@demicodes/agent 0.8.0 → 0.9.1

package/dist/index.d.mts

@@ -2,7 +2,7 @@ import { A as AgentToolContext, B as estimateTranscriptBlockTokens, C as AgentSe
 import { a as AgentClient, i as AgentActionOptions, n as createWebSocketClientTransport, o as AgentClientListener, r as createWebSocketServerTransport, t as JsonWebSocket } from "./websocket-transport-DrsdY12Z.mjs";
 import { BashAuditEvent, BashEnvironment, BashEnvironmentOptions, CommandMetadataRecord, Host, ShellCommandStatus, ShellOutputChunk } from "@demicodes/shell";
 import { AgentProvider, Provider } from "@demicodes/provider";
-import { Block, Model, ModelSelection, ProviderErrorDiagnostics, QueuedMessage, SessionPhase, UserContentBlock } from "@demicodes/core";
+import { Block, Model, ModelMediaKind, ModelSelection, ProviderErrorDiagnostics, QueuedMessage, SessionPhase, UserContentBlock } from "@demicodes/core";
 //#region src/provider-stream-error.d.ts
 /** Error carrying a provider stream's normalized code and bounded diagnostics. */
 declare class ProviderStreamError extends Error {
@@ -272,7 +272,28 @@ interface ShellToolResultOptions {
   exposeCommandHandle?: boolean;
   /** Model receiving this result; gates binary-stream media attachment. */
   model?: Model;
+  /** Per-modality byte caps on attached media; unset modalities keep the defaults. */
+  maxMediaBytes?: Partial<Record<ModelMediaKind, number>>;
 }
+/**
+ * How many bytes of each modality a tool may hand to the model.
+ *
+ * One number cannot serve both, because bytes buy wildly different amounts of
+ * context per modality — measured against a frontier model, a KiB of video
+ * costs ~2 tokens where a KiB of image costs ~50. A cap generous enough to show
+ * a five-minute clip would let a single still eat a six-figure token budget.
+ *
+ * image (4 MiB): well past any sane still — a 4000x3000 PNG lands under it — so
+ *   crossing this line is a mistake, not a use case.
+ * video (16 MiB): roughly ten minutes at a viewing-grade encoding, and
+ *   deliberately under the ~20 MB inline-payload ceiling the major APIs enforce.
+ *   A larger cap buys no reach, only a rejection further downstream where the
+ *   reason is harder to read.
+ *
+ * Bytes are a proxy, not a budget: for video they track cost reasonably at a
+ * fixed encoding, but an image's real driver is its pixel dimensions.
+ */
+declare const DEFAULT_MAX_MEDIA_BYTES: Record<ModelMediaKind, number>;
 declare function shellPreviewBudgetTokens(contextWindow: number): number;
 declare function toShellToolResult(result: ShellCommandStatus, options?: ShellToolResultOptions): AgentToolInvokeResult;
 /** Character budget for a shell view's render window (tail-biased). */
@@ -301,4 +322,4 @@ interface ShellToolView {
 declare function finishShellToolResult<State>(environment: BashEnvironment, result: ShellCommandStatus, ctx: AgentToolInvokeContext<State>): Promise<AgentToolInvokeResult>;
 declare function shellCommandHandleRequired(result: ShellCommandStatus, budgetTokens: number): boolean;
 //#endregion
