dominds 1.17.7 → 1.18.2

package/dist/docs/dialog-system.zh.md

@@ -241,7 +241,8 @@ flowchart TD
 **支线交付规则（规范）**：
 
 - 只有当所有目标完成时，支线对话才可直接正常回复诉请者对话。
-- 若任何目标未完成或关键信息缺失，必须先用 `tellaskBack({ tellaskContent: "..." })` 回问诉请者对话再继续。
+- 若目标尚未完成，不要默认直接 `tellaskBack`；应先按团队规程 / SOP / 职责卡判断能否明确负责人，若能明确且属于执行性处理，直接 `tellask` / `tellaskSessionless` 对应负责人。
+- 只有当必须由上游诉请者补充需求、澄清目标、做业务裁决、确认验收口径、提供缺失输入，或现有规程无法明确判责时，才使用 `tellaskBack({ tellaskContent: "..." })` 回问诉请者对话再继续。
 - **FBR 例外**：FBR 支线对话禁止一切 tellask（包括 `tellaskBack` / `tellask` / `tellaskSessionless` / `askHuman`），只能列出缺口与阻塞原因并直接回复。
 
 **跨对话传递与标记（强制）**：
@@ -262,7 +263,7 @@ flowchart TD
 
 **协议澄清**：
 
-- 需要回问时必须真实调用 `tellaskBack({ tellaskContent: "..." })`；不得发送普通文本中间汇报。
+- 需要回问上游时必须真实调用 `tellaskBack({ tellaskContent: "..." })`；但在此之前应先判断是否已有团队规程可直接判责到其他负责人。不得发送普通文本中间汇报。
 
 **示例**：
 

package/dist/docs/tellask-collab.md

@@ -120,7 +120,8 @@ That is a workflow break. The model should send the Tellask directly.
 
 - If a sideline dialog has completed all goals and can deliver the final result, it MUST reply directly with the response body; do not use `tellaskBack` to send final delivery.
 - Runtime treats that direct reply as the completion delivery to the tellasker dialog and injects the work-language marker automatically (`【Completed】` in English work language, `【最终完成】` in Chinese work language).
-- If any goal is incomplete, the dialog is blocked, or critical context is missing, it MUST issue `tellaskBack({ tellaskContent: "..." })` before proceeding; do not post plain-text intermediate status updates while unfinished.
+- If the work is unfinished, do not default to `tellaskBack`: first use team SOP / role ownership to judge whether a responsible owner is already clear; if yes and the issue is execution work, directly use `tellask` / `tellaskSessionless` for that owner.
+- Use `tellaskBack({ tellaskContent: "..." })` only when the upstream requester must clarify the request, decide a tradeoff, confirm acceptance criteria, provide missing input, or current SOP cannot determine ownership; do not post plain-text intermediate status updates while unfinished.
 - **FBR exception**: FBR forbids all tellask calls (including `tellaskBack` / `tellask` / `tellaskSessionless` / `askHuman`); list missing context + reasoning and return.
 - If a human user inserts a message or asks a follow-up in the sideline: just reply normally; no need to report back to the upstream requester.
 

package/dist/docs/tellask-collab.zh.md

@@ -116,7 +116,8 @@
 **支线交付规则**：
 
 - 只有当所有目标完成时，支线对话才可直接正常回复诉请者对话（直接发正文即可，不需要调用 tellaskBack；运行时自动注入完成标记）。
-- 若任何目标未完成或关键信息缺失，必须先用 `tellaskBack({ tellaskContent: "..." })` 回问诉请者对话再继续；不得发送普通文本中间汇报。
+- 若目标尚未完成，不要默认直接 `tellaskBack`：应先按团队规程 / SOP / 职责卡判断能否明确负责人；若能明确且属于执行性处理，直接 `tellask` / `tellaskSessionless` 对应负责人。
+- 只有当必须由上游诉请者补充需求、澄清目标、做业务裁决、确认验收口径、提供缺失输入，或现有规程无法明确判责时，才使用 `tellaskBack({ tellaskContent: "..." })` 回问诉请者对话；不得发送普通文本中间汇报。
 - **FBR 例外**：FBR 禁止一切 tellask（包括 `tellaskBack` / `tellask` / `tellaskSessionless` / `askHuman`）；只列缺口 + 推理与摘要并直接回复。
 - 若人类用户在支线对话中插入消息或补问：直接正常回复，不需要向诉请者汇报。
 

