diff --git a/.env.example b/.env.example
index 8ff39a2..0821ea4 100644
--- a/.env.example
+++ b/.env.example
@@ -1,3 +1,13 @@
+# Conversion engine: "local" (in-process Docling) or "remote" (Docling Serve)
+# CONVERSION_ENGINE=local
+
+# Docling Serve settings (remote mode only)
+# DOCLING_SERVE_URL=http://localhost:5001
+# DOCLING_SERVE_API_KEY=
+
+# Max seconds per conversion (default: 600)
+# CONVERSION_TIMEOUT=600
+
 # CORS (comma-separated origins, only needed for custom deployments)
 # CORS_ORIGINS=http://localhost:3000,https://your-domain.com
 
diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
index bea8895..50b4605 100644
--- a/CONTRIBUTING.md
+++ b/CONTRIBUTING.md
@@ -22,7 +22,13 @@ Thank you for your interest in contributing to Docling Studio! This guide will h
 ```bash
 cd document-parser
 python -m venv .venv && source .venv/bin/activate
+
+# Remote mode (lightweight — delegates to Docling Serve)
 pip install -r requirements.txt
+
+# Local mode (full — runs Docling in-process)
+pip install -r requirements-local.txt
+
 pip install ruff pytest pytest-asyncio httpx  # dev tools
 uvicorn main:app --reload --port 8000
 ```
@@ -129,11 +135,16 @@ We use [Semantic Versioning](https://semver.org/): `MAJOR.MINOR.PATCH`.
 
 ### Docker Image Tags
 
+Each release produces two image variants:
+
 | Tag | Description |
 |-----|-------------|
-| `X.Y.Z` | Exact version |
-| `X.Y` | Latest patch of this minor |
-| `latest` | Latest stable release |
+| `X.Y.Z-remote` | Exact version — lightweight (Docling Serve) |
+| `X.Y.Z-local` | Exact version — full (in-process Docling) |
+| `X.Y-remote` | Latest patch of this minor — lightweight |
+| `X.Y-local` | Latest patch of this minor — full |
+| `latest-remote` | Latest stable — lightweight |
+| `latest-local` | Latest stable — full |
 
 ### Hotfix
 
diff --git a/README.md b/README.md
index 5ac1323..0a70625 100644
--- a/README.md
+++ b/README.md
@@ -85,22 +85,42 @@ frontend/src/
 
 ## Quick Start
 
-### Docker (fastest)
+Docling Studio ships two Docker image variants:
+
+| Variant | Image tag | Size | Description |
+|---------|-----------|------|-------------|
+| **remote** | `latest-remote` | ~300 MB | Lightweight — delegates to an external [Docling Serve](https://github.com/DS4SD/docling-serve) instance |
+| **local** | `latest-local` | ~2–3 GB | Full — runs Docling in-process (downloads ML models on first run) |
+
+### Docker — remote mode (fastest)
 
 ```bash
-docker run -p 3000:3000 ghcr.io/scub-france/docling-studio:latest
+docker run -p 3000:3000 \
+  -e DOCLING_SERVE_URL=http://your-docling-serve:5001 \
+  ghcr.io/scub-france/docling-studio:latest-remote
 ```
 
-Open [http://localhost:3000](http://localhost:3000)
+### Docker — local mode (self-contained)
+
+```bash
+docker run -p 3000:3000 ghcr.io/scub-france/docling-studio:latest-local
+```
 
 > **Note:** The first analysis takes longer as Docling downloads its ML models (~400 MB). Subsequent runs are fast.
 
+Open [http://localhost:3000](http://localhost:3000)
+
 ### Docker Compose (for development)
 
 ```bash
 git clone https://github.com/scub-france/Docling-Studio.git
 cd Docling-Studio
+
+# Local mode (default)
 docker compose up --build
+
+# Remote mode
+CONVERSION_MODE=remote DOCLING_SERVE_URL=http://your-docling-serve:5001 docker compose up --build
 ```
 
 ### Local Development
@@ -109,7 +129,13 @@ docker compose up --build
 ```bash
 cd document-parser
 python -m venv .venv && source .venv/bin/activate
+
+# Remote mode (lightweight)
 pip install -r requirements.txt
+
+# Local mode (with Docling)
+pip install -r requirements-local.txt
+
 uvicorn main:app --reload --port 8000
 ```
 
@@ -156,6 +182,9 @@ All configuration is done via environment variables. See [`.env.example`](.env.e
 
 | Variable | Default | Description |
 |----------|---------|-------------|
+| `CONVERSION_ENGINE` | `local` | `local` (in-process Docling) or `remote` (Docling Serve) |
+| `DOCLING_SERVE_URL` | `http://localhost:5001` | Docling Serve endpoint (remote mode only) |
+| `DOCLING_SERVE_API_KEY` | — | API key for Docling Serve (optional) |
 | `CORS_ORIGINS` | `http://localhost:3000,...` | CORS allowed origins (comma-separated) |
 | `UPLOAD_DIR` | `./uploads` | File storage directory |
 | `DB_PATH` | `./data/docling_studio.db` | SQLite database path |
@@ -168,7 +197,7 @@ GitHub Actions pipelines (see [`.github/workflows/`](.github/workflows/)):
 | Workflow | Trigger | What it does |
 |----------|---------|--------------|
 | **CI** | push to `main`, pull requests | Lint + type check + Backend tests + Frontend tests + build |
-| **Release** | push tag `v*` | Build & push multi-arch Docker image to `ghcr.io` |
+| **Release** | push tag `v*` | Build & push **two** multi-arch Docker images (`remote` + `local`) to `ghcr.io` |
 | **Docs** | push to `main` (docs changes) | Build & deploy MkDocs to GitHub Pages |
 
 We follow [Semantic Versioning](https://semver.org/) with a simplified Git Flow. See [CONTRIBUTING.md](CONTRIBUTING.md) for the full release process.
@@ -183,12 +212,11 @@ We follow [Semantic Versioning](https://semver.org/) with a simplified Git Flow.
 