-export { AbortResult, AbortTarget, ActiveTurnPhase, AgentActionOptions, AgentClient, AgentClientListener, AgentClientTransport, AgentDisposeContext, AgentHarness, AgentHarnessContext, AgentHarnessRuntime, AgentHostContext, AgentLifecycleEvent, AgentMetadata, AgentPromptContext, AgentReferenceResolveContext, AgentServer, AgentServerOptions, AgentServerSessionOptions, AgentServerTransport, AgentSession, AgentSessionCheckpoint, AgentSessionOptions, AgentSessionParams, AgentSessionRestoreParams, AgentSessionStore, AgentSystemPromptContext, AgentTool, AgentToolContext, AgentToolInvokeContext, AgentToolInvokeResult, AgentTransport, AgentTransportBinding, ClientFrame, ClientSessionEvent, CompactionWindow, ConversationSummary, DEFAULT_TURN_RETRY_POLICY, DrainedTranscriptPatches, ExternalMutationReservation, InProcessTransportPair, JsonWebSocket, PrepareShell, PrepareShellContext, ProviderStreamError, RunCommandLineCommandNotRegisteredError, RunCommandLineOptions, RunCommandLineResult, RunCommandLineShellNotFoundError, RunCommandLineTimeoutError, SHELL_VIEW_MAX_CHARS, ServerFrame, SessionEvent, SessionEventListener, ShellCommandStatusLike, ShellToolResultOptions, ShellToolView, StandardAgentToolOptions, TranscriptLog, TranscriptOptions, TranscriptPatch, TurnRetryPolicy, applyTranscriptPatches, cloneBlocks, createInProcessTransportPair, createStandardAgentTools, createWebSocketClientTransport, createWebSocketServerTransport, estimateTranscriptBlockTokens, finishShellToolResult, isContextLengthExceeded, isRetryableCode, resolveRetryPolicy, retryDelayMs, shellCommandHandleRequired, shellPreviewBudgetTokens, toShellToolResult };
+export { AbortResult, AbortTarget, ActiveTurnPhase, AgentActionOptions, AgentClient, AgentClientListener, AgentClientTransport, AgentDisposeContext, AgentHarness, AgentHarnessContext, AgentHarnessRuntime, AgentHostContext, AgentLifecycleEvent, AgentMetadata, AgentPromptContext, AgentReferenceResolveContext, AgentServer, AgentServerOptions, AgentServerSessionOptions, AgentServerTransport, AgentSession, AgentSessionCheckpoint, AgentSessionOptions, AgentSessionParams, AgentSessionRestoreParams, AgentSessionStore, AgentSystemPromptContext, AgentTool, AgentToolContext, AgentToolInvokeContext, AgentToolInvokeResult, AgentTransport, AgentTransportBinding, ClientFrame, ClientSessionEvent, CompactionWindow, ConversationSummary, DEFAULT_MAX_MEDIA_BYTES, DEFAULT_TURN_RETRY_POLICY, DrainedTranscriptPatches, ExternalMutationReservation, InProcessTransportPair, JsonWebSocket, PrepareShell, PrepareShellContext, ProviderStreamError, RunCommandLineCommandNotRegisteredError, RunCommandLineOptions, RunCommandLineResult, RunCommandLineShellNotFoundError, RunCommandLineTimeoutError, SHELL_VIEW_MAX_CHARS, ServerFrame, SessionEvent, SessionEventListener, ShellCommandStatusLike, ShellToolResultOptions, ShellToolView, StandardAgentToolOptions, TranscriptLog, TranscriptOptions, TranscriptPatch, TurnRetryPolicy, applyTranscriptPatches, cloneBlocks, createInProcessTransportPair, createStandardAgentTools, createWebSocketClientTransport, createWebSocketServerTransport, estimateTranscriptBlockTokens, finishShellToolResult, isContextLengthExceeded, isRetryableCode, resolveRetryPolicy, retryDelayMs, shellCommandHandleRequired, shellPreviewBudgetTokens, toShellToolResult };

