agent-inspect 3.5.4 → 4.0.0

package/CHANGELOG.md

@@ -1,5 +1,19 @@
 # Changelog
 
+## 4.0.0
+
+### Major Changes
+
+- dc4297b: Add the local trace workspace (v4.0): a project-local layout and manifest (`.agent-inspect/workspace.json`) with `workspace` CLI commands (`init`, `status`, `doctor`, `clean`, `path`) and a new experimental `agent-inspect/workspace` subpath export.
+
+  The workspace is additive and backward-compatible: existing trace directories keep working, existing `.agent-inspect` directories are adopted without rewrite, and trace files are never deleted. All manifest-derived paths are traversal-guarded, `workspace clean` is a dry-run by default, and there is no network I/O, daemon, or database dependency.
+
+## 3.5.5
+
+### Patch Changes
+
+- 822da6c: Fix npm README images: use absolute raw GitHub SVG URLs with sanitize=true so the product-loop diagram and logos render on npmjs.com. Harden readme-product-loop.svg for sanitizer compatibility. Docs-only; no runtime API changes.
+
 ## 3.5.4
 
 ### Patch Changes

package/README.md

@@ -1,7 +1,7 @@
 <p align="center">
   <picture>
-    <source media="(prefers-color-scheme: dark)" srcset="docs/assets/agent-inspect-logo-dark.svg">
-    <img src="docs/assets/agent-inspect-logo.svg" width="240" alt="AgentInspect">
+    <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/rajudandigam/agent-inspect/main/docs/assets/agent-inspect-logo-dark.svg?sanitize=true">
+    <img src="https://raw.githubusercontent.com/rajudandigam/agent-inspect/main/docs/assets/agent-inspect-logo.svg?sanitize=true" width="240" alt="AgentInspect">
   </picture>
 </p>
 
@@ -52,7 +52,7 @@ Install → one trace → one failure check → one share-safe artifact in under
 Guide: [First trace in 5 minutes](https://agentinspect.vercel.app/docs/getting-started/) · [repo](docs/FIRST-TRACE-IN-5-MINUTES.md)
 
 <p align="center">
-  <img src="docs/assets/readme-product-loop.svg" alt="Capture, inspect, check, redact — local JSONL only" width="720">
+  <img src="https://raw.githubusercontent.com/rajudandigam/agent-inspect/main/docs/assets/readme-product-loop.svg?sanitize=true" alt="Capture, inspect, check, redact — local JSONL only" width="720">
 </p>
 
