@adhdev/daemon-core 0.9.76-rc.61 → 0.9.76-rc.62

package/src/providers/cli-provider-instance.ts

@@ -10,8 +10,8 @@ import * as path from 'path';
 import * as crypto from 'crypto';
 import * as fs from 'fs';
 import { createRequire } from 'node:module';
-import { normalizeInputEnvelope, type ProviderModule, flattenContent } from './contracts.js';
-import { assertTextOnlyInput } from './provider-input-support.js';
+import { normalizeInputEnvelope, type ProviderModule, flattenContent, type InputEnvelope, type InputPart } from './contracts.js';
+import { assertProviderSupportsDeclaredInput, getEffectiveMessageInputSupport } from './provider-input-support.js';
 import type { ProviderInstance, ProviderState, ProviderEvent, InstanceContext, ProviderErrorReason, HotChatSessionState, SessionModalState } from './provider-instance.js';
 import { ProviderCliAdapter } from '../cli-adapters/provider-cli-adapter.js';
 import type { CliProviderModule } from '../cli-adapters/provider-cli-adapter.js';
@@ -35,6 +35,95 @@ type PersistableCliHistoryMessage = {
     receivedAt?: number;
 };
 
+const IMAGE_MIME_EXTENSIONS: Record<string, string> = {
+    'image/png': '.png',
+    'image/jpeg': '.jpg',
+    'image/jpg': '.jpg',
+    'image/gif': '.gif',
+    'image/webp': '.webp',
+    'image/bmp': '.bmp',
+    'image/tiff': '.tiff',
+    'image/svg+xml': '.svg',
+};
+
+function filePathFromUri(uri: string): string | null {
+    if (!uri) return null;
+    if (uri.startsWith('file://')) {
+        try {
+            return decodeURIComponent(new URL(uri).pathname);
+        } catch {
+            return uri.slice('file://'.length);
+        }
+    }
+    if (path.isAbsolute(uri)) return uri;
+    return null;
+}
+
+function extensionForImageMime(mimeType: string): string {
+    return IMAGE_MIME_EXTENSIONS[mimeType.toLowerCase()] || '.img';
+}
+
+function safeInputImageBasename(index: number, mimeType: string): string {
+    const extension = extensionForImageMime(mimeType);
+    const suffix = crypto.randomBytes(6).toString('hex');
+    return `adhdev-input-image-${Date.now()}-${index}-${suffix}${extension}`;
+}
+
+function materializeImageDataPart(part: Extract<InputPart, { type: 'image' }>, index: number, dir: string): string | null {
+    if (!part.data) return null;
+    const rawData = part.data.includes(',') ? part.data.split(',').pop() || '' : part.data;
+    if (!rawData) return null;
+    fs.mkdirSync(dir, { recursive: true });
+    const filePath = path.join(dir, safeInputImageBasename(index, part.mimeType));
+    fs.writeFileSync(filePath, Buffer.from(rawData, 'base64'));
+    return filePath;
+}
+
+export function buildCliStructuredInputPrompt(
+    input: InputEnvelope,
+    options: { materializeDir?: string } = {},
+): string {
+    const promptParts: string[] = [];
+    const imageRefs: string[] = [];
+    const resourceRefs: string[] = [];
+    const materializeDir = options.materializeDir || path.join(os.tmpdir(), 'adhdev-input-media');
+
+    input.parts.forEach((part, index) => {
+        if (part.type === 'text' && part.text.trim()) {
+            promptParts.push(part.text.trim());
+            return;
+        }
+
+        if (part.type === 'image') {
+            const localPath = typeof part.uri === 'string' ? filePathFromUri(part.uri) : null;
+            const materializedPath = !localPath && part.data ? materializeImageDataPart(part, index, materializeDir) : null;
+            const ref = localPath || materializedPath || part.uri || '';
+            if (ref) imageRefs.push(ref);
+            if (part.alt?.trim()) promptParts.push(part.alt.trim());
+            return;
+        }
+
+        if (part.type === 'resource_link') {
+            resourceRefs.push([part.title, part.name, part.description, part.uri].filter(Boolean).join('\n'));
+            return;
+        }
+
+        if (part.type === 'resource') {
+            resourceRefs.push([part.name, part.text, part.uri].filter(Boolean).join('\n'));
+        }
+    });
+
+    if (input.textFallback.trim()) promptParts.push(input.textFallback.trim());
+
+    const ordered = [
+        ...imageRefs,
+        ...promptParts,
+        ...resourceRefs,
+    ].filter((value, index, values) => value.trim().length > 0 && values.indexOf(value) === index);
+
+    return ordered.join('\n');
+}
+
 function normalizePersistableCliHistoryContent(content: unknown): string {
     return flattenContent(content as any).replace(/\s+/g, ' ').trim();
 }
@@ -476,6 +565,7 @@ export class CliProviderInstance implements ProviderInstance {
             resume: this.provider.resume,
             controlValues: surface.controlValues,
             providerControls: this.provider.controls,
+            messageInput: getEffectiveMessageInputSupport(this.provider),
             summaryMetadata: surface.summaryMetadata as any,
             errorMessage: this.errorMessage,
             errorReason: this.errorReason,
@@ -532,9 +622,10 @@ export class CliProviderInstance implements ProviderInstance {
     onEvent(event: string, data?: any): void {
         if (event === 'send_message') {
             const input = normalizeInputEnvelope(data);
-            assertTextOnlyInput(this.provider, input);
-            if (input.textFallback) {
-                void this.adapter.sendMessage(input.textFallback).catch((e: any) => {
+            assertProviderSupportsDeclaredInput(this.provider, input);
+            const promptText = buildCliStructuredInputPrompt(input);
+            if (promptText) {
+                void this.adapter.sendMessage(promptText).catch((e: any) => {
                     LOG.warn('CLI', `[${this.type}] send_message failed: ${e?.message || e}`);
                 });
             }

