refactor: move res into internal (#20390)

Signed-off-by: Jin Hai <[email protected]>

.gitignore

@@ -31,7 +31,7 @@ Cargo.lock
 docker/ragflow-logs/
 /flask_session
 /logs
-rag/res/deepdoc
+internal/rag/res/deepdoc/
 
 # Exclude sdk generated files
 sdk/python/ragflow.egg-info/

AGENTS.md

@@ -18,7 +18,7 @@ Use this file as the local operating guide for the current codebase. Prefer the
 ## Go-only target
 - Remove all Python code and Python-specific dependencies, scripts, containers, CI jobs, documentation, SDKs, and tests from the repository. Do not add new Python code while this removal is in progress. Replace Python-based preparation of Go native libraries, DeepDoc models, and the BPE table before removing the existing download scripts.
 - Move any still-required behavior and test coverage into the Go backend or the frontend before deleting the old implementation. Do not preserve a Python server, compatibility layer, fallback, or dual-backend path.
-- Before deleting `rag/`, move the Go DeepDoc `.ort` models and `ocr.res` out of `rag/res/deepdoc/`, then update model discovery, dependency preparation, Docker packaging, and tests together.
+- Before deleting `rag/`, move the Go DeepDoc `.ort` models and `ocr.res` out of `internal/rag/res/deepdoc/`, then update model discovery, dependency preparation, Docker packaging, and tests together.
 - Remove the frontend's Go/Python backend variants and update `web/CLAUDE.md` when the frontend no longer needs them. Until then, treat its dual-backend instructions as migration-only guidance; new frontend work should target the Go API.
 - Update Docker's default entrypoint, dependency image, and CI workflows when removing their Python paths. Until then, select Go explicitly with `API_PROXY_SCHEME=go` where that variable controls startup.
 - Keep this guide aligned with the checked-in tree. Once the Python directories are deleted, remove their migration inventory below.

Dockerfile

@@ -16,7 +16,7 @@ WORKDIR /ragflow
 # layout.laws/manual/paper.onnx are byte-identical to layout.onnx, so we
 # exclude them from the tar extract and symlink them to layout.onnx instead,
 # saving ~219MB in the image.
-RUN mkdir -p /ragflow/rag/res/deepdoc /root/.ragflow
+RUN mkdir -p /ragflow/internal/rag/res/deepdoc /root/.ragflow
 RUN --mount=type=bind,from=infiniflow/ragflow_deps:latest,source=/huggingface.co,target=/huggingface.co \
     tar --exclude='.*' \
         --exclude='layout.laws.onnx' \
@@ -32,13 +32,13 @@ RUN --mount=type=bind,from=infiniflow/ragflow_deps:latest,source=/huggingface.co
         -cf - \
         /huggingface.co/InfiniFlow/text_concat_xgb_v1.0 \
         /huggingface.co/InfiniFlow/deepdoc \
-        | tar -xf - --strip-components=3 -C /ragflow/rag/res/deepdoc && \
+        | tar -xf - --strip-components=3 -C /ragflow/internal/rag/res/deepdoc && \
-    #ln -s layout.onnx /ragflow/rag/res/deepdoc/layout.laws.onnx && \
+    #ln -s layout.onnx /ragflow/internal/rag/res/deepdoc/layout.laws.onnx && \
-    #ln -s layout.onnx /ragflow/rag/res/deepdoc/layout.manual.onnx && \
+    #ln -s layout.onnx /ragflow/internal/rag/res/deepdoc/layout.manual.onnx && \
-    #ln -s layout.onnx /ragflow/rag/res/deepdoc/layout.paper.onnx
+    #ln -s layout.onnx /ragflow/internal/rag/res/deepdoc/layout.paper.onnx
-    ln -s layout.ort /ragflow/rag/res/deepdoc/layout.laws.ort && \
+    ln -s layout.ort /ragflow/internal/rag/res/deepdoc/layout.laws.ort && \
-    ln -s layout.ort /ragflow/rag/res/deepdoc/layout.manual.ort && \
+    ln -s layout.ort /ragflow/internal/rag/res/deepdoc/layout.manual.ort && \
-    ln -s layout.ort /ragflow/rag/res/deepdoc/layout.paper.ort
+    ln -s layout.ort /ragflow/internal/rag/res/deepdoc/layout.paper.ort
 
 # Copy the cl100k_base BPE table used by the Go tokenizer (tiktoken-go
 # cl100k_base). The deps image ships it at its root; the Go image previously