 ## Choose your path
@@ -153,7 +153,7 @@ See [Compare](https://agentinspect.vercel.app/docs/compare/).
 
 ## Install details
 
-Current release: **3.5.3** (sixteen linked npm packages). Persisted trace schema **1.0**. Requires **Node.js >= 20**.
+Current release: **3.5.5** (sixteen linked npm packages). Persisted trace schema **1.0**. Requires **Node.js >= 20**.
 
 ```bash
 pnpm add agent-inspect

package/docs/ARCHITECTURE.md

@@ -1,6 +1,6 @@
 ## Architecture
 
-> **Adoption-focused overview:** see [TECHNICAL-GUIDE.md](./TECHNICAL-GUIDE.md) for a comprehensive technical guide (v3.5.3) suitable for blogs and outreach.
+> **Adoption-focused overview:** see [TECHNICAL-GUIDE.md](./TECHNICAL-GUIDE.md) for a comprehensive technical guide (v3.5.5) suitable for blogs and outreach.
 
 AgentInspect is a **local-first execution-tree debugger**: manual traces and log ingest produce inspectable step trees stored as JSONL on disk (default `.agent-inspect-runs/`), with a CLI for list/view/clean/logs/tail/export/diff workflows.
 

package/docs/CLI.md

@@ -29,6 +29,7 @@ Core commands:
 - `migrate` — convert one local AgentInspect JSONL file to schema 1.0 with dry-run or explicit output
 - `init` — scaffold local AgentInspect config and demo files (v3.1+)
 - `doctor` — diagnose local setup without network or installs (v3.1+)
+- `workspace` — manage a project-local trace workspace (`.agent-inspect/workspace.json`) (v4.0+)
 - `check` — run deterministic local trace checks with stable JSON and exit codes
 - `eval` — run deterministic local evals over existing traces
 - `redact` — redact a local JSON/JSONL file or trace copy
@@ -808,6 +809,26 @@ Provider design gate:
 - Provider prompts must ask for concise explanations from redacted facts only. They must not request, expose, or preserve raw chain-of-thought.
 - Cloud provider behavior is never selected by default and must be reviewed before implementation. Local provider support must still be explicit and opt-in.
 
+### 6.22 `workspace`
+
+Manage a project-local trace workspace (`.agent-inspect/workspace.json`). Added in v4.0. Local-only; trace files are never deleted. See [WORKSPACE.md](WORKSPACE.md) for the full model.
+
+```bash
+agent-inspect workspace init [--project <name>] [--redaction-profile <local|share|strict>] [--dry-run] [--json]
+agent-inspect workspace status [--json]
+agent-inspect workspace doctor [--json]
+agent-inspect workspace clean [--yes] [--json]
+agent-inspect workspace path [--json]
+```
+
+- `init` — create or adopt a workspace and its standard folders (`runs/`, `reports/`, `artifacts/`, `bundles/`, `notes/`, `index/`). An existing `workspace.json` is adopted without rewrite; existing traces are preserved.
+- `status` — read-only counts of traces, reports, artifacts, bundles, notes, plus index status.
+- `doctor` — validate the manifest, folder permissions, trace readability, and index freshness. Exits non-zero when a check fails.
+- `clean` — remove generated content (reports/artifacts/bundles/index). Dry-run by default; `--yes` deletes. Trace directories are never targeted.
+- `path` — print resolved workspace paths.
+
+All subcommands accept `--json` for deterministic output.
+
 ## 7. Optional TUI behavior
 
 `view --tui` delegates to `@agent-inspect/tui` and requires an interactive terminal. If the package is not installed, the CLI prints a short install hint.

package/docs/COMPARE.md

@@ -1,15 +1,29 @@
-## Compare AgentInspect
+# Compare AgentInspect
 
-AgentInspect is a local-first execution-tree debugger for TypeScript AI agents. It’s designed for inner-loop debugging, deterministic local eval heuristics, redaction, and quick inspection — not as a replacement for hosted observability, dataset evaluation, or production monitoring platforms.
+AgentInspect is a local-first trace workbench for TypeScript AI agents: capture runs as local JSONL, inspect and diff them from the CLI, gate them in CI, and redact before sharing. It is designed for the inner loop, not as a replacement for hosted observability, dataset evaluation, or production monitoring platforms. Those tools complement AgentInspect; this page maps where each fits.
 
 **Docs site:** [https://agentinspect.vercel.app/docs/compare/](https://agentinspect.vercel.app/docs/compare/)
 
+## Category comparison
+
+| Dimension | AgentInspect | Hosted dashboards (LangSmith, Langfuse) | Eval platforms (Braintrust) | Production APM / OTel pipelines |
