@botbotgo/agent-harness 0.0.287 → 0.0.288

package/README.md

@@ -143,8 +143,8 @@ Additional docs:
 
 The public API spans a full product runtime—persistent records, memory and evidence, protocol surfaces, and governance—not only a thin bootstrap around YAML and tools.
 
-- **Core runtime API:** `createAgentHarness`, `request`, `subscribe`, `resolveApproval`, inspection helpers, and stable persisted runtime records for `requests`, `sessions`, `approvals`, `events`, and artifacts.
-- **Runtime memory and evidence:** `memorize`, `recall`, `listMemories`, memory policy hooks, `listArtifacts`, `getArtifact`, `exportEvaluationBundle`, `replayEvaluationBundle`, and request/session evidence export helpers.
+- **Core runtime API:** `createAgentHarness`, `request`, `subscribe`, `resolveApproval`, `recordArtifact`, inspection helpers, and stable persisted runtime records for `requests`, `sessions`, `approvals`, `events`, and artifacts.
+- **Runtime memory and evidence:** `memorize`, `recall`, `listMemories`, memory policy hooks, `recordArtifact`, `listArtifacts`, `getArtifact`, `exportEvaluationBundle`, `replayEvaluationBundle`, and request/session evidence export helpers.
 - **Protocol and transport surfaces:** `createAcpServer`, `createAcpStdioClient`, `serveAcpStdio`, `serveAcpHttp`, `serveA2aHttp`, `serveAgUiHttp`, and `createRuntimeMcpServer` / `serveRuntimeMcpOverStdio`.
 - **Governed workspace runtime:** YAML-owned routing, concurrency, maintenance, MCP policy, runtime governance bundles, and approval defaults for sensitive memory or write-like MCP side effects.
 - **Policy-shaped approvals:** governed tools can stay on manual review, auto-approve, or auto-reject / deny-and-continue modes while the runtime keeps one inspectable governance decision surface.

package/README.zh.md

@@ -139,8 +139,8 @@ try {
 
 若你想先看「今天能直接用到什么」，可从本节读起。`agent-harness` 提供完整的产品级运行时能力，而不只是「能启动」的脚手架。
 
-- **核心 runtime API：** `createAgentHarness`、`request`、`subscribe`、`resolveApproval`、各类查询与检查辅助 API，以及稳定持久化的 `requests`、`sessions`、`approvals`、`events` 与 artifacts 记录。
-- **运行时 memory 与证据能力：** `memorize`、`recall`、`listMemories`、memory policy hooks、`listArtifacts`、`getArtifact`、`exportEvaluationBundle`、`replayEvaluationBundle`，以及 request / session 级证据导出辅助函数。
+- **核心 runtime API：** `createAgentHarness`、`request`、`subscribe`、`resolveApproval`、`recordArtifact`、各类查询与检查辅助 API，以及稳定持久化的 `requests`、`sessions`、`approvals`、`events` 与 artifacts 记录。
+- **运行时 memory 与证据能力：** `memorize`、`recall`、`listMemories`、memory policy hooks、`recordArtifact`、`listArtifacts`、`getArtifact`、`exportEvaluationBundle`、`replayEvaluationBundle`，以及 request / session 级证据导出辅助函数。
 - **协议与传输层：** `createAcpServer`、`createAcpStdioClient`、`serveAcpStdio`、`serveAcpHttp`、`serveA2aHttp`、`serveAgUiHttp`，以及 `createRuntimeMcpServer` / `serveRuntimeMcpOverStdio`。
 - **受治理的工作区运行时：** 由 YAML 持有的路由、并发、维护、MCP 策略、runtime governance bundles，以及针对敏感 memory 或写类 MCP 副作用的默认审批门槛。
 - **策略化审批：** 受治理工具现在既可以走人工审批，也可以走 `auto-approve`、`auto-reject` 或 `deny-and-continue`，同时继续保留统一可检查的治理决策面。