package/dist/llm/defaults.yaml

@@ -199,24 +199,28 @@ providers:
     models:
       MiniMax-M2.7:
         name: MiniMax M2.7
+        supports_image_input: false
         context_length: 204800
         input_length: 204800
         output_length: 8192
         context_window: '204K'
       MiniMax-M2.5:
         name: MiniMax M2.5
+        supports_image_input: false
         context_length: 204800
         input_length: 204800
         output_length: 8192
         context_window: '204K'
       MiniMax-M2.1:
         name: MiniMax M2.1
+        supports_image_input: false
         context_length: 204800
         input_length: 204800
         output_length: 8192
         context_window: '204K'
       MiniMax-M2:
         name: MiniMax M2
+        supports_image_input: false
         context_length: 204800
         input_length: 204800
         output_length: 8192
@@ -252,24 +256,28 @@ providers:
     models:
       MiniMax-M2.7:
         name: MiniMax M2.7
+        supports_image_input: false
         context_length: 204800
         input_length: 204800
         output_length: 8192
         context_window: '204K'
       MiniMax-M2.5:
         name: MiniMax M2.5
+        supports_image_input: false
         context_length: 204800
         input_length: 204800
         output_length: 8192
         context_window: '204K'
       MiniMax-M2.1:
         name: MiniMax M2.1
+        supports_image_input: false
         context_length: 204800
         input_length: 204800
         output_length: 8192
         context_window: '204K'
       MiniMax-M2:
         name: MiniMax M2 Stable
+        supports_image_input: false
         context_length: 204800
         input_length: 204800
         output_length: 8192
@@ -305,24 +313,28 @@ providers:
     models:
       MiniMax-M2.7:
         name: MiniMax M2.7
+        supports_image_input: false
         context_length: 204800
         input_length: 204800
         output_length: 8192
         context_window: '204K'
       MiniMax-M2.5:
         name: MiniMax M2.5
+        supports_image_input: false
         context_length: 204800
         input_length: 204800
         output_length: 8192
         context_window: '204K'
       MiniMax-M2.1:
         name: MiniMax M2.1
+        supports_image_input: false
         context_length: 204800
         input_length: 204800
         output_length: 8192
         context_window: '204K'
       MiniMax-M2:
         name: MiniMax M2
+        supports_image_input: false
         context_length: 204800
         input_length: 204800
         output_length: 8192
@@ -358,24 +370,28 @@ providers:
     models:
       MiniMax-M2.7:
         name: MiniMax M2.7
+        supports_image_input: false
         context_length: 204800
         input_length: 204800
         output_length: 8192
         context_window: '204K'
       MiniMax-M2.5:
         name: MiniMax M2.5
+        supports_image_input: false
         context_length: 204800
         input_length: 204800
         output_length: 8192
         context_window: '204K'
       MiniMax-M2.1:
         name: MiniMax M2.1
+        supports_image_input: false
         context_length: 204800
         input_length: 204800
         output_length: 8192
         context_window: '204K'
       MiniMax-M2:
         name: MiniMax M2 Stable
+        supports_image_input: false
         context_length: 204800
         input_length: 204800
         output_length: 8192
@@ -390,24 +406,28 @@ providers:
     models:
       glm-4.7:
         name: GLM-4.7
+        supports_image_input: false
         context_length: 200000
         input_length: 200000
         output_length: 8192
         context_window: '200K'
       glm-4.6:
         name: GLM-4.6
+        supports_image_input: false
         context_length: 200000
         input_length: 200000
         output_length: 8192
         context_window: '200K'
       glm-4.5:
         name: GLM-4.5
+        supports_image_input: false
         context_length: 128000
         input_length: 128000
         output_length: 8192
         context_window: '128K'
       glm-4.5-air:
         name: GLM-4.5-Air
+        supports_image_input: false
         context_length: 128000
         input_length: 128000
         output_length: 8192
@@ -422,24 +442,28 @@ providers:
     models:
       glm-4.7:
         name: GLM-4.7
+        supports_image_input: false
         context_length: 200000
         input_length: 200000
         output_length: 8192
         context_window: '200K'
       glm-4.6:
         name: GLM-4.6
+        supports_image_input: false
         context_length: 200000
         input_length: 200000
         output_length: 8192
         context_window: '200K'
       glm-4.5:
         name: GLM-4.5
+        supports_image_input: false
         context_length: 128000
         input_length: 128000
         output_length: 8192
         context_window: '128K'
       glm-4.5-air:
         name: GLM-4.5-Air
+        supports_image_input: false
         context_length: 128000
         input_length: 128000
         output_length: 8192
@@ -454,6 +478,7 @@ providers:
     models:
       doubao-seed-code:
         name: Doubao Seed Code Preview
+        supports_image_input: true
         context_length: 256000
         input_length: 256000
         output_length: 8192
@@ -461,12 +486,14 @@ providers:
         optimization: 专为Agentic Coding任务优化
       deepseek-v3.2:
         name: DeepSeek-V3.2
+        supports_image_input: false
         context_length: 128000
         input_length: 96000
         output_length: 32000
         context_window: '128K'
       glm-4.7:
         name: GLM 4.7
+        supports_image_input: false
         context_length: 256000
         input_length: 256000
         output_length: 8192
@@ -487,42 +514,49 @@ providers:
     models:
       deepseek-v3-2-251201:
         name: DeepSeek-V3.2
+        supports_image_input: false
         context_length: 128000
         input_length: 96000
         output_length: 32000
         context_window: '128K'
       doubao-seed-code-preview-251028:
         name: Doubao Seed Code Preview 251028
+        supports_image_input: true
         context_length: 256000
         input_length: 256000
         output_length: 8192
         context_window: '256K'
       doubao-seed-1-6-251015:
         name: Doubao Seed 1.6
+        supports_image_input: false
         context_length: 256000
         input_length: 256000
         output_length: 8192
         context_window: '256K'
       doubao-seed-1-6-thinking-250715:
         name: Doubao Seed 1.6 Thinking
+        supports_image_input: false
         context_length: 256000
         input_length: 256000
         output_length: 8192
         context_window: '256K'
       doubao-seed-1-6-lite-251015:
         name: Doubao Seed 1.6 Lite
+        supports_image_input: false
         context_length: 32000
         input_length: 32000
         output_length: 8192
         context_window: '32K'
       kimi-k2-250905:
         name: Kimi K2
+        supports_image_input: false
         context_length: 200000
         input_length: 200000
         output_length: 8192
         context_window: '200K (约20万汉字)'
       deepseek-v3-1-terminus:
         name: DeepSeek V3.1 Terminus
+        supports_image_input: false
         context_length: 131072
         input_length: 131072
         output_length: 8192
@@ -621,12 +655,14 @@ providers:
     models:
       deepseek-chat:
         name: DeepSeek-V3.2
+        supports_image_input: false
         context_length: 128000
         input_length: 128000
         output_length: 8192
         context_window: '128K'
       deepseek-reasoner:
         name: DeepSeek-R1 (reasoner, V3.2 thinking mode)
+        supports_image_input: false
         context_length: 128000
         input_length: 128000
         output_length: 65536
@@ -665,42 +701,49 @@ providers:
     models:
       qwen3.5-plus:
         name: Qwen3.5 Plus
+        supports_image_input: true
         context_length: 1000000
         input_length: 983616
         output_length: 65536
         context_window: '1M'
       kimi-k2.5:
         name: Kimi K2.5
+        supports_image_input: true
         context_length: 262144
         input_length: 258048
         output_length: 32768
         context_window: '262K'
       glm-5:
         name: GLM-5
+        supports_image_input: false
         context_length: 202752
         input_length: 202752
         output_length: 16384
         context_window: '202K'
       MiniMax-M2.5:
         name: MiniMax M2.5
+        supports_image_input: false
         context_length: 204800
         input_length: 196608
         output_length: 131072
         context_window: '204K'
       qwen3-coder-next:
         name: Qwen3 Coder Next
+        supports_image_input: false
         context_length: 262144
         input_length: 204800
         output_length: 65536
         context_window: '262K'
       qwen3-coder-plus:
         name: Qwen3 Coder Plus
+        supports_image_input: false
         context_length: 1000000
         input_length: 997952
         output_length: 65536
         context_window: '1M'
       glm-4.7:
         name: GLM-4.7
+        supports_image_input: false
         context_length: 202752
         input_length: 169984
         output_length: 16384

package/dist/llm/gen/anthropic.js

@@ -185,7 +185,7 @@ function parseForcedJsonToolInput(rawJson, fallbackInput, at) {
     }
     throw new Error(`Invalid ${at}: ${errorText}; raw=${JSON.stringify(preview)}`);
 }
-async function funcResultToAnthropicToolResultBlock(chatMsg, limitChars, requestContext, allowedImageKeys, onToolResultImageIngest) {
+async function funcResultToAnthropicToolResultBlock(chatMsg, limitChars, requestContext, allowedImageKeys, supportsImageInput, onToolResultImageIngest) {
     const items = chatMsg.contentItems;
     if (!Array.isArray(items) || items.length === 0) {
         return {
@@ -195,12 +195,29 @@ async function funcResultToAnthropicToolResultBlock(chatMsg, limitChars, request
         };
     }
     const content = [];
-    for (const item of items) {
+    for (const [itemIndex, item] of items.entries()) {
         if (item.type === 'input_text') {
             content.push({ type: 'text', text: item.text });
             continue;
         }
         if (item.type === 'input_image') {
+            if (!supportsImageInput) {
+                if (onToolResultImageIngest) {
+                    await onToolResultImageIngest((0, tool_result_image_ingest_1.buildToolResultImageIngest)({
+                        requestContext,
+                        toolCallId: chatMsg.id,
+                        toolName: chatMsg.name,
+                        artifact: item.artifact,
+                        disposition: 'filtered_model_unsupported',
+                        providerPathLabel: 'Anthropic Messages path',
+                    }));
+                }
+                content.push({
+                    type: 'text',
+                    text: `[image not sent: current model does not support image input]`,
+                });
+                continue;
+            }
             if (!(0, artifacts_1.isVisionImageMimeType)(item.mimeType)) {
                 if (onToolResultImageIngest) {
                     await onToolResultImageIngest((0, tool_result_image_ingest_1.buildToolResultImageIngest)({
@@ -219,7 +236,7 @@ async function funcResultToAnthropicToolResultBlock(chatMsg, limitChars, request
                 });
                 continue;
             }
-            if (!allowedImageKeys.has((0, tool_result_image_ingest_1.buildToolResultImageBudgetKeyForMsg)(chatMsg, item.artifact))) {
+            if (!allowedImageKeys.has((0, tool_result_image_ingest_1.buildImageBudgetKeyForContentItem)({ msg: chatMsg, itemIndex, artifact: item.artifact }))) {
                 if (onToolResultImageIngest) {
                     await onToolResultImageIngest((0, tool_result_image_ingest_1.buildToolResultImageIngest)({
                         requestContext,
@@ -227,7 +244,7 @@ async function funcResultToAnthropicToolResultBlock(chatMsg, limitChars, request
                         toolName: chatMsg.name,
                         artifact: item.artifact,
                         disposition: 'filtered_size_limit',
-                        detail: (0, tool_result_image_ingest_1.buildToolResultImageBudgetLimitDetail)({
+                        detail: (0, tool_result_image_ingest_1.buildImageBudgetLimitDetail)({
                             byteLength: item.byteLength,
                             budgetBytes: tool_result_image_ingest_1.ANTHROPIC_TOOL_RESULT_IMAGE_BUDGET_BYTES,
                         }),
@@ -308,16 +325,135 @@ async function funcResultToAnthropicToolResultBlock(chatMsg, limitChars, request
         content: limitAnthropicToolOutputBlocks(content, chatMsg, limitChars),
     };
 }
-async function chatMessageToContentBlocksAsync(chatMsg, limitChars, requestContext, allowedImageKeys, onToolResultImageIngest) {
+async function chatMessageToContentBlocksAsync(chatMsg, limitChars, requestContext, allowedImageKeys, supportsImageInput, onToolResultImageIngest, onUserImageIngest) {
+    if ((chatMsg.type === 'prompting_msg' ||
+        chatMsg.type === 'tellask_result_msg' ||
+        chatMsg.type === 'tellask_carryover_msg') &&
+        Array.isArray(chatMsg.contentItems) &&
+        chatMsg.contentItems.length > 0) {
+        const content = [
+            { type: 'text', text: chatMsg.content },
+        ];
+        for (const [itemIndex, item] of chatMsg.contentItems.entries()) {
+            if (item.type === 'input_text') {
+                content.push({ type: 'text', text: item.text });
+                continue;
+            }
+            if (item.type === 'input_image') {
+                if (!supportsImageInput) {
+                    if (onUserImageIngest) {
+                        await onUserImageIngest((0, tool_result_image_ingest_1.buildUserImageIngest)({
+                            requestContext,
+                            ...(chatMsg.type === 'prompting_msg' ? { msgId: chatMsg.msgId } : {}),
+                            artifact: item.artifact,