+| --------- | ------------ | --------------------------------------- | --------------------------- | ------------------------------- |
+| Where traces live | Local JSONL on your disk | Vendor or self-hosted backend | Vendor backend | Your collector + backend |
+| Account / setup | None; `npm install` + CLI | Account or deployment | Account | SDK + collector + backend |
+| Primary surface | CLI (`view`, `report`, `diff`, `check`), local viewer, in-repo VS Code extension | Web dashboards | Web dashboards, eval UI | Dashboards, alerting |
+| Evals | Deterministic local heuristics and CI gates (`check`, `eval`, `circuit`, `guardrails`) | Platform evals | Datasets, scoring, regressions at scale | Not the focus |
+| Retention / fleet view | Not the goal | Strong | Strong | Strong |
+| Data sharing | Explicit `redact` + `scan` / `verify-safe` before you share a file | Team access controls | Team access controls | Org pipelines |
+| Network behavior | No upload; everything stays local unless you share a file | Uploads traces by design | Uploads by design | Ships telemetry by design |
+
+If you need hosted retention, fleet dashboards, dataset management, or org-wide pipelines, use one of those platforms alongside AgentInspect. Boundaries are listed in [LIMITATIONS.md](./LIMITATIONS.md); concrete inner-loop workflows in [USE-CASES.md](./USE-CASES.md).
+
 ## AgentInspect vs console.log
 
 - **console.log is flat**: logs are a stream of lines without run grouping or step boundaries.
 - **AgentInspect adds structure**: runs, nested steps, step types (tool/LLM), durations, status summaries, and local trace files you can inspect later.
 - **console.log still matters**: use it for quick values or ad-hoc debugging inside a step.
-- **AgentInspect helps you understand flow**: especially when your agent fans out, retries, or has nested tool/LLM calls.
+- **Existing structured logs work too**: the log ingest path parses JSON logs into trees without instrumentation ([LOGGING-PLAYBOOK.md](./LOGGING-PLAYBOOK.md)).
 
 ## AgentInspect vs LangSmith
 
@@ -17,7 +31,7 @@ LangSmith is a hosted/platform workflow for tracing, evaluation, and observabili
 
 AgentInspect is local-first CLI debugging:
 
-- Use AgentInspect to debug locally before/alongside LangSmith when iterating on agent logic.
+- Use AgentInspect to debug locally before or alongside LangSmith when iterating on agent logic; `@agent-inspect/langchain` captures LangChain callbacks locally.
 - AgentInspect does not provide hosted dashboards, dataset/eval management workflows, or production tracing pipelines.
 
 ## AgentInspect vs Langfuse
@@ -26,7 +40,7 @@ Langfuse provides broader LLM observability (tracing, prompt management, dataset
 
 AgentInspect focuses on local execution trees and CLI workflows:
 
-- Complementary: use AgentInspect for quick local run understanding; use Langfuse for dashboards and longer-lived observability workflows.
+- Complementary: use AgentInspect for quick local run understanding and PR/debug artifacts; use Langfuse for dashboards and longer-lived observability workflows.
 - Not a replacement.
 
 ## AgentInspect vs Braintrust
@@ -35,7 +49,7 @@ Braintrust is strong for evals, regressions, datasets, and production AI quality
 
 AgentInspect is lighter and local-first:
 
-- Use AgentInspect to understand a single run locally and run deterministic trace checks/eval heuristics before sharing artifacts.
+- Use AgentInspect to understand a single run locally and run deterministic trace checks (`check`, `@agent-inspect/eval` heuristics, `@agent-inspect/guardrails`, `@agent-inspect/circuit`) before sharing artifacts.
 - Use Braintrust when you want repeatable evals, comparisons at scale, and production quality workflows.
 
 ## AgentInspect vs Phoenix / OpenInference
@@ -44,7 +58,7 @@ Phoenix and the OpenInference ecosystem are standards-oriented and useful for tr
 
 AgentInspect can export **OpenInference-compatible JSON** and **OTLP JSON** as local files:
 
-- Exports are **compatibility-oriented** and should be validated against your target backend/collector.
+- Exports are **compatibility-oriented** and experimental; validate them against your target backend or collector.
 - AgentInspect does not claim that every backend will accept these exports without configuration.
 
 ## AgentInspect vs OpenTelemetry setup
@@ -53,9 +67,9 @@ OpenTelemetry is powerful, but setup can be heavier (SDK configuration, exporter
 
 AgentInspect avoids SDK/collector setup for local debugging:
 