package/src/providers/contracts.ts

@@ -128,6 +128,7 @@ export type ContentBlock =
   | TextBlock
   | ImageBlock
   | AudioBlock
+  | VideoBlock
   | ResourceLinkBlock
   | ResourceBlock;
 
@@ -144,6 +145,7 @@ export interface ImageBlock {
   data: string;       // base64-encoded
   mimeType: string;   // 'image/png', 'image/jpeg', etc.
   uri?: string;       // optional URL reference
+  alt?: string;
   annotations?: ContentAnnotations;
 }
 
@@ -152,6 +154,19 @@ export interface AudioBlock {
   type: 'audio';
   data: string;       // base64-encoded
   mimeType: string;
+  uri?: string;
+  transcript?: string;
+  annotations?: ContentAnnotations;
+}
+
+/** Video content — ADHDev canonical display block. ACP prompt input degrades video to resource_link/text. */
+export interface VideoBlock {
+  type: 'video';
+  data?: string;      // base64-encoded
+  mimeType: string;
+  uri?: string;
+  transcript?: string;
+  posterUri?: string;
   annotations?: ContentAnnotations;
 }
 
@@ -595,7 +610,16 @@ export interface ProviderModule {
  // ─── Contract version / capability declaration ───
   contractVersion?: number;
   capabilities?: {
-    input?: { multipart?: boolean; mediaTypes?: Array<'text' | 'image' | 'audio' | 'video' | 'resource'> };
+    input?: {
+      multipart?: boolean;
+      mediaTypes?: Array<'text' | 'image' | 'audio' | 'video' | 'resource'>;
+      strategies?: Array<{
+        mediaType: 'text' | 'image' | 'audio' | 'video' | 'resource';
+        strategies?: Array<'native' | 'native_acp' | 'resource_link' | 'text_fallback' | 'paste' | 'upload'>;
+        native?: boolean;
+        degradation?: Array<'native' | 'native_acp' | 'resource_link' | 'text_fallback' | 'paste' | 'upload'>;
+      }>;
+    };
     output?: { richContent?: boolean; mediaTypes?: Array<'text' | 'image' | 'audio' | 'video' | 'resource'> };
     controls?: { typedResults?: boolean };
   };