Dockerfile_base

@@ -10,12 +10,12 @@ ARG GITEE_TOKEN=""
 WORKDIR /ragflow
 
 # copy models downloaded via download_deps.py
-RUN mkdir -p /ragflow/rag/res/deepdoc /root/.ragflow
+RUN mkdir -p /ragflow/internal/rag/res/deepdoc /root/.ragflow
 RUN --mount=type=bind,from=infiniflow/ragflow_deps:latest,source=/huggingface.co,target=/huggingface.co \
     tar --exclude='.*' -cf - \
         /huggingface.co/InfiniFlow/text_concat_xgb_v1.0 \
         /huggingface.co/InfiniFlow/deepdoc \
-        | tar -xf - --strip-components=3 -C /ragflow/rag/res/deepdoc
+        | tar -xf - --strip-components=3 -C /ragflow/internal/rag/res/deepdoc
 
 # https://github.com/chrismattmann/tika-python
 # This is the only way to run python-tika without internet access. Without this set, the default is to check the tika version and pull latest every time from Apache.

README.md

@@ -264,7 +264,7 @@ Install the Go version specified in `go.mod` (currently Go 1.27), Clang 20, LLD
    bash build.sh --all
    ```
 
-   The script prepares native libraries and model resources required for the Go build and needs `requests` and `huggingface-hub`. Skip this step if you have prepared the same resources by other means. When started from the repository root, Go services automatically find `rag/res/deepdoc`; to start from another directory, set `DEEPDOC_MODEL_DIR` to its absolute path.
+   The script prepares native libraries and model resources required for the Go build and needs `requests` and `huggingface-hub`. Skip this step if you have prepared the same resources by other means. When started from the repository root, Go services automatically find `internal/rag/res/deepdoc`; to start from another directory, set `DEEPDOC_MODEL_DIR` to its absolute path.
 
 3. Start the local dependencies and make sure the hosts and ports in **conf/service_conf.yaml** point to addresses accessible from the host. Go source services connect to Compose-exposed Kvrocks at `localhost:6379`, while Go Docker services connect to Kvrocks on the container network. If using the default Elasticsearch engine, set `vm.max_map_count` on the Docker host to at least `262144` first.
 

README_ar.md

@@ -285,7 +285,7 @@
    bash build.sh --all
    ```
 
-   يجهّز البرنامج النصي المكتبات الأصلية وموارد النماذج اللازمة لبناء Go، ويحتاج إلى `requests` و`huggingface-hub`. يمكن تخطي هذه الخطوة إذا جُهزت الموارد نفسها بطريقة أخرى. عند التشغيل من جذر المستودع، تعثر خدمات Go تلقائيًا على `rag/res/deepdoc`؛ وللتشغيل من دليل آخر، اضبط `DEEPDOC_MODEL_DIR` على المسار المطلق لذلك الدليل.
+   يجهّز البرنامج النصي المكتبات الأصلية وموارد النماذج اللازمة لبناء Go، ويحتاج إلى `requests` و`huggingface-hub`. يمكن تخطي هذه الخطوة إذا جُهزت الموارد نفسها بطريقة أخرى. عند التشغيل من جذر المستودع، تعثر خدمات Go تلقائيًا على `internal/rag/res/deepdoc`؛ وللتشغيل من دليل آخر، اضبط `DEEPDOC_MODEL_DIR` على المسار المطلق لذلك الدليل.
 3. ابدأ الخدمات التابعة المطلوبة (Elasticsearch وMySQL وMinIO وNATS وKvrocks وClickHouse) باستخدام Docker Compose:
 
    ```bash

README_fr.md

@@ -273,7 +273,7 @@ Pour changer de moteur de documents, modifier la configuration, redémarrer les
    bash build.sh --all
    ```
 
-   Le script prépare les bibliothèques natives et les ressources de modèles nécessaires à la compilation Go et requiert `requests` et `huggingface-hub`. Ignorez cette étape si vous avez préparé les mêmes ressources autrement. Depuis la racine du dépôt, les services Go trouvent automatiquement `rag/res/deepdoc` ; depuis un autre répertoire, définissez `DEEPDOC_MODEL_DIR` sur son chemin absolu.
+   Le script prépare les bibliothèques natives et les ressources de modèles nécessaires à la compilation Go et requiert `requests` et `huggingface-hub`. Ignorez cette étape si vous avez préparé les mêmes ressources autrement. Depuis la racine du dépôt, les services Go trouvent automatiquement `internal/rag/res/deepdoc` ; depuis un autre répertoire, définissez `DEEPDOC_MODEL_DIR` sur son chemin absolu.
 3. Lancez les dépendances requises (Elasticsearch, MySQL, MinIO, NATS, Kvrocks et ClickHouse) avec Docker Compose :
 