-- Use AgentInspect when you want quick, local execution trees with no collector required.
+- Use AgentInspect when you want quick, local execution trees with no collector required; no OTel SDK is added to the root package.
 - Use OpenTelemetry when you need organization-wide production telemetry pipelines.
-- Exports can help bridge later, but AgentInspect is not an OpenTelemetry SDK replacement.
+- Local OTLP JSON export can help bridge later, but AgentInspect is not an OpenTelemetry SDK replacement.
 
 ## Quick decision table
 
@@ -63,9 +77,12 @@ AgentInspect avoids SDK/collector setup for local debugging:
 | --- | --- |
 | Local agent debugging | Strong fit |
 | No-account CLI tracing | Strong fit |
-| Deterministic local eval heuristics | Good fit |
+| Framework-native capture (AI SDK, OpenAI Agents, LangChain, MCP) | Good fit (optional adapter packages) |
+| Deterministic local eval heuristics and CI gates | Good fit |
+| Loop / retry / timeout analysis (`circuit`) | Good fit |
 | Share-safe local redaction copy | Good fit |
-| VS Code trace sidebar (in-repo / dev host) | Good fit — Marketplace listing separate |
+| Existing structured logs to trees | Good fit |
+| VS Code trace review (in-repo extension) | Good fit; Marketplace listing separate |
 | Production dashboards | Not the goal |
 | Hosted eval datasets | Not the goal |
 | Prompt management | Not the goal |
@@ -74,12 +91,16 @@ AgentInspect avoids SDK/collector setup for local debugging:
 
 ## v3.5 positioning (local inner loop)
 
-AgentInspect v3.5 is the **adoption release** — not new runtime surface. Use it when:
+AgentInspect v3.5 is the **adoption release**. The npm map is one core package (`agent-inspect`: APIs + CLI) plus optional packages that stay out of the root dependency graph: framework adapters (`ai-sdk`, `openai-agents`, `langchain`, `mcp`, `adapter-sdk`), CI and quality gates (`vitest`, `jest`, `eval`, `guardrails`, `circuit`, `harness`), and inspection surfaces (`viewer`, `tui`, `mcp-server`, `redact`). See the [package map](../README.md#package-map).
 
-- You want **traces on disk** before/alongside hosted tools
+Use it when:
+
+- You want **traces on disk** before or alongside hosted tools
 - You need **CI gates** (`check`, `eval`, `circuit`) without a vendor account
 - You want **metadata-only** defaults and explicit `redact` before sharing
 
 Keep using LangSmith, Langfuse, Braintrust, Phoenix, or OTel when you need hosted retention, fleet dashboards, dataset management, or org-wide production pipelines. AgentInspect is complementary.
 
-**Full adoption path:** [ADOPTION.md](./ADOPTION.md)
+Before posting any exported trace to an issue, PR, or chat, follow [SAFE-TRACE-SHARING.md](./SAFE-TRACE-SHARING.md) and [SECURITY.md](../SECURITY.md): redaction is a key-based safeguard, and exports deserve a human review.
+
+**Full adoption path:** [ADOPTION.md](./ADOPTION.md) · **Boundaries:** [LIMITATIONS.md](./LIMITATIONS.md) · **Scenarios:** [USE-CASES.md](./USE-CASES.md)

package/docs/SAFE-TRACE-SHARING.md

@@ -57,6 +57,7 @@ Replace sensitive data with clear placeholders such as `example.test`, `user@exa
 - OpenInference / OTLP JSON exports: check attributes, span names, events, and resource metadata.
 - Structured log ingest configs: confirm mapped keys do not pull in full request bodies, headers, raw prompts, or unbounded output fields.
 - LangChain adapter traces: keep `capture: "metadata-only"` for shareable examples; review `capture: "preview"` traces carefully because previews can include prompt or output fragments.
+- Third-party adapter packages: follow the [Adapter SDK privacy checklist](./ADAPTER-SDK-PRIVACY.md) before sharing adapter traces, examples, or registry submissions.
 
 ## When to use each profile
 

package/docs/TECHNICAL-GUIDE.md

@@ -1,8 +1,10 @@
 # AgentInspect Technical Guide
 
-**Version:** 3.5.3 · **Audience:** TypeScript agent developers, platform engineers, tech leads  
-**Purpose:** Source-of-truth technical overview for adoption content (blogs, articles, emails, demos).  
-**Status:** Grounded in shipped code and docs as of v3.5.3. Not a compliance or marketing guarantee.
+**Version:** 3.5.5 · **Audience:** TypeScript agent developers, platform engineers, tech leads
+
+**Purpose:** Source-of-truth technical overview for adoption content (blogs, articles, emails, demos).
+
+**Status:** Grounded in shipped code and docs as of v3.5.5. Not a compliance or marketing guarantee.
 