package/src/providers/io-contracts.ts

@@ -5,6 +5,7 @@ export type InputPart =
   | ImageInputPart
   | AudioInputPart
   | VideoInputPart
+  | ResourceLinkInputPart
   | ResourceInputPart
 
 export interface TextInputPart {
@@ -33,6 +34,7 @@ export interface VideoInputPart {
   mimeType: string
   uri?: string
   data?: string
+  transcript?: string
   posterUri?: string
 }
 
@@ -45,6 +47,17 @@ export interface ResourceInputPart {
   data?: string
 }
 
+export interface ResourceLinkInputPart {
+  type: 'resource_link'
+  uri: string
+  name: string
+  title?: string
+  description?: string
+  mimeType?: string
+  size?: number
+  annotations?: ContentAnnotations
+}
+
 export interface InputEnvelope {
   parts: InputPart[]
   textFallback: string
@@ -73,6 +86,7 @@ export interface ImageMessagePart {
   mimeType: string
   uri?: string
   data?: string
+  alt?: string
   annotations?: ContentAnnotations
 }
 
@@ -90,6 +104,7 @@ export interface VideoMessagePart {
   mimeType: string
   uri?: string
   data?: string
+  transcript?: string
   posterUri?: string
   annotations?: ContentAnnotations
 }
@@ -98,8 +113,11 @@ export interface ResourceLinkMessagePart {
   type: 'resource_link'
   uri: string
   name: string
+  title?: string
+  description?: string
   mimeType?: string
   size?: number
+  annotations?: ContentAnnotations
 }
 
