docs/env: clarify embedded vs external worker env ownership

.env.example

@@ -79,6 +79,8 @@ IMPORT_LIBRARY_DIRS=
 
 # Heavy compute is always worker-backed.
 # App server calls compute-worker over HTTP for ONNX whisper alignment + PDF layout parsing.
+# Embedded/local (`pnpm dev` / `pnpm start` without COMPUTE_WORKER_URL) uses this same file.
+# `compute/worker/.env*` is only for standalone external worker deployments.
 # In single-container self-host setups, run compute-worker + NATS alongside the app and
 # point COMPUTE_WORKER_URL at that internal worker endpoint.
 # Required:

compute/worker/.env.example

@@ -1,3 +1,7 @@
+# Standalone compute-worker env.
+# Used only when running compute-worker as a separate service/container.
+# Not used by app embedded worker startup via `pnpm dev` / `pnpm start`.
+
 # Compute worker bind
 # Platform note:
 # - Local/manual: keep PORT=8081

docs-site/docs/deploy/compute-worker.md

@@ -1,7 +1,8 @@
 title: Compute Worker (NATS JetStream)
 ---
 
-Use this guide for worker-backed deployments where heavy compute runs outside the Next.js app server.
+Use this guide when compute-worker runs as a standalone service outside the Next.js app server.
+For embedded/local startup (`pnpm dev` / `pnpm start` without `COMPUTE_WORKER_URL`), use root `.env` instead.
 
 ## Overview
 
@@ -30,9 +31,12 @@ Required:
 - `S3_SECRET_ACCESS_KEY`
 
 > [!IMPORTANT]
-> **S3 credentials cannot be left blank/empty** when running in worker mode.
-> While the main Next.js server can generate random, dynamic S3 keys on-the-fly when `USE_EMBEDDED_WEED_MINI=true` and `S3_*` vars are blank, the compute worker runs in a separate process and cannot connect to SeaweedFS using those dynamically generated keys. 
-> To use the compute worker with the embedded SeaweedFS, you **must configure identical, stable S3 credentials** (e.g. `S3_ACCESS_KEY_ID` and `S3_SECRET_ACCESS_KEY`) in both the root `.env` and the compute worker `.env` files.
+> This file (`compute/worker/.env*`) is only for standalone worker deployments.
+> In embedded/local startup, app entrypoint spawns worker with the already-resolved root `.env` values.
+> For standalone worker deployments, keep shared app/worker values aligned:
+> - `COMPUTE_WORKER_TOKEN`
+> - shared object storage settings (`S3_*`)
+> - shared timeout/stale settings (`COMPUTE_WHISPER_TIMEOUT_MS`, `COMPUTE_PDF_TIMEOUT_MS`, `COMPUTE_OP_STALE_MS`)
 
 Common optional:
 
@@ -86,6 +90,16 @@ Set the same value on app + worker envs.
 
 There is no app-local compute fallback. If worker is unavailable, affected requests fail.
 
+## Config ownership summary
+
+- Embedded/local startup (`pnpm dev` / `pnpm start`, no `COMPUTE_WORKER_URL`):
+  - Configure root `.env` only.
+  - `compute/worker/.env*` is ignored.
+- Standalone external worker service:
+  - Configure app root `.env` with `COMPUTE_WORKER_URL` + `COMPUTE_WORKER_TOKEN`.
+  - Configure worker service env (`compute/worker/.env*` or platform env).
+  - Keep shared values aligned (`COMPUTE_WORKER_TOKEN`, `S3_*`, timeout/stale values).
+
 ## Production notes
 
 - Worker mode assumes shared object storage is reachable by both app server and worker.

docs-site/docs/deploy/local-development.md

@@ -127,7 +127,8 @@ If you need mirrors or pinned artifact locations, set `WHISPER_MODEL_BASE_URL` i
 <details>
 <summary><strong>External compute worker dev stack (optional)</strong></summary>
 
-Use this when you want durable compute with NATS JetStream + KV while keeping Next.js on native host `pnpm dev`.
+Use this only when you intentionally run compute-worker as a separate service.
+Default local flow does not need `compute/worker/.env`; embedded worker startup reads root `.env`.
 Full worker deployment details are in [Compute Worker (NATS JetStream)](./compute-worker).
 