package/dist/api.d.ts

@@ -1,4 +1,4 @@
-import type { ArtifactListing, CancelOptions, InvocationEnvelope, ListMemoriesInput, ListMemoriesResult, MemoryRecord, MemorizeInput, MemorizeResult, MessageContent, RecallInput, RecallResult, RemoveMemoryInput, RequestRecord, RequestSummary, ResumeOptions, RunDecisionOptions, RunListeners, RunResult, RunStartOptions, RuntimeHealthSnapshot, RuntimeGovernanceEvidence, RuntimeGovernanceDiagnostics, RuntimeOperatorOverview, RuntimeQueueDiagnostics, RuntimeAdapterOptions, RuntimeEvaluationExport, RuntimeEvaluationExportInput, RuntimeEvaluationReplayInput, RuntimeEvaluationReplayResult as InternalRuntimeEvaluationReplayResult, RuntimeSessionPackage, RuntimeSessionPackageInput, SessionListSummary, SessionRecord, SessionSummary, TranscriptMessage, UpdateMemoryInput, WorkspaceLoadOptions } from "./contracts/types.js";
+import type { ArtifactListing, CancelOptions, InvocationEnvelope, ListMemoriesInput, ListMemoriesResult, MemoryRecord, MemorizeInput, MemorizeResult, MessageContent, RecallInput, RecallResult, RemoveMemoryInput, RequestRecord, RequestSummary, RuntimeArtifactWriteInput, ResumeOptions, RunDecisionOptions, RunListeners, RunResult, RunStartOptions, RuntimeHealthSnapshot, RuntimeGovernanceEvidence, RuntimeGovernanceDiagnostics, RuntimeOperatorOverview, RuntimeQueueDiagnostics, RuntimeAdapterOptions, RuntimeEvaluationExport, RuntimeEvaluationExportInput, RuntimeEvaluationArtifact, RuntimeEvaluationReplayInput, RuntimeEvaluationReplayResult as InternalRuntimeEvaluationReplayResult, RuntimeSessionPackage, RuntimeSessionPackageInput, SessionListSummary, SessionRecord, SessionSummary, TranscriptMessage, UpdateMemoryInput, WorkspaceLoadOptions } from "./contracts/types.js";
 import { AgentHarnessRuntime } from "./runtime/harness.js";
 import type { InventoryAgentRecord, InventorySkillRecord } from "./runtime/harness/system/inventory.js";
 import type { RequirementAssessmentOptions } from "./runtime/harness/system/skill-requirements.js";
@@ -72,10 +72,11 @@ export type RequestPackage = {
     approvals: Approval[];
     transcript: TranscriptMessage[];
     events: RequestEvent[];
-    artifacts: RequestArtifactListing["items"];
+    artifacts: RuntimeEvaluationArtifact[];
     governance: RuntimeGovernanceEvidence;
     runtimeHealth?: RuntimeHealthSnapshot;
 };
+export type RecordArtifactInput = RuntimeArtifactWriteInput;
 export type RuntimeEvaluationReplayResult = Omit<InternalRuntimeEvaluationReplayResult, "result"> & {
     result: PublicRunResult;
 };
@@ -162,6 +163,7 @@ export declare function getArtifact(runtime: AgentHarnessRuntime, input: {
     requestId: string;
     artifactPath: string;
 }): Promise<unknown>;
+export declare function recordArtifact(runtime: AgentHarnessRuntime, input: RecordArtifactInput): Promise<import("./contracts/types.js").ArtifactRecord>;
 export declare function listRequestEvents(runtime: AgentHarnessRuntime, input: {
     sessionId: string;
     requestId: string;

package/dist/api.js

@@ -285,6 +285,9 @@ export async function listArtifacts(runtime, input) {
 export async function getArtifact(runtime, input) {
     return runtime.readArtifact(input.sessionId, input.requestId, input.artifactPath);
 }
+export async function recordArtifact(runtime, input) {
+    return runtime.recordArtifact(input);
+}
 export async function listRequestEvents(runtime, input) {
     return (await runtime.listRequestEvents(input.sessionId, input.requestId)).map(toPublicEvent);
 }