 export interface ResourceMessagePart {
@@ -149,6 +167,10 @@ export function flattenMessageParts(parts: MessagePart[]): string {
     .map((part) => {
       if (part.type === 'text') return part.text
       if (part.type === 'resource') return part.resource.text || ''
+      if (part.type === 'image') return part.alt || (part.data ? `[image: ${part.mimeType}]` : '')
+      if (part.type === 'audio') return part.transcript || (part.data ? `[audio: ${part.mimeType}]` : '')
+      if (part.type === 'video') return part.transcript || (part.data ? `[video: ${part.mimeType}]` : '')
+      if (part.type === 'resource_link') return [part.name, part.description].filter(Boolean).join('\n')
       return ''
     })
     .filter((value) => value.length > 0)
@@ -247,6 +269,7 @@ function normalizeInputPartObject(raw: Record<string, unknown>): InputPart | nul
       mimeType: raw.mimeType,
       ...(typeof raw.uri === 'string' ? { uri: raw.uri } : {}),
       ...(typeof raw.data === 'string' ? { data: raw.data } : {}),
+      ...(typeof raw.transcript === 'string' ? { transcript: raw.transcript } : {}),
       ...(typeof raw.posterUri === 'string' ? { posterUri: raw.posterUri } : {}),
     }
   }
@@ -262,10 +285,14 @@ function normalizeInputPartObject(raw: Record<string, unknown>): InputPart | nul
   }
   if (type === 'resource_link' && typeof raw.uri === 'string') {
     return {
-      type: 'resource',
+      type,
       uri: raw.uri,
+      name: typeof raw.name === 'string' ? raw.name : getUriDisplayName(raw.uri, 'resource'),
+      ...(typeof raw.title === 'string' ? { title: raw.title } : {}),
+      ...(typeof raw.description === 'string' ? { description: raw.description } : {}),
       ...(typeof raw.mimeType === 'string' ? { mimeType: raw.mimeType } : {}),
-      ...(typeof raw.name === 'string' ? { name: raw.name } : {}),
+      ...(typeof raw.size === 'number' && Number.isFinite(raw.size) ? { size: raw.size } : {}),
+      ...normalizeAnnotationsProperty(raw.annotations),
     }
   }
   return null
@@ -282,6 +309,7 @@ function normalizeMessagePartObject(raw: Record<string, unknown>): MessagePart |
       mimeType: raw.mimeType,
       ...(typeof raw.uri === 'string' ? { uri: raw.uri } : {}),
       ...(typeof raw.data === 'string' ? { data: raw.data } : {}),
+      ...(typeof raw.alt === 'string' ? { alt: raw.alt } : {}),
     }
   }
   if (type === 'audio' && typeof raw.mimeType === 'string') {
@@ -299,6 +327,7 @@ function normalizeMessagePartObject(raw: Record<string, unknown>): MessagePart |
       mimeType: raw.mimeType,
       ...(typeof raw.uri === 'string' ? { uri: raw.uri } : {}),
       ...(typeof raw.data === 'string' ? { data: raw.data } : {}),
+      ...(typeof raw.transcript === 'string' ? { transcript: raw.transcript } : {}),
       ...(typeof raw.posterUri === 'string' ? { posterUri: raw.posterUri } : {}),
     }
   }
@@ -307,8 +336,11 @@ function normalizeMessagePartObject(raw: Record<string, unknown>): MessagePart |
       type,
       uri: raw.uri,
       name: raw.name,
+      ...(typeof raw.title === 'string' ? { title: raw.title } : {}),
+      ...(typeof raw.description === 'string' ? { description: raw.description } : {}),
       ...(typeof raw.mimeType === 'string' ? { mimeType: raw.mimeType } : {}),
-      ...(typeof raw.size === 'number' ? { size: raw.size } : {}),
+      ...(typeof raw.size === 'number' && Number.isFinite(raw.size) ? { size: raw.size } : {}),
+      ...normalizeAnnotationsProperty(raw.annotations),
     }
   }
   if (type === 'resource' && raw.resource && typeof raw.resource === 'object') {
@@ -331,10 +363,36 @@ function flattenInputParts(parts: InputPart[]): string {
   return parts
     .map((part) => {
       if (part.type === 'text') return part.text
-      if (part.type === 'audio') return part.transcript || ''
-      if (part.type === 'resource') return part.text || ''
+      if (part.type === 'image') return part.alt || (part.data ? `[image: ${part.mimeType}]` : '')
+      if (part.type === 'audio') return part.transcript || (part.data ? `[audio: ${part.mimeType}]` : '')
+      if (part.type === 'video') return part.transcript || (part.data ? `[video: ${part.mimeType}]` : '')
+      if (part.type === 'resource_link') return [part.title, part.name, part.description, part.uri].filter(Boolean).join('\n')
+      if (part.type === 'resource') return part.text || part.name || part.uri
       return ''
     })
     .filter((value) => value.length > 0)
     .join('\n')
 }
+
+function getUriDisplayName(uri: string, fallback: string): string {
+  try {
+    const pathname = uri.startsWith('file://') ? new URL(uri).pathname : uri
+    return pathname.split(/[\\/]/).filter(Boolean).pop() || fallback
+  } catch {
+    return uri.split(/[\\/]/).filter(Boolean).pop() || fallback
+  }
+}
+
+function normalizeAnnotationsProperty(value: unknown): { annotations?: ContentAnnotations } {
+  if (!value || typeof value !== 'object') return {}
+  const record = value as Record<string, unknown>
+  const annotations: ContentAnnotations = {}
+  if (Array.isArray(record.audience)) {
+    const audience = record.audience.filter((item): item is 'user' | 'assistant' => item === 'user' || item === 'assistant')
+    if (audience.length > 0) annotations.audience = audience
+  }
+  if (typeof record.priority === 'number' && Number.isFinite(record.priority)) {
+    annotations.priority = record.priority
+  }
+  return Object.keys(annotations).length > 0 ? { annotations } : {}
+}

package/src/providers/provider-input-support.ts

@@ -1,8 +1,38 @@
 import type { InputEnvelope, ProviderModule } from './contracts.js'
 
