Merge branch 'main' into mmqna-image-query

.github/workflows/_example-workflow.yml

@@ -79,7 +79,6 @@ jobs:
           fi
           if [[ $(grep -c "vllm-gaudi:" ${docker_compose_path}) != 0 ]]; then
                git clone https://github.com/HabanaAI/vllm-fork.git
-               cd vllm-fork && git checkout 3c39626 && cd ../
           fi
           git clone https://github.com/opea-project/GenAIComps.git
           cd GenAIComps && git checkout ${{ inputs.opea_branch }} && git rev-parse HEAD && cd ../

ChatQnA/tests/test_compose_vllm_on_gaudi.sh

@@ -17,7 +17,7 @@ ip_address=$(hostname -I | awk '{print $1}')
 function build_docker_images() {
     cd $WORKPATH/docker_image_build
     git clone https://github.com/opea-project/GenAIComps.git && cd GenAIComps && git checkout "${opea_branch:-"main"}" && cd ../
-    git clone https://github.com/HabanaAI/vllm-fork.git && cd vllm-fork && git checkout 3c39626 && cd ../
+    git clone https://github.com/HabanaAI/vllm-fork.git
 
     echo "Build all the images with --no-cache, check docker_image_build.log for details..."
     service_list="chatqna chatqna-ui dataprep-redis retriever-redis vllm-gaudi nginx"

EdgeCraftRAG/README.md

@@ -8,7 +8,7 @@ quality and performance.
 ## What's New in this release?
 
 - Support image/url data retrieval and display in EC-RAG
-- Support display of LLM-used context sources in UI
+- Support display of document source used by LLM in UI
 - Support pipeline remove operation in RESTful API and UI
 - Support RAG pipeline performance benchmark and display in UI
 - Fixed known issues in EC-RAG UI and server
@@ -77,7 +77,7 @@ export RENDERGROUPID=$(getent group render | cut -d: -f3)
 pip install --upgrade --upgrade-strategy eager "optimum[openvino]"
 
 optimum-cli export openvino -m BAAI/bge-small-en-v1.5 ${MODEL_PATH}/BAAI/bge-small-en-v1.5 --task sentence-similarity
-optimum-cli export openvino -m BAAI/bge-reranker-large ${MODEL_PATH}/BAAI/bge-reranker-large --task sentence-similarity
+optimum-cli export openvino -m BAAI/bge-reranker-large ${MODEL_PATH}/BAAI/bge-reranker-large --task text-classification
 optimum-cli export openvino -m Qwen/Qwen2-7B-Instruct ${MODEL_PATH}/Qwen/Qwen2-7B-Instruct/INT4_compressed_weights --weight-format int4
 
 ```

EdgeCraftRAG/docker_compose/intel/gpu/arc/compose_vllm.yaml

@@ -67,14 +67,14 @@ services:
     ports:
       - ${VLLM_SERVICE_PORT:-8008}:80
     environment:
-      HTTPS_PROXY: ${https_proxy}
-      HTTP_PROXY: ${https_proxy}
+      no_proxy: ${no_proxy}
+      http_proxy: ${http_proxy}
+      https_proxy: ${https_proxy}
       VLLM_OPENVINO_DEVICE: GPU
       HF_ENDPOINT: ${HF_ENDPOINT}
       HF_TOKEN: ${HUGGINGFACEHUB_API_TOKEN}
     volumes:
-      - /dev/dri/by-path:/dev/dri/by-path
-      - $HOME/.cache/huggingface:/root/.cache/huggingface
+      - ${HF_CACHE:-${HOME}/.cache}:/root/.cache
     devices:
       - /dev/dri
     group_add:

EdgeCraftRAG/edgecraftrag/api/v1/data.py

@@ -53,9 +53,9 @@ async def get_files():
 
 
 # GET a file
-@data_app.get(path="/v1/data/files")
+@data_app.get(path="/v1/data/files/{name}")
 async def get_file_docs(name):
-    return ctx.get_file_mgr().get_docs_by_file(name)
+    return ctx.get_file_mgr().get_file_by_name_or_id(name)
 
 
 # DELETE a file

EdgeCraftRAG/edgecraftrag/api/v1/pipeline.py

@@ -54,7 +54,11 @@ async def add_pipeline(request: PipelineCreateIn):
             pass
         else:
             return "Unable to patch an active pipeline..."
-    update_pipeline_handler(pl, request)
+    try:
+        update_pipeline_handler(pl, request)
+    except ValueError as e:
+        ctx.get_pipeline_mgr().remove_pipeline_by_name_or_id(request.name)
+        return str(e)
     return pl
 
 
@@ -71,7 +75,10 @@ async def update_pipeline(name, request: PipelineCreateIn):
         else:
             return "Unable to patch an active pipeline..."
     async with ctx.get_pipeline_mgr()._lock:
-        update_pipeline_handler(pl, request)
+        try:
+            update_pipeline_handler(pl, request)
+        except ValueError as e:
+            return str(e)
     return pl
 
 

EdgeCraftRAG/edgecraftrag/api_schema.py

@@ -19,7 +19,7 @@ class NodeParserIn(BaseModel):
     chunk_overlap: Optional[int] = None
     chunk_sizes: Optional[list] = None
     parser_type: str
-    window_size: Optional[int] = None
+    window_size: Optional[int] = 3
 
 
 class IndexerIn(BaseModel):

EdgeCraftRAG/edgecraftrag/base.py

@@ -40,7 +40,6 @@ class FileType(str, Enum):
 
 class NodeParserType(str, Enum):
 
-    DEFAULT = "default"
     SIMPLE = "simple"
     HIERARCHY = "hierarchical"
     SENTENCEWINDOW = "sentencewindow"
@@ -49,14 +48,12 @@ class NodeParserType(str, Enum):
 
 class IndexerType(str, Enum):
 
-    DEFAULT = "default"
     FAISS_VECTOR = "faiss_vector"
     DEFAULT_VECTOR = "vector"
 
 
 class RetrieverType(str, Enum):
 
-    DEFAULT = "default"
     VECTORSIMILARITY = "vectorsimilarity"
     AUTOMERGE = "auto_merge"
     BM25 = "bm25"

EdgeCraftRAG/edgecraftrag/components/retriever.py

@@ -98,7 +98,8 @@ def run(self, **kwargs) -> Any:
         for k, v in kwargs.items():
             if k == "query":
                 nodes = cast(List[BaseNode], list(self._docstore.docs.values()))
-                bm25_retr = BM25Retriever.from_defaults(nodes=nodes, similarity_top_k=self.topk)
+                similarity_top_k = min(len(nodes), self.topk)
+                bm25_retr = BM25Retriever.from_defaults(nodes=nodes, similarity_top_k=similarity_top_k)
                 return bm25_retr.retrieve(v)
 
         return None

EdgeCraftRAG/edgecraftrag/controllers/nodemgr.py

@@ -25,7 +25,8 @@ def del_nodes(self, nodes):
         pass
 
     def del_nodes_by_np_idx(self, np_idx):
-        del self.nodes[np_idx]
+        if np_idx in self.nodes:
+            del self.nodes[np_idx]
 
     def get_nodes(self, np_idx) -> List[BaseNode]:
         if np_idx in self.nodes: