Merge branch 'dev'

Author: Ahmath-Gadji

.env.example

@@ -18,6 +18,8 @@ JUDGE_API_KEY=
 
 # App
 APP_PORT=8080 # this is the forwarded port
+API_NUM_WORKERS=1 # Number of uvicorn workers for the FastAPI app
+# ENABLE_RAY_SERVE=true # Use Ray Serve for the API instead of Uvicorn
 
 # # VLLM
 # VLLM_PORT=8000
@@ -70,7 +72,9 @@ RAY_NUM_GPUS=0.1
 RAY_POOL_SIZE=1 # Number of serializer actor instances
 RAY_MAX_TASKS_PER_WORKER=5 # Number of tasks per serializer instance
 RAY_DASHBOARD_PORT=8265
-RAY_RUNTIME_ENV_HOOK=ray._private.runtime_env.uv_runtime_env_hook.hook
+RAY_ENABLE_RECORD_ACTOR_TASK_LOGGING=1
+RAY_task_retry_delay_ms=3000
+RAY_ENABLE_UV_RUN_RUNTIME_ENV=0
 
 # SAVE UPLOADED FILES
 SAVE_UPLOADED_FILES=true # usefull for chainlit source viewing
@@ -92,6 +96,7 @@ SAVE_UPLOADED_FILES=true # usefull for chainlit source viewing
 # UV_LINK_MODE=copy
 # UV_CACHE_DIR=/tmp/uv-cache 
 # RAY_ADDRESS=ray://X.X.X.X:10001
+# HEAD_NODE_IP=X.X.X.X
 
 # # Chainlit UI authentication
 # CHAINLIT_AUTH_SECRET=... # has to be generated with with this command: 'uv run chainlit create-secret' but a random value works too.

.github/workflows/build.yml

@@ -0,0 +1,75 @@
+name: Build and Release
+
+on:
+  push:
+    tags: ["v*"]
+  workflow_dispatch:
+
+env:
+  REGISTRY: ghcr.io
+  IMAGE_NAME: ${{ github.repository }}
+  PYTHON_VERSION: "3.12"
+
+jobs:
+  build-and-push-image:
+    runs-on: ubuntu-latest
+    if: startsWith(github.ref, 'refs/tags/') && github.event.base_ref == 'refs/heads/main'
+    permissions:
+      contents: read
+      packages: write
+      attestations: write
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Set up QEMU
+        uses: docker/setup-qemu-action@v3
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+
+      - name: Log in to the Container registry
+        uses: docker/login-action@v3
+        with:
+          registry: ${{ env.REGISTRY }}
+          username: ${{ github.actor }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: Extract metadata (tags, labels) for Docker
+        id: meta
+        uses: docker/metadata-action@v5
+        with:
+          images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}
+          tags: |
+            type=ref,event=tag
+            type=raw,value=latest,enable={{is_default_branch}}
+
+      - name: Extract tag version
+        id: get_version
+        run: echo "version=${GITHUB_REF#refs/tags/v}" >> $GITHUB_OUTPUT
+
+      - name: Build and push Docker image
+        uses: docker/build-push-action@v5
+        with:
+          context: .
+          platforms: linux/amd64
+          push: true
+          tags: ${{ steps.meta.outputs.tags }}
+          labels: ${{ steps.meta.outputs.labels }}
+          cache-from: type=gha
+          cache-to: type=gha,mode=max
+
+  release:
+    runs-on: ubuntu-latest
+    if: startsWith(github.ref, 'refs/tags/') && github.event.base_ref == 'refs/heads/main'
+    needs: [build-and-push-image]
+    permissions:
+      contents: write
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Create GitHub Release
+        uses: ncipollo/release-action@v1
+        with:
+          generateReleaseNotes: true
+          draft: false
+          prerelease: false

.hydra_config/chunker/base.yaml

@@ -1 +1,3 @@
-contextual_retrieval: ${oc.decode:${oc.env:CONTEXTUAL_RETRIEVAL, true}}
+contextual_retrieval: ${oc.decode:${oc.env:CONTEXTUAL_RETRIEVAL, true}}
+chunk_size: 512
+chunk_overlap_rate: 0.2

.hydra_config/chunker/markdown_splitter.yaml

@@ -1,8 +1,6 @@
 defaults:
   - base
 name: markdown_splitter
-chunk_size: 512
-chunk_overlap: 100
 
 
 # https://chat.deepseek.com/a/chat/s/28913c5d-1f62-40b0-9247-4655994fe16b

.hydra_config/chunker/recursive_splitter.yaml

@@ -1,8 +1,6 @@
 defaults:
   - base
 name: recursive_splitter
-chunk_size: 512
-chunk_overlap: 100
 
 
 # https://chat.deepseek.com/a/chat/s/28913c5d-1f62-40b0-9247-4655994fe16b

.hydra_config/chunker/semantic_splitter.yaml

@@ -1,5 +1,4 @@
 defaults:
   - base
 name: semantic_splitter
-min_chunk_size: 1000
 breakpoint_threshold_amount: 95

.hydra_config/config.yaml

@@ -36,10 +36,16 @@ vectordb:
   host: ${oc.env:VDB_HOST, milvus}
   port: ${oc.env:VDB_iPORT, 19530}
   connector_name: ${oc.env:VDB_CONNECTOR_NAME, milvus}
-  collection_name: vdb_test 
-  hybrid_mode: true
+  collection_name: ${oc.env:VDB_COLLECTION_NAME, vdb_test} 
+  hybrid_search: true
   enable: true
 