package/dist/contracts/runtime.d.ts

@@ -699,8 +699,31 @@ export type RuntimeEvaluationExportInput = {
     tags?: string[];
     metadata?: Record<string, unknown>;
 };
+export type RuntimeArtifactWriteInput = {
+    sessionId: string;
+    requestId: string;
+    kind: string;
+    path: string;
+    content: unknown;
+    artifactId?: string;
+    createdAt?: string;
+};
+export type RuntimeEvaluationArtifactFormat = "json" | "markdown" | "text" | "unknown";
+export type RuntimeEvaluationArtifactRole = "product-spec" | "sprint-contract" | "qa-report" | "handoff" | "other";
 export type RuntimeEvaluationArtifact = ArtifactRecord & {
     content?: unknown;
+    format: RuntimeEvaluationArtifactFormat;
+    role: RuntimeEvaluationArtifactRole;
+};
+export type RuntimeEvaluationArtifactSummary = {
+    productSpecPaths: string[];
+    sprintContractPaths: string[];
+    qaReportPaths: string[];
+    handoffPaths: string[];
+    qaVerdicts: Array<{
+        path: string;
+        verdict: string;
+    }>;
 };
 export type RuntimeEvaluationExport = {
     session: SessionRecord | null;
@@ -714,6 +737,7 @@ export type RuntimeEvaluationExport = {
     transcript: TranscriptMessage[];
     events: HarnessEvent[];
     artifacts: RuntimeEvaluationArtifact[];
+    artifactSummary: RuntimeEvaluationArtifactSummary;
     runtimeHealth: RuntimeHealthSnapshot;
     expectedOutput?: string;
     rubric: string[];

package/dist/index.d.ts

@@ -1,7 +1,7 @@
-export { AgentHarnessAcpServer, AgentHarnessRuntime, cancelRun, createAgentHarness, createAcpServer, createAcpStdioClient, createRuntimeMcpServer, createUpstreamTimelineReducer, createToolMcpServer, deleteSession, describeInventory, exportEvaluationBundle, exportFlow, exportSequence, exportRequestPackage, exportSessionPackage, replayEvaluationBundle, getArtifact, getAgent, getApproval, getOperatorOverview, getRequest, getHealth, listMemories, listRequestTraceItems, getSession, listAgentSkills, listArtifacts, listApprovals, listRequests, listRequestEvents, listSessionSummaries, listSessions, memorize, normalizeUserChatInput, request, recall, removeMemory, resolveApproval, serveA2aHttp, serveAcpHttp, serveAcpStdio, serveAgUiHttp, serveRuntimeMcpOverStdio, serveToolsOverStdio, subscribe, stop, updateMemory, } from "./api.js";
+export { AgentHarnessAcpServer, AgentHarnessRuntime, cancelRun, createAgentHarness, createAcpServer, createAcpStdioClient, createRuntimeMcpServer, createUpstreamTimelineReducer, createToolMcpServer, deleteSession, describeInventory, exportEvaluationBundle, exportFlow, exportSequence, exportRequestPackage, exportSessionPackage, replayEvaluationBundle, getArtifact, getAgent, getApproval, getOperatorOverview, getRequest, getHealth, listMemories, listRequestTraceItems, getSession, listAgentSkills, listArtifacts, listApprovals, listRequests, listRequestEvents, listSessionSummaries, listSessions, memorize, normalizeUserChatInput, recordArtifact, request, recall, removeMemory, resolveApproval, serveA2aHttp, serveAcpHttp, serveAcpStdio, serveAgUiHttp, serveRuntimeMcpOverStdio, serveToolsOverStdio, subscribe, stop, updateMemory, } from "./api.js";
 export { createKnowledgeModule, readKnowledgeRuntimeConfig } from "./knowledge/index.js";
 export type { AcpApproval, AcpArtifact, AcpEventNotification, AcpJsonRpcError, AcpJsonRpcRequest, AcpJsonRpcResponse, AcpJsonRpcSuccess, AcpRequestRecord, AcpRunRequestParams, AcpServerCapabilities, AcpSessionRecord, AcpStdioClient, AcpStdioClientOptions, } from "./acp.js";
-export type { Approval, ListMemoriesInput, ListMemoriesResult, MemoryDecision, MemoryKind, MemoryRecord, MemoryScope, MemorizeInput, MemorizeResult, NormalizeUserChatInputOptions, OperatorOverview, PublicRunListeners, RequestArtifactListing, RequestEvent, RequestEventType, RequestPackage, RequestPackageInput, RequestFlowGraphInput, RequestResult, RequestTraceItem, RecallInput, RecallResult, RemoveMemoryInput, RuntimeEvaluationExport, RuntimeEvaluationExportInput, RuntimeEvaluationReplayInput, RuntimeEvaluationReplayResult, SessionListSummary, RuntimeSessionPackage, RuntimeSessionPackageInput, UpdateMemoryInput, UserChatInput, UserChatMessage, } from "./api.js";
+export type { Approval, ListMemoriesInput, ListMemoriesResult, MemoryDecision, MemoryKind, MemoryRecord, MemoryScope, MemorizeInput, MemorizeResult, NormalizeUserChatInputOptions, OperatorOverview, PublicRunListeners, RecordArtifactInput, RequestArtifactListing, RequestEvent, RequestEventType, RequestPackage, RequestPackageInput, RequestFlowGraphInput, RequestResult, RequestTraceItem, RecallInput, RecallResult, RemoveMemoryInput, RuntimeEvaluationExport, RuntimeEvaluationExportInput, RuntimeEvaluationReplayInput, RuntimeEvaluationReplayResult, SessionListSummary, RuntimeSessionPackage, RuntimeSessionPackageInput, UpdateMemoryInput, UserChatInput, UserChatMessage, } from "./api.js";
 export type { KnowledgeListInput, KnowledgeMemorizeInput, KnowledgeModule, KnowledgeModuleDependencies, KnowledgeRecallInput, KnowledgeRuntimeConfig, KnowledgeRuntimeContext, } from "./knowledge/index.js";
 export type { A2aAgentCard, A2aHttpServer, A2aHttpServerOptions, A2aTask, A2aTaskState, AcpHttpServer, AcpHttpServerOptions, AcpStdioServer, AcpStdioServerOptions, AgUiEvent, AgUiHttpServer, AgUiHttpServerOptions, AgUiRunAgentInput, } from "./api.js";
 export type { RuntimeMcpServerOptions, ToolMcpServerOptions } from "./mcp.js";