 ---
 
@@ -107,9 +109,9 @@ flowchart LR
 
 The root `agent-inspect` tarball ships core + CLI. Heavy optional deps stay in scoped packages.
 
-### 4.2 Public package ecosystem (v3.5.3)
+### 4.2 Public package ecosystem (v3.5.5)
 
-Sixteen linked npm packages at **3.5.3**:
+Sixteen linked npm packages at **3.5.5**:
 
 | Tier | Package | Purpose |
 | ---- | ------- | ------- |
@@ -433,7 +435,7 @@ Daily workflow: capture → inspect → verify → share safely → scale. See [
 
 ## 16. Content creation appendix
 
-Reuse these blocks in blogs, emails, and comments. All statements are accurate to v3.5.3; adjust version numbers when publishing.
+Reuse these blocks in blogs, emails, and comments. All statements are accurate to v3.5.5; adjust version numbers when publishing.
 
 ### 16.1 Elevator pitches
 
@@ -523,4 +525,4 @@ Vitest reporter → upload artifact → `verify-safe` → reviewer workflow →
 
 ---
 
-*AgentInspect v3.5.3 — local-first trace workbench for TypeScript AI agents. MIT license.*
+*AgentInspect v3.5.5 — local-first trace workbench for TypeScript AI agents. MIT license.*

package/docs/WORKSPACE.md

@@ -0,0 +1,96 @@
+# Local trace workspace
+
+The **workspace** turns a project's local AgentInspect output into one coherent, local-first workbench: traces, reports, artifacts, bundles, an optional index, and notes — described by a small manifest at `.agent-inspect/workspace.json`.
+
+It is **opt-in and additive**. Existing trace directories keep working exactly as before; the workspace is a layout + config layer, not a schema change. Trace files are never deleted by workspace commands.
+
+- **Added in:** v4.0.0
+- **Stability:** experimental
+- **Non-goals:** no daemon, no database dependency, no cloud sync, no default migration of existing traces.
+
+## Quick start
+
+```bash
+# Create (or adopt) a workspace in the current project
+agent-inspect workspace init --project my-agent
+
+# See what's inside
+agent-inspect workspace status
+
+# Validate config, permissions, and index freshness
+agent-inspect workspace doctor
+
+# Preview generated content that could be cleaned (dry-run)
+agent-inspect workspace clean
+# Actually remove generated content (traces are always preserved)
+agent-inspect workspace clean --yes
+
+# Print resolved paths
+agent-inspect workspace path
+```
+
+All commands accept `--json` for deterministic machine-readable output.
+
+## Layout
+
+```text
+.agent-inspect/
+  workspace.json
+  runs/        # JSONL traces
+  reports/     # generated reports
+  artifacts/   # CI artifacts
+  bundles/     # share-safe bundles
+  index/       # optional local index (v4.1+)
+  notes/       # decision notes
+```
+
+## Manifest: `workspace.json`
+
+```json
+{
+  "schemaVersion": "1.0",
+  "project": "my-agent",
+  "createdAt": "2026-07-08T00:00:00.000Z",
+  "traceDirs": ["runs"],
+  "reportsDir": "reports",
+  "artifactsDir": "artifacts",
+  "bundlesDir": "bundles",
+  "notesDir": "notes",
+  "redactionProfile": "share",
+  "index": { "enabled": false, "type": "none" }
+}
+```
+
+All directory fields are relative to the `.agent-inspect` workspace directory and must resolve inside it (absolute paths and `..` traversal are rejected). `redactionProfile` sets the default share-safety posture used by later bundle/export flows.
+
+## Adoption and compatibility
+
+- **New project:** `workspace init` creates `workspace.json` and the standard folders.
+- **Existing `.agent-inspect` directory:** detected and adopted without rewriting or moving existing traces. Top-level `*.jsonl` traces are preserved and included via a `"."` trace directory.
+- **Existing manifest:** left untouched; missing folders are created.
+- **No manifest:** existing commands continue to operate against explicit trace directories.
+
+## Programmatic API
+
+The same model is available on the `agent-inspect/workspace` subpath (experimental):
+
+```ts
+import {
+  createWorkspace,
+  getWorkspaceStatus,
+  readWorkspaceManifestFile,
+  resolveWorkspaceLocation,
+} from "agent-inspect/workspace";
+
+const result = await createWorkspace({ project: "my-agent" });
+const status = await getWorkspaceStatus(result.location, result.manifest);
+console.log(status.traceFiles, status.reports);
+```
+
+Helpers are local-only, never perform network I/O, and never delete trace files.
+
+## Safety
+
+- Manifest parsing validates shape and bounds input size; invalid input is rejected with clear messages rather than throwing.