    ```bash

README_id.md

@@ -281,7 +281,7 @@ Untuk mengganti mesin dokumen, mengubah konfigurasi, memulai ulang layanan, sert
    bash build.sh --all
    ```
 
-   Skrip tersebut menyiapkan pustaka native dan sumber daya model yang diperlukan untuk build Go serta memerlukan `requests` dan `huggingface-hub`. Lewati langkah ini jika sumber daya yang sama telah disiapkan dengan cara lain. Saat dijalankan dari root repositori, layanan Go otomatis menemukan `rag/res/deepdoc`; untuk menjalankan dari direktori lain, atur `DEEPDOC_MODEL_DIR` ke path absolutnya.
+   Skrip tersebut menyiapkan pustaka native dan sumber daya model yang diperlukan untuk build Go serta memerlukan `requests` dan `huggingface-hub`. Lewati langkah ini jika sumber daya yang sama telah disiapkan dengan cara lain. Saat dijalankan dari root repositori, layanan Go otomatis menemukan `internal/rag/res/deepdoc`; untuk menjalankan dari direktori lain, atur `DEEPDOC_MODEL_DIR` ke path absolutnya.
 3. Jalankan dependensi yang diperlukan (Elasticsearch, MySQL, MinIO, NATS, Kvrocks, dan ClickHouse) menggunakan Docker Compose:
 
    ```bash

README_ja.md

@@ -263,7 +263,7 @@ Go版のDockerデプロイでは `docker/.env` と `docker/docker-compose.yml`
    bash build.sh --all
    ```
 
-   このスクリプトはGoビルドに必要なネイティブライブラリとモデルリソースを準備し、`requests`と`huggingface-hub`を使用します。同じリソースを別の方法で準備済みの場合は、この手順を省略できます。リポジトリルートから起動するとGoサービスは`rag/res/deepdoc`を自動検出します。別のディレクトリから起動する場合は、`DEEPDOC_MODEL_DIR`にその絶対パスを設定してください。
+   このスクリプトはGoビルドに必要なネイティブライブラリとモデルリソースを準備し、`requests`と`huggingface-hub`を使用します。同じリソースを別の方法で準備済みの場合は、この手順を省略できます。リポジトリルートから起動するとGoサービスは`internal/rag/res/deepdoc`を自動検出します。別のディレクトリから起動する場合は、`DEEPDOC_MODEL_DIR`にその絶対パスを設定してください。
 3. Docker Composeで必要な依存サービス（Elasticsearch、MySQL、MinIO、NATS、Kvrocks、ClickHouse）を起動します:
 
    ```bash

README_ko.md

@@ -263,7 +263,7 @@ Go Docker 배포는 `docker/.env`와 `docker/docker-compose.yml`을 사용하며
    bash build.sh --all
    ```
 
-   이 스크립트는 Go 빌드에 필요한 네이티브 라이브러리와 모델 리소스를 준비하며 `requests`와 `huggingface-hub`가 필요합니다. 같은 리소스를 다른 방법으로 준비했다면 이 단계를 건너뛸 수 있습니다. 저장소 루트에서 시작하면 Go 서비스가 `rag/res/deepdoc`을 자동으로 찾습니다. 다른 디렉터리에서 시작하려면 `DEEPDOC_MODEL_DIR`을 해당 디렉터리의 절대 경로로 설정하세요.
+   이 스크립트는 Go 빌드에 필요한 네이티브 라이브러리와 모델 리소스를 준비하며 `requests`와 `huggingface-hub`가 필요합니다. 같은 리소스를 다른 방법으로 준비했다면 이 단계를 건너뛸 수 있습니다. 저장소 루트에서 시작하면 Go 서비스가 `internal/rag/res/deepdoc`을 자동으로 찾습니다. 다른 디렉터리에서 시작하려면 `DEEPDOC_MODEL_DIR`을 해당 디렉터리의 절대 경로로 설정하세요.
 