package/dist/index.js

@@ -1,3 +1,3 @@
-export { AgentHarnessAcpServer, AgentHarnessRuntime, cancelRun, createAgentHarness, createAcpServer, createAcpStdioClient, createRuntimeMcpServer, createUpstreamTimelineReducer, createToolMcpServer, deleteSession, describeInventory, exportEvaluationBundle, exportFlow, exportSequence, exportRequestPackage, exportSessionPackage, replayEvaluationBundle, getArtifact, getAgent, getApproval, getOperatorOverview, getRequest, getHealth, listMemories, listRequestTraceItems, getSession, listAgentSkills, listArtifacts, listApprovals, listRequests, listRequestEvents, listSessionSummaries, listSessions, memorize, normalizeUserChatInput, request, recall, removeMemory, resolveApproval, serveA2aHttp, serveAcpHttp, serveAcpStdio, serveAgUiHttp, serveRuntimeMcpOverStdio, serveToolsOverStdio, subscribe, stop, updateMemory, } from "./api.js";
+export { AgentHarnessAcpServer, AgentHarnessRuntime, cancelRun, createAgentHarness, createAcpServer, createAcpStdioClient, createRuntimeMcpServer, createUpstreamTimelineReducer, createToolMcpServer, deleteSession, describeInventory, exportEvaluationBundle, exportFlow, exportSequence, exportRequestPackage, exportSessionPackage, replayEvaluationBundle, getArtifact, getAgent, getApproval, getOperatorOverview, getRequest, getHealth, listMemories, listRequestTraceItems, getSession, listAgentSkills, listArtifacts, listApprovals, listRequests, listRequestEvents, listSessionSummaries, listSessions, memorize, normalizeUserChatInput, recordArtifact, request, recall, removeMemory, resolveApproval, serveA2aHttp, serveAcpHttp, serveAcpStdio, serveAgUiHttp, serveRuntimeMcpOverStdio, serveToolsOverStdio, subscribe, stop, updateMemory, } from "./api.js";
 export { createKnowledgeModule, readKnowledgeRuntimeConfig } from "./knowledge/index.js";
 export { tool } from "./tools.js";