 ### Docker Desktop settings
 
-The document parser needs **at least 4 GB of RAM**:
-
-| Resource | Minimum | Recommended |
-|----------|---------|-------------|
-| Memory   | 6 GB    | 8 GB+       |
-| CPUs     | 4       | 8+          |
+| | Remote image | Local image |
+|---|---|---|
+| **Image size** | ~300 MB | ~2–3 GB |
+| **Memory** | 2 GB | 6 GB (recommended 8 GB+) |
+| **CPUs** | 2 | 4 (recommended 8+) |
 
 ### Platform support
 
diff --git a/docs/contributing.md b/docs/contributing.md
index 17c6e7d..44412e3 100644
--- a/docs/contributing.md
+++ b/docs/contributing.md
@@ -20,7 +20,13 @@
     ```bash
     cd document-parser
     python -m venv .venv && source .venv/bin/activate
+
+    # Remote mode (lightweight — delegates to Docling Serve)
     pip install -r requirements.txt
+
+    # Local mode (full — runs Docling in-process)
+    pip install -r requirements-local.txt
+
     pip install ruff pytest pytest-asyncio httpx
     uvicorn main:app --reload --port 8000
     ```
diff --git a/docs/getting-started.md b/docs/getting-started.md
index cd38397..8e8bbcf 100644
--- a/docs/getting-started.md
+++ b/docs/getting-started.md
@@ -1,33 +1,67 @@
 # Getting Started
 
-## Docker Compose (recommended)
+Docling Studio ships two Docker image variants:
+
+| Variant | Image tag | Size | Description |
+|---------|-----------|------|-------------|
+| **remote** | `latest-remote` | ~300 MB | Lightweight — delegates to an external [Docling Serve](https://github.com/DS4SD/docling-serve) instance |
+| **local** | `latest-local` | ~2–3 GB | Full — runs Docling in-process (downloads ML models on first run) |
+
+## Docker — remote mode (fastest)
+
+```bash
+docker run -p 3000:3000 \
+  -e DOCLING_SERVE_URL=http://your-docling-serve:5001 \
+  ghcr.io/scub-france/docling-studio:latest-remote
+```
+
+## Docker — local mode (self-contained)
+
+```bash
+docker run -p 3000:3000 ghcr.io/scub-france/docling-studio:latest-local
+```
+
+> **Note:** The first analysis takes longer as Docling downloads its ML models (~400 MB). Subsequent runs are fast.
+
+Open [http://localhost:3000](http://localhost:3000).
+
+## Docker Compose (recommended for development)
 
 ```bash
 git clone https://github.com/scub-france/Docling-Studio.git
 cd Docling-Studio
-docker compose up --build
-```
 
-Open [http://localhost:3000](http://localhost:3000).
+# Local mode (default)
+docker compose up --build
+
+# Remote mode
+CONVERSION_MODE=remote DOCLING_SERVE_URL=http://your-docling-serve:5001 docker compose up --build
+```
 
 ## Local Development
 
-### Backend (Python 3.12+)
+=== "Backend (Python 3.12+)"
 
-```bash
-cd document-parser
-python -m venv .venv && source .venv/bin/activate
-pip install -r requirements.txt
-uvicorn main:app --reload --port 8000
-```
+    ```bash
+    cd document-parser
+    python -m venv .venv && source .venv/bin/activate
 
-### Frontend (Node 20+)
+    # Remote mode (lightweight)
+    pip install -r requirements.txt
 
-```bash
-cd frontend
-npm install
-npm run dev
-```
+    # Local mode (with Docling)
+    pip install -r requirements-local.txt
+
+    uvicorn main:app --reload --port 8000
+    ```
+
+=== "Frontend (Node 20+)"
+
+    ```bash
+    cd frontend
+    npm install
+    npm run dev
+    ```
 
 The frontend runs on `http://localhost:3000` and proxies API calls to `http://localhost:8000`.
 
@@ -71,6 +105,9 @@ All configuration is done via environment variables:
 
 | Variable | Default | Description |
 |----------|---------|-------------|
+| `CONVERSION_ENGINE` | `local` | `local` (in-process Docling) or `remote` (Docling Serve) |
+| `DOCLING_SERVE_URL` | `http://localhost:5001` | Docling Serve endpoint (remote mode only) |
+| `DOCLING_SERVE_API_KEY` | — | API key for Docling Serve (optional) |
 | `CORS_ORIGINS` | `http://localhost:3000,...` | CORS allowed origins |
 | `UPLOAD_DIR` | `./uploads` | File storage directory |
 | `DB_PATH` | `./data/docling_studio.db` | SQLite database path |
@@ -78,9 +115,10 @@ All configuration is done via environment variables:
 
 ## System Requirements
 
-| Resource | Minimum | Recommended |
-|----------|---------|-------------|
-| Memory   | 6 GB    | 8 GB+       |
-| CPUs     | 4       | 8+          |
+| | Remote image | Local image |
+|---|---|---|
+| **Image size** | ~300 MB | ~2–3 GB |
+| **Memory** | 2 GB | 6 GB (recommended 8 GB+) |
+| **CPUs** | 2 | 4 (recommended 8+) |
 
 All Docker images are multi-arch (`linux/amd64` + `linux/arm64`). No GPU required.