 3. Docker Compose를 사용하여 필요한 의존 서비스(Elasticsearch, MySQL, MinIO, NATS, Kvrocks, ClickHouse)를 시작합니다:
 

README_pt_br.md

@@ -269,7 +269,7 @@ Para trocar o mecanismo de documentos, alterar configurações, reiniciar servi
    bash build.sh --all
    ```
 
-   O script prepara as bibliotecas nativas e os recursos de modelo necessários para o build Go e requer `requests` e `huggingface-hub`. Ignore esta etapa se você já preparou os mesmos recursos de outra forma. Quando iniciados na raiz do repositório, os serviços Go encontram automaticamente `rag/res/deepdoc`; para iniciar em outro diretório, defina `DEEPDOC_MODEL_DIR` como o caminho absoluto correspondente.
+   O script prepara as bibliotecas nativas e os recursos de modelo necessários para o build Go e requer `requests` e `huggingface-hub`. Ignore esta etapa se você já preparou os mesmos recursos de outra forma. Quando iniciados na raiz do repositório, os serviços Go encontram automaticamente `internal/rag/res/deepdoc`; para iniciar em outro diretório, defina `DEEPDOC_MODEL_DIR` como o caminho absoluto correspondente.
 3. Inicie as dependências necessárias (Elasticsearch, MySQL, MinIO, NATS, Kvrocks e ClickHouse) usando Docker Compose:
 
    ```bash

README_ru.md

@@ -279,7 +279,7 @@
    bash build.sh --all
    ```
 
-   Скрипт подготавливает нативные библиотеки и ресурсы моделей для сборки Go и требует `requests` и `huggingface-hub`. Пропустите этот шаг, если те же ресурсы подготовлены другим способом. При запуске из корня репозитория сервисы Go автоматически находят `rag/res/deepdoc`; для запуска из другого каталога задайте `DEEPDOC_MODEL_DIR` как абсолютный путь к нему.
+   Скрипт подготавливает нативные библиотеки и ресурсы моделей для сборки Go и требует `requests` и `huggingface-hub`. Пропустите этот шаг, если те же ресурсы подготовлены другим способом. При запуске из корня репозитория сервисы Go автоматически находят `internal/rag/res/deepdoc`; для запуска из другого каталога задайте `DEEPDOC_MODEL_DIR` как абсолютный путь к нему.
 
 3. Запустите необходимые зависимости (Elasticsearch, MySQL, MinIO, NATS, Kvrocks и ClickHouse):
 

README_tr.md

@@ -281,7 +281,7 @@ Belge motorunu değiştirme, yapılandırma değişikliklerinden sonra hizmetler
    bash build.sh --all
    ```
 
-   Betik, Go derlemesi için gerekli yerel kitaplıkları ve model kaynaklarını hazırlar ve `requests` ile `huggingface-hub` gerektirir. Aynı kaynakları başka bir yöntemle hazırladıysanız bu adımı atlayabilirsiniz. Depo kökünden başlatıldığında Go hizmetleri `rag/res/deepdoc` dizinini otomatik olarak bulur; başka bir dizinden başlatmak için `DEEPDOC_MODEL_DIR` değişkenini mutlak yola ayarlayın.
+   Betik, Go derlemesi için gerekli yerel kitaplıkları ve model kaynaklarını hazırlar ve `requests` ile `huggingface-hub` gerektirir. Aynı kaynakları başka bir yöntemle hazırladıysanız bu adımı atlayabilirsiniz. Depo kökünden başlatıldığında Go hizmetleri `internal/rag/res/deepdoc` dizinini otomatik olarak bulur; başka bir dizinden başlatmak için `DEEPDOC_MODEL_DIR` değişkenini mutlak yola ayarlayın.
 3. Gerekli bağımlılıkları (Elasticsearch, MySQL, MinIO, NATS, Kvrocks ve ClickHouse) Docker Compose ile başlatın:
 
    ```bash

README_tzh.md

@@ -281,7 +281,7 @@ Go 版 Docker 部署使用 `docker/.env` 和 `docker/docker-compose.yml`，以 K
    bash build.sh --all
    ```
 