-type InputMediaType = 'text' | 'image' | 'audio' | 'video' | 'resource'
+export type InputMediaType = 'text' | 'image' | 'audio' | 'video' | 'resource'
+
+export type InputAttachmentStrategy =
+  | 'native'
+  | 'native_acp'
+  | 'resource_link'
+  | 'text_fallback'
+  | 'paste'
+  | 'upload'
+
+export interface InputMediaStrategyDescriptor {
+  mediaType: InputMediaType
+  strategies: InputAttachmentStrategy[]
+  native?: boolean
+  degradation?: InputAttachmentStrategy[]
+}
+
+export interface MessageInputSupport {
+  text: boolean
+  multipart: boolean
+  mediaTypes: InputMediaType[]
+  strategies: InputMediaStrategyDescriptor[]
+}
 
 const VALID_INPUT_MEDIA_TYPES = new Set<InputMediaType>(['text', 'image', 'audio', 'video', 'resource'])
+const VALID_INPUT_STRATEGIES = new Set<InputAttachmentStrategy>(['native', 'native_acp', 'resource_link', 'text_fallback', 'paste', 'upload'])
+
+export const TEXT_ONLY_MESSAGE_INPUT_SUPPORT: MessageInputSupport = Object.freeze<MessageInputSupport>({
+  text: true,
+  multipart: false,
+  mediaTypes: ['text'],
+  strategies: [],
+})
 
 function getProviderLabel(provider?: Pick<ProviderModule, 'name' | 'type'> | null): string {
   return provider?.name || provider?.type || 'This provider'
@@ -44,14 +74,108 @@ export function assertTextOnlyInput(provider: Pick<ProviderModule, 'name' | 'typ
 export function getDeclaredProviderInputSupport(provider?: Pick<ProviderModule, 'capabilities'> | null): {
   multipart: boolean
   mediaTypes: Set<InputMediaType>
+  strategies: InputMediaStrategyDescriptor[]
 } {
   const rawMediaTypes = Array.isArray(provider?.capabilities?.input?.mediaTypes)
     ? provider?.capabilities?.input?.mediaTypes.filter((type): type is InputMediaType => VALID_INPUT_MEDIA_TYPES.has(type as InputMediaType))
     : []
+  const strategies = normalizeInputStrategyDescriptors(provider?.capabilities?.input?.strategies)
 
