/diffex/pull/316)).
+- Rejected oversized, truncated, empty, cancelled, failed, or tool-producing summaries before committing an archive ([#312](https://github.com/diffexai/diffex/pull/312)).
+- Added bounded retries for transient registry failures during pnpm release acceptance ([#312](https://github.com/diffexai/diffex/pull/312)).
+- Fixed development launchers to start without tsx IPC sockets and preserve the caller's working directory ([#321](https://github.com/diffexai/diffex/pull/321)).
+
 ## [0.4.1] - 2026-10-29
 
 ### New Features