-   下載腳本使用 `requests` 和 `huggingface-hub` 來準備 Go 建置所需的原生程式庫和模型資源；如已透過其他方式準備好相同資源，可略過此步驟。從倉庫根目錄啟動時，Go 服務會自動尋找 `rag/res/deepdoc`；如需從其他目錄啟動，請將 `DEEPDOC_MODEL_DIR` 設為該目錄的絕對路徑。
+   下載腳本使用 `requests` 和 `huggingface-hub` 來準備 Go 建置所需的原生程式庫和模型資源；如已透過其他方式準備好相同資源，可略過此步驟。從倉庫根目錄啟動時，Go 服務會自動尋找 `internal/rag/res/deepdoc`；如需從其他目錄啟動，請將 `DEEPDOC_MODEL_DIR` 設為該目錄的絕對路徑。
 3. 啟動本機依賴服務，並確認 **conf/service_conf.yaml** 中的主機與連接埠可從主機存取：
 
    ```bash

README_zh.md

@@ -288,7 +288,7 @@ Go 版 Docker 部署使用 `docker/.env` 和 `docker/docker-compose.yml`，使
    bash build.sh --all
    ```
 
-   下载脚本负责准备 Go 构建所需的原生库和模型资源，并需要 `requests` 和 `huggingface-hub`。如果已通过其他方式准备好相同资源，可跳过这一步。Go 服务从仓库根目录启动时会自动查找 `rag/res/deepdoc`；如需从其他目录启动，请将 `DEEPDOC_MODEL_DIR` 设为该目录的绝对路径。
+   下载脚本负责准备 Go 构建所需的原生库和模型资源，并需要 `requests` 和 `huggingface-hub`。如果已通过其他方式准备好相同资源，可跳过这一步。Go 服务从仓库根目录启动时会自动查找 `internal/rag/res/deepdoc`；如需从其他目录启动，请将 `DEEPDOC_MODEL_DIR` 设为该目录的绝对路径。
 
 3. 启动本地依赖服务，并确认 **conf/service_conf.yaml** 中的主机与端口对应宿主机可访问的地址。Go 源码服务通过 `localhost:6379` 连接 Compose 暴露的 Kvrocks；Go Docker 服务则在容器网络中连接 Kvrocks。如果使用默认 Elasticsearch，还需先将 Docker 主机的 `vm.max_map_count` 设为至少 `262144`。
 

build.sh

@@ -1181,10 +1181,10 @@ main() {
                 pkgs=(./internal/deepdoc/parser/pdf/inference/native_analyzer/...)
             fi
             # The in-process (Go) DeepDoc backend needs the .ort weights. Default
-            # MODEL_DIR to the canonical repo model dir (rag/res/deepdoc) so a
+            # MODEL_DIR to the canonical repo model dir (internal/rag/res/deepdoc) so a
             # local run needs no MODEL_DIR export after `download_go_deps.py`.
             REPO_ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
-            export MODEL_DIR="${MODEL_DIR:-$REPO_ROOT/rag/res/deepdoc}"
+            export MODEL_DIR="${MODEL_DIR:-$REPO_ROOT/internal/rag/res/deepdoc}"
             run_go_tests_tagged "cgo integration" "${pkgs[@]}"
             run_native_integration_tests
             ;;

cmd/ragflow_server.go

@@ -1678,10 +1678,6 @@ func logTokenizerCounters() {
 	}
 }
 
-// resolveDeepDocModelDir picks the model directory: the explicit DEEPDOC_MODEL_DIR
-// env, else the RAGFlow default model dir (rag/res/deepdoc),
-// else the snapshot fetched by ragflow_deps/download_deps.py. The first
-// candidate that actually contains the required weights wins.
 func resolveDeepDocModelDir() string {
 	if v := strings.TrimSpace(common.GetEnv(common.EnvDeepDocModelDir)); v != "" {
 		return v
@@ -1691,7 +1687,7 @@ func resolveDeepDocModelDir() string {
 	// variable can point at the DeepDoc weights as well as the embedding tokenizers.
 	candidates := append([]string(nil), common.ModelAssetCandidates("huggingface.co/InfiniFlow/deepdoc")...)
 	candidates = append(candidates,
-		filepath.Join(wd, "rag", "res", "deepdoc"),
+		filepath.Join(wd, "internal", "rag", "res", "deepdoc"),
 		filepath.Join(wd, "huggingface.co", "InfiniFlow", "deepdoc"),
 	)
 	for _, c := range candidates {
@@ -1701,7 +1697,7 @@ func resolveDeepDocModelDir() string {
 	}
 	// None verified; return the canonical default so any error message points
 	// at the conventional location.
-	return filepath.Join(wd, "rag", "res", "deepdoc")
+	return filepath.Join(wd, "internal", "rag", "res", "deepdoc")
 }
 