+                            disposition: 'filtered_model_unsupported',
+                            providerPathLabel: 'Anthropic Messages path',
+                        }));
+                    }
+                    content.push({
+                        type: 'text',
+                        text: `[image not sent: current model does not support image input]`,
+                    });
+                    continue;
+                }
+                if (!(0, artifacts_1.isVisionImageMimeType)(item.mimeType)) {
+                    if (onUserImageIngest) {
+                        await onUserImageIngest((0, tool_result_image_ingest_1.buildUserImageIngest)({
+                            requestContext,
+                            ...(chatMsg.type === 'prompting_msg' ? { msgId: chatMsg.msgId } : {}),
+                            artifact: item.artifact,
+                            disposition: 'filtered_mime_unsupported',
+                            mimeType: item.mimeType,
+                            providerPathLabel: 'Anthropic Messages path',
+                        }));
+                    }
+                    content.push({
+                        type: 'text',
+                        text: `[image not sent: unsupported mimeType=${item.mimeType}]`,
+                    });
+                    continue;
+                }
+                if (!allowedImageKeys.has((0, tool_result_image_ingest_1.buildImageBudgetKeyForContentItem)({ msg: chatMsg, itemIndex, artifact: item.artifact }))) {
+                    if (onUserImageIngest) {
+                        await onUserImageIngest((0, tool_result_image_ingest_1.buildUserImageIngest)({
+                            requestContext,
+                            ...(chatMsg.type === 'prompting_msg' ? { msgId: chatMsg.msgId } : {}),
+                            artifact: item.artifact,
+                            disposition: 'filtered_size_limit',
+                            detail: (0, tool_result_image_ingest_1.buildImageBudgetLimitDetail)({
+                                byteLength: item.byteLength,
+                                budgetBytes: tool_result_image_ingest_1.ANTHROPIC_TOOL_RESULT_IMAGE_BUDGET_BYTES,
+                            }),
+                            providerPathLabel: 'Anthropic Messages path',
+                        }));
+                    }
+                    content.push({
+                        type: 'text',
+                        text: `[image not sent: request image budget exceeded bytes=${String(item.byteLength)} budget=${String(tool_result_image_ingest_1.ANTHROPIC_TOOL_RESULT_IMAGE_BUDGET_BYTES)}]`,
+                    });
+                    continue;
+                }
+                const bytesResult = await (0, tool_result_image_ingest_1.readToolResultImageBytesSafe)(item.artifact);
+                if (bytesResult.kind === 'missing') {
+                    if (onUserImageIngest) {
+                        await onUserImageIngest((0, tool_result_image_ingest_1.buildUserImageIngest)({
+                            requestContext,
+                            ...(chatMsg.type === 'prompting_msg' ? { msgId: chatMsg.msgId } : {}),
+                            artifact: item.artifact,
+                            disposition: 'filtered_missing',
+                            providerPathLabel: 'Anthropic Messages path',
+                        }));
+                    }
+                    content.push({ type: 'text', text: `[image missing: ${item.artifact.relPath}]` });
+                    continue;
+                }
+                if (bytesResult.kind === 'read_failed') {
+                    if (onUserImageIngest) {
+                        await onUserImageIngest((0, tool_result_image_ingest_1.buildUserImageIngest)({
+                            requestContext,
+                            ...(chatMsg.type === 'prompting_msg' ? { msgId: chatMsg.msgId } : {}),
+                            artifact: item.artifact,
+                            disposition: 'filtered_read_failed',
+                            detail: bytesResult.detail,
+                            providerPathLabel: 'Anthropic Messages path',
+                        }));
+                    }
+                    content.push({ type: 'text', text: `[image unreadable: ${item.artifact.relPath}]` });
+                    continue;
+                }
+                if (onUserImageIngest) {
+                    await onUserImageIngest((0, tool_result_image_ingest_1.buildUserImageIngest)({
+                        requestContext,
+                        ...(chatMsg.type === 'prompting_msg' ? { msgId: chatMsg.msgId } : {}),
+                        artifact: item.artifact,
+                        disposition: 'fed_native',
+                        providerPathLabel: 'Anthropic Messages path',
+                    }));
+                }
+                content.push({