package/dist/package-version.d.ts

@@ -1 +1 @@
-export declare const AGENT_HARNESS_VERSION = "0.0.286";
+export declare const AGENT_HARNESS_VERSION = "0.0.287";

package/dist/package-version.js

@@ -1 +1 @@
-export const AGENT_HARNESS_VERSION = "0.0.286";
+export const AGENT_HARNESS_VERSION = "0.0.287";

package/dist/runtime/harness/run/evaluation-artifacts.d.ts

@@ -0,0 +1,8 @@
+import type { RuntimeEvaluationArtifact, RuntimeEvaluationArtifactSummary } from "../../../contracts/types.js";
+export declare function normalizeRecordedArtifactPath(input: string): string;
+export declare function enrichEvaluationArtifact<T extends {
+    path: string;
+    kind: string;
+    content?: unknown;
+}>(artifact: T): T & Pick<RuntimeEvaluationArtifact, "format" | "role">;
+export declare function summarizeEvaluationArtifacts(artifacts: RuntimeEvaluationArtifact[]): RuntimeEvaluationArtifactSummary;

package/dist/runtime/harness/run/evaluation-artifacts.js

@@ -0,0 +1,108 @@
+import path from "node:path";
+function isObject(value) {
+    return Boolean(value) && typeof value === "object" && !Array.isArray(value);
+}
+function inferArtifactFormat(artifactPath, content) {
+    const normalized = artifactPath.toLowerCase();
+    if (normalized.endsWith(".json")) {
+        return "json";
+    }
+    if (normalized.endsWith(".md") || normalized.endsWith(".markdown")) {
+        return "markdown";
+    }
+    if (normalized.endsWith(".txt")) {
+        return "text";
+    }
+    if (typeof content === "string") {
+        return content.includes("#") || content.includes("##") ? "markdown" : "text";
+    }
+    if (content !== undefined) {
+        return "json";
+    }
+    return "unknown";
+}
+function inferArtifactRole(artifactPath, kind, content) {
+    const normalized = artifactPath.toLowerCase();
+    const loweredKind = kind.toLowerCase();
+    const contentObject = isObject(content) ? content : null;
+    if (normalized.includes("product-spec") || normalized.endsWith("/spec.md") || loweredKind === "product-spec") {
+        return "product-spec";
+    }
+    if (normalized.includes("sprint-contract")
+        || loweredKind === "sprint-contract"
+        || (contentObject && Array.isArray(contentObject.acceptanceCriteria))) {
+        return "sprint-contract";
+    }
+    if (normalized.includes("qa-report")
+        || loweredKind === "qa-report"
+        || (contentObject && (typeof contentObject.verdict === "string" || typeof contentObject.status === "string"))) {
+        return "qa-report";
+    }
+    if (normalized.includes("handoff") || loweredKind === "handoff") {
+        return "handoff";
+    }
+    return "other";
+}
+function extractQaVerdict(content) {
+    if (!isObject(content)) {
+        return null;
+    }
+    for (const key of ["verdict", "status", "decision"]) {
+        if (typeof content[key] === "string" && String(content[key]).trim().length > 0) {
+            return String(content[key]).trim();
+        }
+    }
+    return null;
+}
+export function normalizeRecordedArtifactPath(input) {
+    const trimmed = input.trim().replaceAll("\\", "/").replace(/^\.\/+/, "");
+    if (!trimmed) {
+        throw new Error("Artifact path must be a non-empty relative path.");
+    }
+    if (trimmed.startsWith("/") || /^[a-z]:\//i.test(trimmed)) {
+        throw new Error("Artifact path must stay relative to the persisted run artifact directory.");
+    }
+    const normalized = path.posix.normalize(trimmed);
+    if (normalized === "." || normalized.startsWith("../") || normalized.includes("/../")) {
+        throw new Error("Artifact path must stay inside the persisted run artifact directory.");
+    }
+    return normalized;
+}
+export function enrichEvaluationArtifact(artifact) {
+    return {
+        ...artifact,
+        format: inferArtifactFormat(artifact.path, artifact.content),
+        role: inferArtifactRole(artifact.path, artifact.kind, artifact.content),
+    };
+}
+export function summarizeEvaluationArtifacts(artifacts) {
+    const summary = {
+        productSpecPaths: [],
+        sprintContractPaths: [],
+        qaReportPaths: [],
+        handoffPaths: [],
+        qaVerdicts: [],
+    };
+    for (const artifact of artifacts) {
+        if (artifact.role === "product-spec") {
+            summary.productSpecPaths.push(artifact.path);
+            continue;
+        }
+        if (artifact.role === "sprint-contract") {
+            summary.sprintContractPaths.push(artifact.path);
+            continue;
+        }
+        if (artifact.role === "qa-report") {
+            summary.qaReportPaths.push(artifact.path);
+            const verdict = extractQaVerdict(artifact.content);
+            if (verdict) {
+                summary.qaVerdicts.push({ path: artifact.path, verdict });
+            }
+            continue;
+        }
+        if (artifact.role === "handoff") {
+            summary.handoffPaths.push(artifact.path);
+        }
+    }
+    return summary;
+}