 // resolveDeepDocDropScore returns the explicit DEEPDOC_DROP_SCORE env, else the

docker/README.md

@@ -154,7 +154,7 @@ The optional `tei-cpu` and `tei-gpu` profiles start a local text-embeddings-infe
 
 ### DeepDoc (in-process)
 
-DeepDoc layout analysis (DLA), OCR (text detection/recognition), and table structure recognition (TSR) run **in-process** inside the RAGFlow server using ONNX Runtime — there is no separate DeepDoc service to deploy. ONNX Runtime is statically linked into the server binary (resolved at runtime via dlopen(NULL); no `libonnxruntime.so` is required) and the models are loaded at runtime; `DEEPDOC_MODEL_DIR` overrides the default model directory. `Dockerfile` copies the required model assets into `/ragflow/rag/res/deepdoc`.
+DeepDoc layout analysis (DLA), OCR (text detection/recognition), and table structure recognition (TSR) run **in-process** inside the RAGFlow server using ONNX Runtime — there is no separate DeepDoc service to deploy. ONNX Runtime is statically linked into the server binary (resolved at runtime via dlopen(NULL); no `libonnxruntime.so` is required) and the models are loaded at runtime; `DEEPDOC_MODEL_DIR` overrides the default model directory. `Dockerfile` copies the required model assets into `/ragflow/internal/rag/res/deepdoc`.
 
 In the RAGFlow open-source 1.0 release, DeepDoc uses CPU inference.
 