+                    type: 'image',
+                    source: {
+                        type: 'base64',
+                        media_type: item.mimeType,
+                        data: bytesResult.bytes.toString('base64'),
+                    },
+                });
+                continue;
+            }
+            const _exhaustive = item;
+            throw new Error(`Unsupported user content item: ${String(_exhaustive)}`);
+        }
+        return content;
+    }
     if (chatMsg.type !== 'func_result_msg') {
         return chatMessageToContentBlocks(chatMsg);
     }
     return [
-        await funcResultToAnthropicToolResultBlock(chatMsg, limitChars, requestContext, allowedImageKeys, onToolResultImageIngest),
+        await funcResultToAnthropicToolResultBlock(chatMsg, limitChars, requestContext, allowedImageKeys, supportsImageInput, onToolResultImageIngest),
     ];
 }
-async function chatMessageToAnthropicAsync(chatMsg, limitChars, requestContext, allowedImageKeys, onToolResultImageIngest) {
-    const contentBlocks = await chatMessageToContentBlocksAsync(chatMsg, limitChars, requestContext, allowedImageKeys, onToolResultImageIngest);
+async function chatMessageToAnthropicAsync(chatMsg, limitChars, requestContext, allowedImageKeys, supportsImageInput, onToolResultImageIngest, onUserImageIngest) {
+    const contentBlocks = await chatMessageToContentBlocksAsync(chatMsg, limitChars, requestContext, allowedImageKeys, supportsImageInput, onToolResultImageIngest, onUserImageIngest);
     if (contentBlocks.length === 0) {
         throw new Error(`No content blocks generated for message: ${JSON.stringify(chatMsg)}`);
     }
@@ -330,7 +466,7 @@ async function chatMessageToAnthropicAsync(chatMsg, limitChars, requestContext,
         content: contentBlocks.length === 1 ? contentBlocks : contentBlocks,
     };
 }
-async function buildAnthropicRequestMessages(context, requestContext, providerConfig, onToolResultImageIngest) {
+async function buildAnthropicRequestMessages(context, requestContext, providerConfig, onToolResultImageIngest, onUserImageIngest) {
     // We keep the async path for func_result_msg because it may contain image artifacts.
     const normalized = (0, tool_call_context_1.normalizeToolCallPairs)(context);
     const violation = (0, tool_call_context_1.findFirstToolCallAdjacencyViolation)(normalized);
@@ -346,9 +482,12 @@ async function buildAnthropicRequestMessages(context, requestContext, providerCo
     }
     const messages = [];
     const toolResultMaxChars = (0, tool_output_limit_1.resolveProviderToolResultMaxChars)(providerConfig);
-    const allowedImageKeys = (0, tool_result_image_ingest_1.selectLatestToolResultImagesWithinBudget)(normalized, tool_result_image_ingest_1.ANTHROPIC_TOOL_RESULT_IMAGE_BUDGET_BYTES);
+    const allowedImageKeys = (0, tool_result_image_ingest_1.selectLatestImagesWithinBudget)(normalized, tool_result_image_ingest_1.ANTHROPIC_TOOL_RESULT_IMAGE_BUDGET_BYTES);
+    const supportsImageInput = (0, tool_result_image_ingest_1.resolveModelImageInputSupport)(requestContext.modelKey === undefined
+        ? undefined
+        : providerConfig?.models[requestContext.modelKey], true);
     for (const msg of normalized) {
-        messages.push(await chatMessageToAnthropicAsync(msg, toolResultMaxChars, requestContext, allowedImageKeys, onToolResultImageIngest));
+        messages.push(await chatMessageToAnthropicAsync(msg, toolResultMaxChars, requestContext, allowedImageKeys, supportsImageInput, onToolResultImageIngest, onUserImageIngest));
     }
     return assembleAnthropicTurns(messages);
 }
@@ -919,7 +1058,7 @@ class AnthropicGen {
         if (!apiKey)
             throw new Error(`Missing API key env var ${providerConfig.apiKeyEnvVar}`);
         const client = new sdk_1.Anthropic({ apiKey, baseURL: providerConfig.baseUrl });
-        const requestMessages = await buildAnthropicRequestMessages(context, requestContext, providerConfig, receiver.toolResultImageIngest);
+        const requestMessages = await buildAnthropicRequestMessages(context, requestContext, providerConfig, receiver.toolResultImageIngest, receiver.userImageIngest);
         const anthropicParams = agent.model_params?.anthropic || {};
         const forceJsonResponse = resolveAnthropicJsonResponseEnabled(agent);
         const maxTokens = agent.model_params?.max_tokens;
@@ -972,6 +1111,8 @@ class AnthropicGen {
         const outputs = [];
         const requestMessages = await buildAnthropicRequestMessages(context, requestContext, providerConfig, async (ingest) => {
             outputs.push({ kind: 'tool_result_image_ingest', ingest });
+        }, async (ingest) => {
+            outputs.push({ kind: 'user_image_ingest', ingest });
         });
         const anthropicParams = agent.model_params?.anthropic || {};
         const forceJsonResponse = resolveAnthropicJsonResponseEnabled(agent);