+rdb:
+  host: ${oc.env:POSTGRES_HOST, rdb}
+  port: ${oc.env:POSTGRES_PORT, 5432}
+  user: ${oc.env:POSTGRES_USER, root}
+  password: ${oc.env:POSTGRES_PASSWORD, root_password}
+
 reranker:
   enable: ${oc.decode:${oc.env:RERANKER_ENABLED, true}}
   model_name: ${oc.env:RERANKER_MODEL, Alibaba-NLP/gte-multilingual-reranker-base}
@@ -69,9 +75,21 @@ loader:
   image_captioning: true
   save_markdown: false
   audio_model: ${oc.env:WHISPER_MODEL, base} # tiny, base, small, medium, large-v1, large-v2, large-v3
+  mimetypes:
+    text/plain: .txt
+    text/markdown: .md
+    application/pdf: .pdf
+    message/rfc822: .eml
+    application/vnd.openxmlformats-officedocument.wordprocessingml.document: .docx
+    application/vnd.openxmlformats-officedocument.presentationml.presentation: .pptx
+    application/msword: .doc
+    image/png: .png
+    image/jpeg: .jpeg
+    audio/vnd.wav: .wav
+    audio/mpeg: .mp3
   file_loaders:
     txt: TextLoader
-    pdf: ${oc.env:PDFLoader, DoclingLoader}  # DoclingLoader # MarkerLoader # PyMuPDFLoader # Custompymupdf4llm
+    pdf: ${oc.env:PDFLoader, MarkerLoader}  # DoclingLoader # MarkerLoader # PyMuPDFLoader # Custompymupdf4llm
     eml: EmlLoader
     docx: MarkItDownLoader
     pptx: PPTXLoader
@@ -93,8 +111,26 @@ loader:
   marker_max_processes: ${oc.decode:${oc.env:MARKER_MAX_PROCESSES, 2}}
   marker_min_processes: ${oc.decode:${oc.env:MARKER_MIN_PROCESSES, 1}}
   marker_num_gpus: ${oc.decode:${oc.env:MARKER_NUM_GPUS, 0.01}}
+  marker_timeout: ${oc.decode:${oc.env:MARKER_TIMEOUT, 3600}}
 
 ray:
   num_gpus: ${oc.decode:${oc.env:RAY_NUM_GPUS, 0.01}}
   pool_size: ${oc.decode:${oc.env:RAY_POOL_SIZE, 1}}
   max_tasks_per_worker: ${oc.decode:${oc.env:RAY_MAX_TASKS_PER_WORKER, 5}}
+  indexer:
+    max_task_retries: ${oc.decode:${oc.env:RAY_MAX_TASK_RETRIES, 2}}
+    serialize_timeout: ${oc.decode:${oc.env:INDEXER_SERIALIZE_TIMEOUT, 36000}}
+    concurrency_groups:
+      default: ${oc.decode:${oc.env:INDEXER_DEFAULT_CONCURRENCY, 1000}}
+      update: ${oc.decode:${oc.env:INDEXER_UPDATE_CONCURRENCY, 100}}
+      search: ${oc.decode:${oc.env:INDEXER_SEARCH_CONCURRENCY, 100}}
+      delete: ${oc.decode:${oc.env:INDEXER_DELETE_CONCURRENCY, 100}}
+      chunk: ${oc.decode:${oc.env:INDEXER_CHUNK_CONCURRENCY, 1000}}
+      insert: ${oc.decode:${oc.env:INDEXER_INSERT_CONCURRENCY, 1}}
+  semaphore:
+    concurrency: ${oc.decode:${oc.env:RAY_SEMAPHORE_CONCURRENCY, 100000}}
+  serve:
+    enable: ${oc.decode:${oc.env:ENABLE_RAY_SERVE, false}}
+    num_replicas: ${oc.decode:${oc.env:RAY_SERVE_NUM_REPLICAS, 1}}
+    host: ${oc.env:RAY_SERVE_HOST, 0.0.0.0}
+    port: ${oc.env:RAY_SERVE_PORT, 8080}

.hydra_config/rag/ChatBotRag.yaml

@@ -1,3 +1,5 @@
 # Config for chatbot RAG
 mode: ChatBotRag
-chat_history_depth: 4
+chat_history_depth: 4
+max_contextualized_query_len: 512
+

Dockerfile

@@ -10,6 +10,7 @@ RUN apt-get update && apt-get install -y \
     gcc \
     cmake \
     make \
+    libpq-dev python3-dev \
     && rm -rf /var/lib/apt/lists/*
 
 # install ffmpeg
@@ -44,4 +45,4 @@ COPY prompts/ /app/prompts/
 COPY .hydra_config/ /app/.hydra_config/
 ENV PYTHONPATH=/app/openrag/
 ENV APP_iPORT=${APP_iPORT:-8080}
-ENTRYPOINT ../entrypoint.sh
+ENTRYPOINT ../entrypoint.sh

Dockerfile.ray

@@ -12,6 +12,7 @@ RUN apt-get update && apt-get install -y \
     make \
     ssh \ 
     rsync \
+    libpq-dev python3-dev \
     && rm -rf /var/lib/apt/lists/*
 
 # install ffmpeg
@@ -38,8 +39,6 @@ RUN pip3 install uv && \
 COPY entrypoint.sh /app/entrypoint.sh
 RUN chmod +x /app/entrypoint.sh
 
-COPY ray-cluster/ /app/ray-cluster/
-
 # Set workdir for source code
 WORKDIR /app/openrag