package/dist/runtime/harness.d.ts

@@ -1,4 +1,4 @@
-import type { ApprovalRecord, ArtifactListing, CancelOptions, HarnessEvent, HarnessStreamItem, RuntimeHealthSnapshot, RuntimeOperatorOverview, ListMemoriesInput, ListMemoriesResult, MessageContent, RemoveMemoryInput, RunRecord, RunStartOptions, RestartConversationOptions, RuntimeAdapterOptions, RuntimeEvaluationExport, RuntimeEvaluationExportInput, RuntimeEvaluationReplayInput, RuntimeEvaluationReplayResult, RuntimeRunPackage, RuntimeRunPackageInput, RuntimeSessionPackage, RuntimeSessionPackageInput, ResumeOptions, RunOptions, RunResult, RunSummary, MemoryRecord, MemorizeInput, MemorizeResult, RecallInput, RecallResult, UpdateMemoryInput, SessionSummary, ThreadSummary, ThreadRecord, SessionRecord, RequestRecord, RequestSummary, SessionListSummary, WorkspaceBundle } from "../contracts/types.js";
+import type { ApprovalRecord, ArtifactListing, CancelOptions, HarnessEvent, HarnessStreamItem, RuntimeHealthSnapshot, RuntimeOperatorOverview, ListMemoriesInput, ListMemoriesResult, MessageContent, RemoveMemoryInput, RunRecord, RunStartOptions, RestartConversationOptions, RuntimeAdapterOptions, RuntimeArtifactWriteInput, RuntimeEvaluationExport, RuntimeEvaluationExportInput, RuntimeEvaluationReplayInput, RuntimeEvaluationReplayResult, RuntimeRunPackage, RuntimeRunPackageInput, RuntimeSessionPackage, RuntimeSessionPackageInput, ResumeOptions, RunOptions, RunResult, RunSummary, MemoryRecord, MemorizeInput, MemorizeResult, RecallInput, RecallResult, UpdateMemoryInput, SessionSummary, ThreadSummary, ThreadRecord, SessionRecord, RequestRecord, RequestSummary, SessionListSummary, WorkspaceBundle } from "../contracts/types.js";
 import { type RuntimeMcpServerOptions, type ToolMcpServerOptions } from "../mcp.js";
 import { type InventoryAgentRecord, type InventorySkillRecord } from "./harness/system/inventory.js";
 import type { RequirementAssessmentOptions } from "./harness/system/skill-requirements.js";
@@ -102,6 +102,7 @@ export declare class AgentHarnessRuntime {
     getApproval(approvalId: string): Promise<ApprovalRecord | null>;
     listArtifacts(threadId: string, runId: string): Promise<ArtifactListing>;
     readArtifact(threadId: string, runId: string, artifactPath: string): Promise<unknown>;
+    recordArtifact(input: RuntimeArtifactWriteInput): Promise<import("../contracts/types.js").ArtifactRecord>;
     listRequestEvents(sessionId: string, requestId: string): Promise<HarnessEvent[]>;
     listRunEvents(threadId: string, runId: string): Promise<HarnessEvent[]>;
     exportRequestPackage(input: RuntimeRunPackageInput): Promise<RuntimeRunPackage>;