+- All manifest-derived paths are resolved and confirmed to stay within the workspace directory.
+- `workspace clean` is a dry-run by default and never targets trace directories.

package/docs/assets/readme-product-loop.svg

@@ -1,31 +1,26 @@
-<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 760 140" role="img" aria-label="AgentInspect product loop: capture, inspect, check, redact">
-  <rect width="760" height="140" fill="transparent"/>
-  <g font-family="ui-sans-serif,system-ui,-apple-system,Segoe UI,sans-serif">
-    <!-- Capture -->
+<svg xmlns="http://www.w3.org/2000/svg" width="760" height="140" viewBox="0 0 760 140" role="img" aria-label="AgentInspect product loop: capture, inspect, check, redact">
+  <defs>
+    <marker id="arr" markerWidth="8" markerHeight="8" refX="6" refY="3" orient="auto">
+      <path d="M0,0 L6,3 L0,6 Z" fill="#64748b"/>
+    </marker>
+  </defs>
+  <rect width="760" height="140" fill="#ffffff"/>
+  <g font-family="ui-sans-serif, system-ui, -apple-system, Segoe UI, sans-serif">
     <rect x="16" y="28" width="150" height="72" rx="12" fill="#eef2ff" stroke="#6366f1" stroke-width="1.5"/>
     <text x="91" y="58" text-anchor="middle" font-size="14" font-weight="600" fill="#312e81">Capture</text>
-    <text x="91" y="78" text-anchor="middle" font-size="11" fill="#475569">run � tools � adapters</text>
-    <!-- arrow -->
+    <text x="91" y="78" text-anchor="middle" font-size="11" fill="#475569">run, tools, adapters</text>
     <path d="M174 64h28" stroke="#64748b" stroke-width="2" marker-end="url(#arr)"/>
-    <!-- Inspect -->
     <rect x="210" y="28" width="150" height="72" rx="12" fill="#ecfdf5" stroke="#10b981" stroke-width="1.5"/>
     <text x="285" y="58" text-anchor="middle" font-size="14" font-weight="600" fill="#064e3b">Inspect</text>
-    <text x="285" y="78" text-anchor="middle" font-size="11" fill="#475569">tree � report � diff</text>
+    <text x="285" y="78" text-anchor="middle" font-size="11" fill="#475569">tree, report, diff</text>
     <path d="M368 64h28" stroke="#64748b" stroke-width="2" marker-end="url(#arr)"/>
-    <!-- Check -->
     <rect x="404" y="28" width="150" height="72" rx="12" fill="#fff7ed" stroke="#f59e0b" stroke-width="1.5"/>
     <text x="479" y="58" text-anchor="middle" font-size="14" font-weight="600" fill="#78350f">Check</text>
-    <text x="479" y="78" text-anchor="middle" font-size="11" fill="#475569">CI � eval � stalls</text>
+    <text x="479" y="78" text-anchor="middle" font-size="11" fill="#475569">CI, eval, stalls</text>
     <path d="M562 64h28" stroke="#64748b" stroke-width="2" marker-end="url(#arr)"/>
-    <!-- Redact -->
     <rect x="598" y="28" width="146" height="72" rx="12" fill="#f8fafc" stroke="#64748b" stroke-width="1.5"/>
     <text x="671" y="58" text-anchor="middle" font-size="14" font-weight="600" fill="#0f172a">Redact</text>
     <text x="671" y="78" text-anchor="middle" font-size="11" fill="#475569">share-safe artifact</text>
-    <text x="380" y="124" text-anchor="middle" font-size="12" fill="#64748b">local JSONL only � no account � no upload</text>
+    <text x="380" y="124" text-anchor="middle" font-size="12" fill="#64748b">local JSONL only - no account - no upload</text>
   </g>
-  <defs>
-    <marker id="arr" markerWidth="8" markerHeight="8" refX="6" refY="3" orient="auto">
-      <path d="M0,0 L6,3 L0,6 Z" fill="#64748b"/>
-    </marker>
-  </defs>
 </svg>

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agent-inspect",
-  "version": "3.5.4",
+  "version": "4.0.0",
   "license": "MIT",
   "type": "module",
   "description": "Trace, check, and safely share TypeScript AI-agent runs locally — no account, no upload, metadata-only by default",
@@ -115,6 +115,16 @@
         "types": "./packages/core/dist/reporters.d.cts",
         "default": "./packages/core/dist/reporters.cjs"
       }
+    },
+    "./workspace": {
+      "import": {
+        "types": "./packages/core/dist/workspace.d.ts",
+        "default": "./packages/core/dist/workspace.mjs"
+      },
+      "require": {
+        "types": "./packages/core/dist/workspace.d.cts",
+        "default": "./packages/core/dist/workspace.cjs"
+      }
     }
   },
   "bin": {
@@ -153,6 +163,7 @@
     "docs/DESIGN-PARTNER-GUIDE.md",
     "docs/DEMO-SCRIPT.md",
     "docs/TECHNICAL-GUIDE.md",
+    "docs/WORKSPACE.md",
     "docs/AI-SDK-ADOPTION.md",
     "docs/OPENAI-AGENTS-LOCAL.md",
     "docs/NESTJS.md",