   return {
     multipart: provider?.capabilities?.input?.multipart === true,
     mediaTypes: new Set<InputMediaType>(rawMediaTypes.length > 0 ? rawMediaTypes : ['text']),
+    strategies,
+  }
+}
+
+export function normalizeInputStrategyDescriptors(raw: unknown): InputMediaStrategyDescriptor[] {
+  if (!Array.isArray(raw)) return []
+  const result: InputMediaStrategyDescriptor[] = []
+  for (const entry of raw) {
+    if (!entry || typeof entry !== 'object') continue
+    const record = entry as Record<string, unknown>
+    const mediaType = record.mediaType
+    if (typeof mediaType !== 'string' || !VALID_INPUT_MEDIA_TYPES.has(mediaType as InputMediaType)) continue
+    const strategies = Array.isArray(record.strategies)
+      ? record.strategies.filter((value): value is InputAttachmentStrategy => typeof value === 'string' && VALID_INPUT_STRATEGIES.has(value as InputAttachmentStrategy))
+      : []
+    const degradation = Array.isArray(record.degradation)
+      ? record.degradation.filter((value): value is InputAttachmentStrategy => typeof value === 'string' && VALID_INPUT_STRATEGIES.has(value as InputAttachmentStrategy))
+      : []
+    if (strategies.length === 0 && degradation.length === 0) continue
+    result.push({
+      mediaType: mediaType as InputMediaType,
+      strategies,
+      ...(typeof record.native === 'boolean' ? { native: record.native } : {}),
+      ...(degradation.length > 0 ? { degradation } : {}),
+    })
+  }
+  return result
+}
+
+function promptCapabilityFlags(runtimeCapabilities?: Record<string, any> | null): { image: boolean; audio: boolean; embeddedContext: boolean } {
+  const prompt = runtimeCapabilities?.promptCapabilities || {}
+  return {
+    image: prompt.image === true,
+    audio: prompt.audio === true,
+    embeddedContext: prompt.embeddedContext === true,
+  }
+}
+
+function supportFromDeclared(provider?: Pick<ProviderModule, 'capabilities'> | null): MessageInputSupport {
+  const declared = getDeclaredProviderInputSupport(provider)
+  return {
+    text: true,
+    multipart: declared.multipart,
+    mediaTypes: Array.from(declared.mediaTypes),
+    strategies: declared.strategies,
+  }
+}
+
+export function getEffectiveMessageInputSupport(
+  provider?: Pick<ProviderModule, 'category' | 'capabilities'> | null,
+  runtimeCapabilities?: Record<string, any> | null,
+): MessageInputSupport {
+  if (provider?.category !== 'acp') {
+    const declared = supportFromDeclared(provider)
+    return {
+      ...declared,
+      mediaTypes: [...declared.mediaTypes],
+      strategies: declared.strategies.map((strategy) => ({
+        ...strategy,
+        strategies: [...strategy.strategies],
+        ...(strategy.degradation ? { degradation: [...strategy.degradation] } : {}),
+      })),
+    }
+  }
+
+  const declared = supportFromDeclared(provider)
+  const caps = promptCapabilityFlags(runtimeCapabilities)
+  const mediaTypes = new Set<InputMediaType>(['text'])
+  const strategies: InputMediaStrategyDescriptor[] = []
+
+  if (declared.mediaTypes.includes('resource')) {
+    mediaTypes.add('resource')
+    strategies.push({ mediaType: 'resource', strategies: caps.embeddedContext ? ['native_acp', 'resource_link', 'text_fallback'] : ['resource_link', 'text_fallback'], native: caps.embeddedContext, degradation: ['resource_link', 'text_fallback'] })
+  }
+  if (declared.mediaTypes.includes('video')) {
+    mediaTypes.add('video')
+    strategies.push({ mediaType: 'video', strategies: ['resource_link', 'text_fallback'], native: false, degradation: ['resource_link', 'text_fallback'] })
+  }
+  if (declared.mediaTypes.includes('image')) {
+    mediaTypes.add('image')
+    strategies.push({ mediaType: 'image', strategies: caps.image ? ['native_acp', 'resource_link', 'text_fallback'] : ['resource_link', 'text_fallback'], native: caps.image, degradation: ['resource_link', 'text_fallback'] })
+  }
+  if (declared.mediaTypes.includes('audio')) {
+    mediaTypes.add('audio')
+    strategies.push({ mediaType: 'audio', strategies: caps.audio ? ['native_acp', 'resource_link', 'text_fallback'] : ['resource_link', 'text_fallback'], native: caps.audio, degradation: ['resource_link', 'text_fallback'] })
+  }
+
+  return {
+    text: true,
+    multipart: declared.multipart && mediaTypes.size > 1,
+    mediaTypes: Array.from(mediaTypes),
+    strategies,
   }
 }
 

package/src/providers/provider-instance.ts

@@ -10,6 +10,7 @@
 
 import type { ProviderModule, ProviderSettingDef, ProviderResumeCapability } from './contracts.js';
 import type { AcpConfigOption, AcpMode, ProviderControlSchema, ProviderSummaryMetadata, SessionCapability } from '../shared-types.js';
+import type { MessageInputSupport } from './provider-input-support.js';
 import type { ChatMessage } from '../types.js';
 