docs/develop/launch_ragflow_from_source.md

@@ -62,7 +62,7 @@ python3 -m venv /tmp/ragflow-go-download-venv
 /tmp/ragflow-go-download-venv/bin/python ragflow_deps/download_go_deps.py
 ```
 
-The downloader fetches the static libraries needed by `build.sh`, plus `det.ort`, `layout.ort`, `tsr.ort`, `rec.ort`, and `ocr.res` into `rag/res/deepdoc/`. These files are required by the in-process Go DeepDoc backend. Keep the server's working directory at the repository root so it can find them automatically; if you launch it elsewhere, set `DEEPDOC_MODEL_DIR` to the absolute path of `rag/res/deepdoc`.
+The downloader fetches the static libraries needed by `build.sh`, plus `det.ort`, `layout.ort`, `tsr.ort`, `rec.ort`, and `ocr.res` into `internal/rag/res/deepdoc/`. These files are required by the in-process Go DeepDoc backend. Keep the server's working directory at the repository root so it can find them automatically; if you launch it elsewhere, set `DEEPDOC_MODEL_DIR` to the absolute path of `rag/res/deepdoc`.
 
 The isolated download environment can be removed after the resources have been prepared.
 
@@ -147,7 +147,7 @@ If you changed a dependency's published port or credentials, update `conf/servic
 
 `RAGFLOW_DEV_MODE=true` bypasses the code-versus-database version check. Use it only for development, not to run an older production binary against a newer database. The standalone `--migrate` action must still complete first.
 
-If startup reports `no in-process DeepDoc backend serving`, check that all five model files above are in `rag/res/deepdoc/`, then re-run the Go dependency downloader and rebuild if necessary. If the tokenizer reports a missing `cl100k_base.tiktoken`, ensure the BPE table is in `ragflow_deps/`.
+If startup reports `no in-process DeepDoc backend serving`, check that all five model files above are in `internal/rag/res/deepdoc/`, then re-run the Go dependency downloader and rebuild if necessary. If the tokenizer reports a missing `cl100k_base.tiktoken`, ensure the BPE table is in `ragflow_deps/`.
 
 ## 4. Start the frontend
 

internal/common/environments.go

@@ -249,7 +249,7 @@ const (
 
 	// EnvDeepDocModelDir points the in-process (Go) DeepDoc backend at the
 	// model snapshot (see common.DeepDocModelFiles); mirrors
-	// the RAGFlow default model dir (rag/res/deepdoc).
+	// the RAGFlow default model dir (internal/rag/res/deepdoc).
 	EnvDeepDocModelDir = "DEEPDOC_MODEL_DIR"
 	// EnvDeepDocDropScore overrides the confidence threshold below which the
 	// in-process (Go) DeepDoc backend blanks recognized text while preserving

internal/deepdoc/native/run.sh

@@ -14,11 +14,11 @@ cd "$SCRIPT_DIR"
 # ONNX Runtime is statically linked (libonnxruntime.a) and resolved via
 # dlopen(NULL); no ORT_LIB is needed.
 #
-# Default MODEL_DIR to the canonical repo model dir (rag/res/deepdoc) so running
+# Default MODEL_DIR to the canonical repo model dir (internal/rag/res/deepdoc) so running
 # `download_go_deps.py` is enough — no MODEL_DIR export is required for local
 # Go DeepDoc runs. Override with `MODEL_DIR=... bash run.sh` if needed.
 REPO_ROOT="$(cd "$SCRIPT_DIR/../../.." && pwd)"
-MODEL_DIR="${MODEL_DIR:-$REPO_ROOT/rag/res/deepdoc}"
+MODEL_DIR="${MODEL_DIR:-$REPO_ROOT/internal/rag/res/deepdoc}"
 export MODEL_DIR
 
 run_task() {

internal/deepdoc/parser/pdf/inference/native_analyzer/native_analyzer_test.go

@@ -47,7 +47,7 @@ func TestNativeAnalyzerUninitializedNegative(t *testing.T) {
 	}
 	modelDir := os.Getenv("MODEL_DIR")
 	if modelDir == "" {
-		modelDir = filepath.Join("..", "..", "..", "..", "rag", "res", "deepdoc")
+		modelDir = filepath.Join("..", "..", "..", "..", "internal", "rag", "res", "deepdoc")
 	}
 	if _, err := NewAnalyzer(modelDir, DefaultDropScore); err == nil {
 		t.Error("NewAnalyzer succeeded before ONNX Runtime init; expected error")

internal/deepdoc/parser/pdf/inference/native_analyzer/raster_alignment_test.go

@@ -58,7 +58,7 @@ func repoRoot(t *testing.T) string {
 	}
 	// .../internal/deepdoc/parser/pdf/inference/native_analyzer -> repo root is 6 levels up.
 	root := filepath.Join(dir, "..", "..", "..", "..", "..", "..")
-	if _, err := os.Stat(filepath.Join(root, "rag", "res", "deepdoc")); err != nil {
+	if _, err := os.Stat(filepath.Join(root, "internal", "rag", "res", "deepdoc")); err != nil {
 		t.Fatalf("repoRoot %s missing model dir: %v", root, err)
 	}
 	return root

internal/deepdoc/parser/pdf/util/garbled.go

@@ -215,13 +215,13 @@ var (
 )
 
 // loadOCRAlphabet reads the recogniser's character dictionary once
-// (rag/res/deepdoc/ocr.res, provisioned next to the OCR model). On any error it
+// (internal/rag/res/deepdoc/ocr.res, provisioned next to the OCR model). On any error it
 // returns an empty set, which OcrCanRepresent treats as "unknown alphabet" and
 // so allows the existing fallback behaviour.
 func loadOCRAlphabet() map[rune]struct{} {
 	ocrAlphabetOnce.Do(func() {
 		ocrAlphabet = map[rune]struct{}{}
-		data, err := os.ReadFile(filepath.Join(utility.GetProjectRoot(), "rag", "res", "deepdoc", "ocr.res"))