package/dist/index.mjs

@@ -2271,6 +2271,28 @@ function createStandardAgentTools(options) {
 function resolveEnvironment(source, ctx, handle) {
 	return typeof source === "function" ? source(ctx, handle) : source;
 }
+/**
+* How many bytes of each modality a tool may hand to the model.
+*
+* One number cannot serve both, because bytes buy wildly different amounts of
+* context per modality — measured against a frontier model, a KiB of video
+* costs ~2 tokens where a KiB of image costs ~50. A cap generous enough to show
+* a five-minute clip would let a single still eat a six-figure token budget.
+*
+* image (4 MiB): well past any sane still — a 4000x3000 PNG lands under it — so
+*   crossing this line is a mistake, not a use case.
+* video (16 MiB): roughly ten minutes at a viewing-grade encoding, and
+*   deliberately under the ~20 MB inline-payload ceiling the major APIs enforce.
+*   A larger cap buys no reach, only a rejection further downstream where the
+*   reason is harder to read.
+*
+* Bytes are a proxy, not a budget: for video they track cost reasonably at a
+* fixed encoding, but an image's real driver is its pixel dimensions.
+*/
+const DEFAULT_MAX_MEDIA_BYTES = {
+	image: 4 * 1024 * 1024,
+	video: 16 * 1024 * 1024
+};
 function shellPreviewBudgetTokens(contextWindow) {
 	return contextWindow >= LARGE_CONTEXT_THRESHOLD_TOKENS ? LARGE_CONTEXT_PREVIEW_TOKENS : SMALL_CONTEXT_PREVIEW_TOKENS;
 }
@@ -2280,7 +2302,10 @@ function toShellToolResult(result, options = {}) {
 		text: formatShellToolResult(result, options)
 	}];
 	if (result.status === "exited" && result.binaryStdout) {
-		const verdict = binaryStreamVerdict(result.binaryStdout, result.commandId, options.model);
+		const verdict = binaryStreamVerdict(result.binaryStdout, result.commandId, options.model, {
+			...DEFAULT_MAX_MEDIA_BYTES,
+			...options.maxMediaBytes
+		});
 		if (verdict.block) output.push(verdict.block);
 		if (verdict.note) output.push({
 			type: "text",
@@ -2352,12 +2377,14 @@ function tailChunkWindow(chunks, maxChars) {
 * magic matches the closed model-media set, the model accepts the type, and
 * the stream was not truncated; otherwise explain why nothing was attached.
 */
-function binaryStreamVerdict(binary, commandId, model) {
+function binaryStreamVerdict(binary, commandId, model, maxMediaBytes) {
 	const media = sniffModelMediaType(binary.data);
 	const binPath = `/@/commands/${commandId}/stdout.bin`;
-	if (binary.truncated) return { note: `Binary stdout was truncated at the output limit (${binary.data.length} of ${binary.totalBytes} bytes)${media ? ` and looks like ${media.mediaType}` : ""}; it was not attached. Redirect to a file or raise the output limit and re-run.` };
+	if (binary.truncated) return { note: `Binary stdout (${binary.totalBytes} bytes${media ? `, ${media.mediaType}` : ""}) exceeded the shell's ${binary.limitBytes}-byte binary limit (maxBinaryBytes) and was not attached; the raw bytes remain readable at ${binPath}. Produce a smaller version and re-run.` };
 	if (!media) return { note: `Binary stdout does not match any model-viewable media type; the raw bytes remain readable at ${binPath}.` };
 	if (!model || !modelAcceptsMediaType(model, media.mediaType)) return { note: `Binary stdout is ${media.mediaType}, which this model does not accept natively; the raw bytes remain readable at ${binPath}.` };
+	const cap = maxMediaBytes[media.kind];
+	if (binary.totalBytes > cap) return { note: `Binary stdout is ${media.mediaType} (${binary.totalBytes} bytes), over the ${cap}-byte ${media.kind} cap; it was not attached and the raw bytes remain readable at ${binPath}. Produce a smaller version — for video, fewer frames or a lower resolution — and re-run.` };
 	const source = {
 		mediaType: media.mediaType,
 		data: bytesToBase64(binary.data)
@@ -3312,4 +3339,4 @@ function createProviderMap(providers) {
 	return map;
 }
 //#endregion
-export { AgentClient, AgentServer, AgentSession, DEFAULT_TURN_RETRY_POLICY, ProviderStreamError, RunCommandLineCommandNotRegisteredError, RunCommandLineShellNotFoundError, RunCommandLineTimeoutError, SHELL_VIEW_MAX_CHARS, TranscriptLog, applyTranscriptPatches, cloneBlocks, createInProcessTransportPair, createStandardAgentTools, createWebSocketClientTransport, createWebSocketServerTransport, estimateTranscriptBlockTokens, finishShellToolResult, isContextLengthExceeded, isRetryableCode, resolveRetryPolicy, retryDelayMs, shellCommandHandleRequired, shellPreviewBudgetTokens, toShellToolResult };
+export { AgentClient, AgentServer, AgentSession, DEFAULT_MAX_MEDIA_BYTES, DEFAULT_TURN_RETRY_POLICY, ProviderStreamError, RunCommandLineCommandNotRegisteredError, RunCommandLineShellNotFoundError, RunCommandLineTimeoutError, SHELL_VIEW_MAX_CHARS, TranscriptLog, applyTranscriptPatches, cloneBlocks, createInProcessTransportPair, createStandardAgentTools, createWebSocketClientTransport, createWebSocketServerTransport, estimateTranscriptBlockTokens, finishShellToolResult, isContextLengthExceeded, isRetryableCode, resolveRetryPolicy, retryDelayMs, shellCommandHandleRequired, shellPreviewBudgetTokens, toShellToolResult };

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@demicodes/agent",
   "description": "Session runtime and transport-neutral client and server protocol for Demi.",
-  "version": "0.8.0",
+  "version": "0.9.1",
   "private": false,
   "type": "module",
   "exports": {
@@ -19,10 +19,10 @@
     }
   },
   "dependencies": {
-    "@demicodes/core": "^0.8.0",
-    "@demicodes/provider": "^0.8.0",
-    "@demicodes/shell": "^0.8.0",
-    "@demicodes/utils": "^0.8.0"
+    "@demicodes/core": "^0.9.1",
+    "@demicodes/provider": "^0.9.1",
+    "@demicodes/shell": "^0.9.1",
+    "@demicodes/utils": "^0.9.1"
   },
   "license": "Apache-2.0",
   "main": "./dist/index.mjs",