 Start only NATS + compute-worker via compose watch:
@@ -137,7 +138,7 @@ docker compose --env-file compute/worker/.env -f compute/worker/docker-compose.y
 # or: pnpm compute:dev:watch
 ```
 
-`compute/worker/.env.example` contains a starter config. Copy it to `compute/worker/.env` and adjust values for your environment.
+`compute/worker/.env.example` contains a starter config for standalone worker service deployments.
 
 Run the main app separately on the host:
 
@@ -145,7 +146,7 @@ Run the main app separately on the host:
 pnpm dev
 ```
 
-For app server worker mode, set:
+For app -> external worker routing, set in root `.env`:
 
 ```env
 COMPUTE_WORKER_URL=http://localhost:8081
@@ -153,7 +154,7 @@ COMPUTE_WORKER_TOKEN=<same-token-used-by-worker>
 ```
 
 Worker mode requires worker-reachable shared object storage (S3-compatible endpoint).
-Non-exposed embedded `weed mini` is not a supported topology for external worker mode.
+For external worker mode, object storage must be shared/reachable by both app and worker services.
 
 </details>
 
@@ -182,7 +183,15 @@ cp .env.example .env
 
 Then edit `.env`.
 
-Required compute worker connectivity (all modes):
+Default embedded worker flow (no external worker URL):
+
+```env
+# Leave COMPUTE_WORKER_URL unset.
+# Entry point auto-starts embedded worker+NATS when available.
+START_EMBEDDED_COMPUTE_WORKER=
+```
+
+External worker flow:
 
 ```env
 COMPUTE_WORKER_URL=http://localhost:8081
@@ -245,7 +254,7 @@ S3_SECRET_ACCESS_KEY=your-secret-key
 ```
 
   </TabItem>
-  <TabItem value="worker-mode" label="External Worker + NATS">
+  <TabItem value="worker-mode" label="External Worker Service">
 
 ```env
 API_BASE=http://host.docker.internal:8880/v1
@@ -292,6 +301,9 @@ Learn about migration behavior and commands in [Migrations](../configure/migrati
 pnpm dev
 ```
 
+If you use embedded worker startup (no `COMPUTE_WORKER_URL`) and the host is missing `nats-server`,
+install `nats-server` locally or switch to external worker mode.
+
   </TabItem>
   <TabItem value="prod" label="Build + Start">
 