package/dist/runtime/harness.js

@@ -3,6 +3,7 @@ import { SqlitePersistence } from "../persistence/sqlite-store.js";
 import { createPersistentId } from "../utils/id.js";
 import { extractMessageText } from "../utils/message-content.js";
 import { AgentRuntimeAdapter } from "./agent-runtime-adapter.js";
+import { enrichEvaluationArtifact, normalizeRecordedArtifactPath, summarizeEvaluationArtifacts, } from "./harness/run/evaluation-artifacts.js";
 import { EventBus } from "./harness/events/event-bus.js";
 import { createBackgroundEventRuntime } from "./harness/background-runtime.js";
 import { PolicyEngine } from "./harness/system/policy-engine.js";
@@ -633,7 +634,28 @@ export class AgentHarnessRuntime {
         };
     }
     async readArtifact(threadId, runId, artifactPath) {
-        return this.persistence.readArtifact(threadId, runId, artifactPath);
+        return this.persistence.readArtifact(threadId, runId, normalizeRecordedArtifactPath(artifactPath));
+    }
+    async recordArtifact(input) {
+        const session = await this.getSessionRecord(input.sessionId);
+        if (!session) {
+            throw new Error(`Cannot record artifact for unknown session: ${input.sessionId}`);
+        }
+        const request = await this.getRequest(input.requestId);
+        if (!request || request.sessionId !== input.sessionId) {
+            throw new Error(`Cannot record artifact for unknown request: ${input.requestId}`);
+        }
+        const artifact = {
+            artifactId: typeof input.artifactId === "string" && input.artifactId.trim().length > 0
+                ? input.artifactId.trim()
+                : createPersistentId(),
+            kind: input.kind.trim(),
+            path: normalizeRecordedArtifactPath(input.path),
+            createdAt: typeof input.createdAt === "string" && input.createdAt.trim().length > 0
+                ? input.createdAt.trim()
+                : new Date().toISOString(),
+        };
+        return this.persistence.createArtifact(input.sessionId, input.requestId, artifact, input.content);
     }
     async listRequestEvents(sessionId, requestId) {
         return this.persistence.listRunEvents(sessionId, requestId);
@@ -650,7 +672,7 @@ export class AgentHarnessRuntime {
         const artifactsListing = input.includeArtifacts === false
             ? { items: [] }
             : await this.persistence.listArtifacts(input.sessionId, input.requestId);
-        const artifacts = await Promise.all(artifactsListing.items.map(async (artifact) => ({
+        const artifacts = await Promise.all(artifactsListing.items.map(async (artifact) => enrichEvaluationArtifact({
             ...artifact,
             ...(input.includeArtifactContents === true
                 ? { content: await this.persistence.readArtifact(input.sessionId, input.requestId, artifact.path) }
@@ -727,10 +749,12 @@ export class AgentHarnessRuntime {
             ? { items: [] }
             : await this.persistence.listArtifacts(input.sessionId, input.requestId);
         const artifacts = await Promise.all(artifactsListing.items.map(async (artifact) => ({
-            ...artifact,
-            ...(input.includeArtifactContents === true
-                ? { content: await this.persistence.readArtifact(input.sessionId, input.requestId, artifact.path) }
-                : {}),
+            ...enrichEvaluationArtifact({
+                ...artifact,
+                ...(input.includeArtifactContents === true
+                    ? { content: await this.persistence.readArtifact(input.sessionId, input.requestId, artifact.path) }
+                    : {}),
+            }),
         })));
         return {
             session: thread ? toSessionRecord(thread) : null,
@@ -746,6 +770,7 @@ export class AgentHarnessRuntime {
             transcript,
             events,
             artifacts,
+            artifactSummary: summarizeEvaluationArtifacts(artifacts),
             runtimeHealth,
             ...(typeof input.expectedOutput === "string" && input.expectedOutput.trim().length > 0
                 ? { expectedOutput: input.expectedOutput.trim() }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@botbotgo/agent-harness",
-  "version": "0.0.287",
+  "version": "0.0.288",
   "description": "Workspace runtime for multi-agent applications",
   "license": "MIT",
   "type": "module",