Update EdgeCraftRAG README and ENV (opea-project#2052)

Signed-off-by: Yongbozzz <yongbo.zhu@intel.com>

Author: Yongbozzz

EdgeCraftRAG/README.md

@@ -7,11 +7,9 @@ quality and performance.
 
 ## What's New in this release?
 
-- A sleek new UI with enhanced user experience, built on Vue and Ant Design
-- Support concurrent multi-requests handling on vLLM inference backend
-- Support pipeline configuration through json file
-- Support system prompt modification through API
-- Fixed known issues in EC-RAG UI and server
+- Chat history support for multi-session chatqna
+- Knowledge base support for EC-RAG
+- Multi-Arc support for multiple LLM inference serving engine
 
 ## Quick Start Guide
 
@@ -76,6 +74,7 @@ export RENDERGROUPID=$(getent group render | cut -d: -f3)
 # Here is the example:
 pip install --upgrade --upgrade-strategy eager "optimum[openvino]"
 
+# If below optimum-cli commands show errors, please set transformers==4.49.0 to fix: pip install transformers==4.49.0
 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 text-classification
 optimum-cli export openvino -m Qwen/Qwen2-7B-Instruct ${MODEL_PATH}/Qwen/Qwen2-7B-Instruct/INT4_compressed_weights --weight-format int4
@@ -149,6 +148,15 @@ export SELECTED_XPU_0=<which GPU to select to run for container 0>
 export SELECTED_XPU_1=<which GPU to select to run for container 1>
 ```
 
+```bash
+# Below are the extra env you can set for vllm
+export MAX_NUM_SEQS=<MAX_NUM_SEQS value>
+export MAX_NUM_BATCHED_TOKENS=<MAX_NUM_BATCHED_TOKENS value>
+export MAX_MODEL_LEN=<MAX_MODEL_LEN value>
+export LOAD_IN_LOW_BIT=<the weight type value> # expected: sym_int4, asym_int4, sym_int5, asym_int5 or sym_int8
+export CCL_DG2_USM=<CCL_DG2_USM value> # Needed on Core to enable USM (Shared Memory GPUDirect). Xeon supports P2P and doesn't need this.
+```
+
 start with compose_vllm_multi-arc.yaml
 
 ```bash

EdgeCraftRAG/docker_compose/intel/gpu/arc/compose_vllm_multi-arc.yaml

@@ -108,6 +108,11 @@ services:
       MODEL_PATH: "/llm/models"
       SERVED_MODEL_NAME: ${LLM_MODEL}
       TENSOR_PARALLEL_SIZE: ${TENSOR_PARALLEL_SIZE:-1}
+      MAX_NUM_SEQS: ${MAX_NUM_SEQS:-256}
+      MAX_NUM_BATCHED_TOKENS: ${MAX_NUM_BATCHED_TOKENS:-4000}
+      MAX_MODEL_LEN: ${MAX_MODEL_LEN:-3000}
+      IPEX_LLM_LOWBIT: ${LOAD_IN_LOW_BIT:-"fp8"}
+      CCL_DG2_USM: ${CCL_DG2_USM:-""}
       PORT: ${VLLM_SERVICE_PORT_0:-8100}
       ZE_AFFINITY_MASK: ${SELECTED_XPU_0:-0}
     shm_size: '32g'
@@ -140,6 +145,11 @@ services:
       MODEL_PATH: "/llm/models"
       SERVED_MODEL_NAME: ${LLM_MODEL}
       TENSOR_PARALLEL_SIZE: ${TENSOR_PARALLEL_SIZE:-1}
+      MAX_NUM_SEQS: ${MAX_NUM_SEQS:-256}
+      MAX_NUM_BATCHED_TOKENS: ${MAX_NUM_BATCHED_TOKENS:-4000}
+      MAX_MODEL_LEN: ${MAX_MODEL_LEN:-3000}
+      IPEX_LLM_LOWBIT: ${LOAD_IN_LOW_BIT:-"fp8"}
+      CCL_DG2_USM: ${CCL_DG2_USM:-""}
       PORT: ${VLLM_SERVICE_PORT_1:-8200}
       ZE_AFFINITY_MASK: ${SELECTED_XPU_1:-1}
     shm_size: '32g'