+		data, err := os.ReadFile(filepath.Join(utility.GetProjectRoot(), "internal", "rag", "res", "deepdoc", "ocr.res"))
 		if err != nil {
 			return
 		}

internal/development.md

@@ -166,7 +166,7 @@ binary carries it and registers the backend at startup.
 #### Model weights
 
 The Go backend loads **`.ort`** (FlatBuffer) weights; the Python side loads
-**`.onnx`**. Both formats live side by side in `rag/res/deepdoc/` — neither
+**`.onnx`**. Both formats live side by side in `internal/rag/res/deepdoc/` — neither
 supersedes the other, so do not delete one to "clean up".
 
 |        | Go (in-process)                                          | Python                                                       |
@@ -175,12 +175,12 @@ supersedes the other, so do not delete one to "clean up".
 | Files  | `det.ort`, `layout.ort`, `tsr.ort`, `rec.ort`, `ocr.res` | `det.onnx`, `layout.onnx`, `tsr.onnx`, `rec.onnx`, `ocr.res` |
 
 `download_go_deps.py` (§1.4) fetches the five required files — four `.ort` plus
-`ocr.res` — into `rag/res/deepdoc/`; `download_deps.py` snapshots the whole
+`ocr.res` — into `internal/rag/res/deepdoc/`; `download_deps.py` snapshots the whole
 `InfiniFlow/deepdoc` repo and therefore carries both formats.
 
 Auto-discovery is **relative to the server process's working directory**:
 `resolveDeepDocModelDir()` (`cmd/ragflow_server.go`) probes
-`<cwd>/rag/res/deepdoc`, then `<cwd>/huggingface.co/InfiniFlow/deepdoc`.
+`<cwd>/internal/rag/res/deepdoc`, then `<cwd>/huggingface.co/InfiniFlow/deepdoc`.
 Launching `./bin/ragflow_server` from the repo root therefore needs no
 `MODEL_DIR` / `DEEPDOC_MODEL_DIR` export; from any other CWD — or an image with
 a different WORKDIR — set `MODEL_DIR` explicitly.
@@ -189,7 +189,7 @@ a different WORKDIR — set `MODEL_DIR` explicitly.
 authoritative list — `HasModelFiles()` refuses to serve when any file in it is
 missing from the model directory.
 
-> **Note**: A `rag/res/deepdoc/` populated before the `.ort` switch holds only
+> **Note**: A `internal/rag/res/deepdoc/` populated before the `.ort` switch holds only
 > `.onnx` and will NOT serve the Go backend, even though the directory looks
 > fully populated. Re-run `download_go_deps.py` after updating.
 

internal/service/nlp/synonym.go

@@ -61,7 +61,7 @@ func NewSynonym(redis Client, resPath string, wordnetDir string) *Synonym {
 	s.lookupNum.Store(100000000)
 
 	if resPath == "" {
-		s.resPath = "rag/res"
+		s.resPath = "internal/rag/res"
 	}
 
 	// Initialize WordNet with provided path

internal/service/nlp/term_weight.go

@@ -54,12 +54,12 @@ func NewTermWeightDealer(resPath string) *TermWeightDealer {
 
 	// Load named entity dictionary
 	if resPath == "" {
-		resPath = "rag/res"
+		resPath = "internal/rag/res"
 	}
 
 	nerPath := filepath.Join(resPath, "ner.json")
 	if data, err := os.ReadFile(nerPath); err == nil {
-		if err := json.Unmarshal(data, &d.ne); err != nil {
+		if err = json.Unmarshal(data, &d.ne); err != nil {
 			common.Warn("Failed to load ner.json", zap.Error(err))
 		}
 	} else {

ragflow_deps/download_go_deps.py

@@ -35,7 +35,7 @@
 #
 # Go DeepDoc weights: in addition to the native libs, this script downloads the
 # five Go model files (internal/common.DeepDocModelFiles) from InfiniFlow/deepdoc
-# straight into the repo's canonical model directory `rag/res/deepdoc/` (one level
+# straight into the repo's canonical model directory `internal/rag/res/deepdoc/` (one level
 # up from this script). The Go server auto-discovers that directory via
 # resolveDeepDocModelDir(), and build.sh --test-native / internal/deepdoc/native/run.sh
 # default MODEL_DIR there too — so after running this script NO MODEL_DIR env needs
@@ -287,7 +287,7 @@ def download_go_models(use_china_mirrors=False):
     backend with no further setup.
 
     The files are written into the repo's canonical model directory
-    `rag/res/deepdoc/` (relative to the repo root, one level up from this
+    `internal/rag/res/deepdoc/` (relative to the repo root, one level up from this
     script). The Go server auto-discovers that directory via
     resolveDeepDocModelDir(), and build.sh --test-native / run.sh default
     MODEL_DIR there too — so after this script runs, no MODEL_DIR env needs to
@@ -305,7 +305,7 @@ def download_go_models(use_china_mirrors=False):
 
     # Canonical local-dev model dir the Go backend auto-discovers.
     repo_root = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
-    target_dir = os.path.join(repo_root, "rag", "res", "deepdoc")
+    target_dir = os.path.join(repo_root, "internal", "rag", "res", "deepdoc")
     os.makedirs(target_dir, exist_ok=True)
 
     missing = []

tools/render_diff/warp_align.py

@@ -41,7 +41,7 @@ Run
     WARP_PDF=test/benchmark/test_docs/Doc1.pdf \
         bash build.sh --test-manual ./internal/deepdoc/parser/pdf/ -run TestWarpAlignGo
     .venv/bin/python tools/render_diff/warp_align.py compare
-    # (needs HF weights in rag/res/deepdoc)
+    # (needs HF weights in internal/rag/res/deepdoc)
     .venv/bin/python tools/render_diff/warp_align.py rec
 """