 // ─── ProviderState — Discriminated union by category ─────────────
@@ -87,6 +88,7 @@ interface ProviderStateBase {
     runtime?: ProviderRuntimeInfo;
     resume?: ProviderResumeCapability;
     sessionCapabilities?: SessionCapability[];
+    messageInput?: MessageInputSupport;
  /** Dynamic control current values */
     controlValues?: Record<string, string | number | boolean>;
     /** Provider-declared controls schema (from provider.controls) */

package/src/providers/provider-schema.ts

@@ -2,6 +2,7 @@ import type { ProviderControlDef, ProviderControlType, ProviderModule } from './
 import { providerHasOpenPanelSupport } from './open-panel-support.js'
 
 const VALID_CAPABILITY_MEDIA_TYPES = new Set(['text', 'image', 'audio', 'video', 'resource'])
+const VALID_INPUT_STRATEGIES = new Set(['native', 'native_acp', 'resource_link', 'text_fallback', 'paste', 'upload'])
 
 const KNOWN_PROVIDER_FIELDS = new Set<string>([
   'type',
@@ -158,16 +159,45 @@ function validateCapabilities(provider: ProviderModule, controls: ProviderContro
   }
 
   const input = capabilities.input
-  if (!input || typeof input !== 'object') {
-    errors.push('capabilities.input is required')
-  } else {
-    if (typeof input.multipart !== 'boolean') {
+  if (input !== undefined) {
+    if (!input || typeof input !== 'object') {
+      errors.push('capabilities.input must be an object when provided')
+    } else if (typeof input.multipart !== 'boolean') {
       errors.push('capabilities.input.multipart must be boolean')
     }
-    if (!Array.isArray(input.mediaTypes) || input.mediaTypes.length === 0) {
-      errors.push('capabilities.input.mediaTypes must be a non-empty array')
-    } else if (input.mediaTypes.some((type) => typeof type !== 'string' || !VALID_CAPABILITY_MEDIA_TYPES.has(type))) {
-      errors.push(`capabilities.input.mediaTypes must only include: ${Array.from(VALID_CAPABILITY_MEDIA_TYPES).join(', ')}`)
+    if (input && typeof input === 'object') {
+      const mediaTypes = Array.isArray(input.mediaTypes) ? input.mediaTypes : undefined
+      if (!mediaTypes || mediaTypes.length === 0) {
+        errors.push('capabilities.input.mediaTypes must be a non-empty array')
+      } else if (mediaTypes.some((type) => typeof type !== 'string' || !VALID_CAPABILITY_MEDIA_TYPES.has(type))) {
+        errors.push(`capabilities.input.mediaTypes must only include: ${Array.from(VALID_CAPABILITY_MEDIA_TYPES).join(', ')}`)
+      }
+    }
+    if (input && typeof input === 'object' && input.strategies !== undefined) {
+      if (!Array.isArray(input.strategies)) {
+        errors.push('capabilities.input.strategies must be an array when provided')
+      } else {
+        for (const strategy of input.strategies) {
+          if (!strategy || typeof strategy !== 'object' || Array.isArray(strategy)) {
+            errors.push('capabilities.input.strategies entries must be objects')
+            continue
+          }
+          const entry = strategy as Record<string, unknown>
+          if (typeof entry.mediaType !== 'string' || !VALID_CAPABILITY_MEDIA_TYPES.has(entry.mediaType)) {
+            errors.push(`capabilities.input.strategies.mediaType must only include: ${Array.from(VALID_CAPABILITY_MEDIA_TYPES).join(', ')}`)
+          }
+          for (const field of ['strategies', 'degradation'] as const) {
+            const values = entry[field]
+            if (values === undefined) continue
+            if (!Array.isArray(values) || values.some((value) => typeof value !== 'string' || !VALID_INPUT_STRATEGIES.has(value))) {
+              errors.push(`capabilities.input.strategies.${field} must only include: ${Array.from(VALID_INPUT_STRATEGIES).join(', ')}`)
+            }
+          }
+          if (entry.native !== undefined && typeof entry.native !== 'boolean') {
+            errors.push('capabilities.input.strategies.native must be boolean when provided')
+          }
+        }
+      }
     }
   }
 

package/src/shared-types.ts

@@ -336,7 +336,9 @@ export type SessionCapability =
     | 'mark_notification_unread';
 