docs-site/docs/reference/environment-variables.md

@@ -53,8 +53,14 @@ For auth-enabled deployments, use **Settings → Admin** as the primary source o
 | `RUN_FS_MIGRATIONS` | Storage migrations | `true` | Set `false` to skip startup filesystem -> S3/DB migration pass |
 | `IMPORT_LIBRARY_DIR` | Library import | `docstore/library` fallback | Set a single server library root |
 | `IMPORT_LIBRARY_DIRS` | Library import | unset | Set multiple roots (comma/colon/semicolon separated) |
-| `COMPUTE_WORKER_URL` | Heavy compute backend | unset | Required; base URL for external compute worker |
-| `COMPUTE_WORKER_TOKEN` | Heavy compute backend | unset | Required bearer token for external compute worker auth |
+| `COMPUTE_WORKER_URL` | Heavy compute backend | unset | Set only for standalone external compute worker; leave unset for embedded worker startup |
+| `COMPUTE_WORKER_TOKEN` | Heavy compute backend | unset (auto-generated in embedded startup) | Required for standalone external compute worker auth; must match worker |
+| `START_EMBEDDED_COMPUTE_WORKER` | Heavy compute backend | auto (`true` when `COMPUTE_WORKER_URL` unset) | Set `false` to disable embedded worker startup and require external worker URL/token |
+| `EMBEDDED_COMPUTE_WORKER_PORT` | Heavy compute backend | `8081` | Override embedded worker bind port |
+| `EMBEDDED_NATS_PORT` | Heavy compute backend | `4222` | Override embedded NATS client port |
+| `EMBEDDED_NATS_MONITOR_PORT` | Heavy compute backend | `8222` | Override embedded NATS monitor port |
+| `EMBEDDED_NATS_STORE_DIR` | Heavy compute backend | `docstore/nats/jetstream` | Override embedded JetStream storage directory |
+| `NATS_URL` | Heavy compute backend | `nats://127.0.0.1:4222` in embedded startup | Optional override for embedded startup or required on standalone worker service |
 | `COMPUTE_JOB_CONCURRENCY` | Heavy compute backend | `1` | Worker-side shared compute concurrency cap |
 | `COMPUTE_WHISPER_TIMEOUT_MS` | Heavy compute backend | `30000` | Shared whisper alignment timeout budget (worker + worker client wait budget) |
 | `COMPUTE_PDF_TIMEOUT_MS` | Heavy compute backend | `300000` | Shared PDF idle-timeout budget (worker + worker client wait budget) |
@@ -355,17 +361,57 @@ Multiple library roots for server library import.
 
 ### COMPUTE_WORKER_URL
 
-Base URL for external compute worker mode.
+Base URL for standalone external compute worker mode.
 
-- Required
+- Leave unset for embedded/local startup (`pnpm dev` / `pnpm start`) so entrypoint can start embedded worker+NATS.
+- Required only when using a standalone external worker service.
 - Example: `http://localhost:8081`
 
 ### COMPUTE_WORKER_TOKEN
 
-Bearer token for external compute worker auth.
+Bearer token for compute-worker auth.
 
-- Required
-- Must match worker service `COMPUTE_WORKER_TOKEN`
+- Required for standalone external worker service mode.
+- Must match worker service `COMPUTE_WORKER_TOKEN`.
+- In embedded startup, entrypoint auto-generates one if unset.
+
+### START_EMBEDDED_COMPUTE_WORKER
+
+Controls whether entrypoint auto-starts embedded compute-worker + NATS.
+
+- Default behavior: enabled when `COMPUTE_WORKER_URL` is unset
+- Set `false` to disable embedded startup and require external worker URL/token
+
+### EMBEDDED_COMPUTE_WORKER_PORT
+
+Embedded compute-worker HTTP port.
+
+- Default: `8081`
+
+### EMBEDDED_NATS_PORT
+
+Embedded NATS client port.
+
+- Default: `4222`
+
+### EMBEDDED_NATS_MONITOR_PORT
+
+Embedded NATS monitor (`/healthz`) port.
+
+- Default: `8222`
+
+### EMBEDDED_NATS_STORE_DIR
+
+Embedded JetStream storage directory.
+
+- Default: `docstore/nats/jetstream`
+
+### NATS_URL
+
+NATS connection URL used by compute worker runtime.
+
+- Embedded startup default: `nats://127.0.0.1:4222`
+- Standalone worker service: set in worker service env (`compute/worker/.env*` or platform env)
 
 ### COMPUTE_JOB_CONCURRENCY
 

docs-site/docs/reference/stack.md

@@ -47,7 +47,7 @@ title: Stack
 
 Monorepo packages under `compute/`:
 
-- **`@openreader/compute-core`** — ONNX runtime lifecycle, model management, and inference logic shared between local and worker modes
+- **`@openreader/compute-core`** — ONNX runtime lifecycle, model management, and inference logic shared by compute worker runtime + app/worker contracts
   - ONNX runtime: `onnxruntime-node` with `@huggingface/tokenizers`
   - Whisper alignment: `onnx-community/whisper-base_timestamped` (int8) for word-level timestamps
   - PDF layout: `Bei0001/PP-DocLayoutV3-ONNX` for document block detection and layout parsing