 import type { RuntimeWriteOwner, RuntimeAttachedClient, SessionStatus } from './shared-types-extra.js';
+import type { MessageInputSupport } from './providers/provider-input-support.js';
 export type { RuntimeWriteOwner, RuntimeAttachedClient, SessionStatus } from './shared-types-extra.js';
+export type { MessageInputSupport, InputMediaStrategyDescriptor, InputAttachmentStrategy, InputMediaType } from './providers/provider-input-support.js';
 
 export interface SessionEntry {
     id: string;
@@ -364,6 +366,8 @@ export interface SessionEntry {
     resume?: ProviderResumeCapability;
     activeChat: SessionActiveChatData | null;
     capabilities?: SessionCapability[];
+    /** Effective message input/media support for this session. Defaults fail-closed to text-only. */
+    messageInput?: MessageInputSupport;
     cdpConnected?: boolean;
     /** Dynamic control current values (generic key-value) */
     controlValues?: Record<string, string | number | boolean>;

package/src/status/builders.ts

@@ -30,6 +30,7 @@ import {
     IDE_PROVIDER_SESSION_CAPABILITIES_BASE,
     EXTENSION_PROVIDER_SESSION_CAPABILITIES_BASE,
 } from '../providers/open-panel-support.js';
+import { TEXT_ONLY_MESSAGE_INPUT_SUPPORT } from '../providers/provider-input-support.js';
 
 export type SessionEntryProfile = 'full' | 'live' | 'metadata';
 
@@ -189,7 +190,7 @@ function buildIdeWorkspaceSession(
         ...(git && { git }),
         activeChat,
         ...(summaryMetadata && { summaryMetadata }),
-        ...(includeSessionMetadata && { capabilities: state.sessionCapabilities || IDE_SESSION_CAPABILITIES }),
+        ...(includeSessionMetadata && { capabilities: state.sessionCapabilities || IDE_SESSION_CAPABILITIES, messageInput: state.messageInput || TEXT_ONLY_MESSAGE_INPUT_SUPPORT }),
         cdpConnected: state.cdpConnected ?? isCdpConnected(cdpManagers, state.type),
         ...(includeSessionControls && {
             ...(controlValues && { controlValues }),
@@ -229,7 +230,7 @@ function buildExtensionAgentSession(
         ...(git && { git }),
         activeChat,
         ...(summaryMetadata && { summaryMetadata }),
-        ...(includeSessionMetadata && { capabilities: ext.sessionCapabilities || EXTENSION_SESSION_CAPABILITIES }),
+        ...(includeSessionMetadata && { capabilities: ext.sessionCapabilities || EXTENSION_SESSION_CAPABILITIES, messageInput: ext.messageInput || TEXT_ONLY_MESSAGE_INPUT_SUPPORT }),
         ...(includeSessionControls && {
             ...(controlValues && { controlValues }),
             providerControls: ext.providerControls,
@@ -307,6 +308,7 @@ function buildCliSession(state: CliProviderState, options: SessionEntryBuildOpti
         ...(summaryMetadata && { summaryMetadata }),
         ...(includeSessionMetadata && {
             capabilities: state.mode === 'terminal' ? PTY_SESSION_CAPABILITIES : CLI_CHAT_SESSION_CAPABILITIES,
+            messageInput: state.messageInput || TEXT_ONLY_MESSAGE_INPUT_SUPPORT,
         }),
         ...(includeSessionControls && {
             ...(controlValues && { controlValues }),
@@ -341,7 +343,7 @@ function buildAcpSession(state: AcpProviderState, options: SessionEntryBuildOpti
         ...(git && { git }),
         activeChat,
         ...(summaryMetadata && { summaryMetadata }),
-        ...(includeSessionMetadata && { capabilities: ACP_SESSION_CAPABILITIES }),
+        ...(includeSessionMetadata && { capabilities: ACP_SESSION_CAPABILITIES, messageInput: state.messageInput || TEXT_ONLY_MESSAGE_INPUT_SUPPORT }),
         ...(includeSessionControls && {
             ...(controlValues && { controlValues }),
             providerControls: state.providerControls,