diff --git a/.github/scripts/e2e-test-llama-linux-and-mac.sh b/.github/scripts/e2e-test-llama-linux-and-mac.sh
index e97c51f63..5b7b9771d 100644
--- a/.github/scripts/e2e-test-llama-linux-and-mac.sh
+++ b/.github/scripts/e2e-test-llama-linux-and-mac.sh
@@ -4,15 +4,16 @@
 # ./linux-and-mac.sh './jan/plugins/@janhq/inference-plugin/dist/nitro/nitro_mac_arm64' https://huggingface.co/TheBloke/TinyLlama-1.1B-Chat-v0.3-GGUF/resolve/main/tinyllama-1.1b-chat-v0.3.Q2_K.gguf
 
 # Check for required arguments
-if [[ $# -ne 2 ]]; then
-    echo "Usage: $0 <path_to_binary> <url_to_download>"
+if [[ $# -ne 3 ]]; then
+    echo "Usage: $0 <path_to_binary> <url_to_download_llm> <url_to_download_embedding>"
     exit 1
 fi
 
-rm /tmp/response1.log /tmp/response2.log /tmp/nitro.log
+rm /tmp/load-llm-model-res.log /tmp/completion-res.log /tmp/unload-model-res.log /tmp/load-embedding-model-res.log /tmp/embedding-res.log /tmp/nitro.log
 
 BINARY_PATH=$1
-DOWNLOAD_URL=$2
+DOWNLOAD_LLM_URL=$2
+DOWNLOAD_EMBEDDING_URL=$3
 
 # Random port to ensure it's not used
 min=10000
@@ -37,11 +38,16 @@ sleep 5
 
 # Check if /tmp/testllm exists, if not, download it
 if [[ ! -f "/tmp/testllm" ]]; then
-    curl --connect-timeout 300 $DOWNLOAD_URL --output /tmp/testllm
+    curl --connect-timeout 300 $DOWNLOAD_LLM_URL --output /tmp/testllm
+fi
+
+# Check if /tmp/test-embedding exists, if not, download it
+if [[ ! -f "/tmp/test-embedding" ]]; then
+    curl --connect-timeout 300 $DOWNLOAD_EMBEDDING_URL --output /tmp/test-embedding
 fi
 
 # Run the curl commands
-response1=$(curl --connect-timeout 60 -o /tmp/response1.log -s -w "%{http_code}" --location "http://127.0.0.1:$PORT/inferences/llamacpp/loadModel" \
+response1=$(curl --connect-timeout 60 -o /tmp/load-llm-model-res.log -s -w "%{http_code}" --location "http://127.0.0.1:$PORT/inferences/llamacpp/loadModel" \
     --header 'Content-Type: application/json' \
     --data '{
     "llama_model_path": "/tmp/testllm",
@@ -57,7 +63,7 @@ if ! ps -p $pid >/dev/null; then
 fi
 
 response2=$(
-    curl --connect-timeout 60 -o /tmp/response2.log -s -w "%{http_code}" --location "http://127.0.0.1:$PORT/v1/chat/completions" \
+    curl --connect-timeout 60 -o /tmp/completion-res.log -s -w "%{http_code}" --location "http://127.0.0.1:$PORT/v1/chat/completions" \
         --header 'Content-Type: application/json' \
         --header 'Accept: text/event-stream' \
         --header 'Access-Control-Allow-Origin: *' \
@@ -76,16 +82,65 @@ response2=$(
      }'
 )
 
+# unload model
+response3=$(curl --connect-timeout 60 -o /tmp/unload-model-res.log --request GET -s -w "%{http_code}" --location "http://127.0.0.1:$PORT/inferences/llamacpp/unloadModel" \
+    --header 'Content-Type: application/json' \
+    --data '{
+    "llama_model_path": "/tmp/testllm"
+}')
+
+# load embedding model
+response4=$(curl --connect-timeout 60 -o /tmp/load-embedding-model-res.log -s -w "%{http_code}" --location "http://127.0.0.1:$PORT/inferences/llamacpp/loadModel" \
+    --header 'Content-Type: application/json' \
+    --data '{
+    "llama_model_path": "/tmp/test-embedding",
+    "ctx_len": 50,
+    "ngl": 32,
+    "embedding": true,
+    "model_type": "embedding"
+}')
+
+# request embedding
+response5=$(
+    curl --connect-timeout 60 -o /tmp/embedding-res.log -s -w "%{http_code}" --location "http://127.0.0.1:$PORT/v1/embeddings" \
+        --header 'Content-Type: application/json' \
+        --header 'Accept: text/event-stream' \
+        --header 'Access-Control-Allow-Origin: *' \
+        --data '{
+          "input": "Hello",
+          "model": "test-embedding",
+          "encoding_format": "float"       
+     }'
+)
+
 error_occurred=0
 if [[ "$response1" -ne 200 ]]; then
-    echo "The first curl command failed with status code: $response1"
-    cat /tmp/response1.log
+    echo "The load llm model curl command failed with status code: $response1"
+    cat /tmp/load-llm-model-res.log
     error_occurred=1
 fi
 
 if [[ "$response2" -ne 200 ]]; then
-    echo "The second curl command failed with status code: $response2"
-    cat /tmp/response2.log
+    echo "The completion curl command failed with status code: $response2"
+    cat /tmp/completion-res.log
+    error_occurred=1
+fi
+
+if [[ "$response3" -ne 200 ]]; then
+    echo "The unload model curl command failed with status code: $response3"
+    cat /tmp/unload-model-res.log
+    error_occurred=1
+fi
+
+if [[ "$response4" -ne 200 ]]; then
+    echo "The load embedding model curl command failed with status code: $response4"
+    cat /tmp/load-embedding-model-res.log
+    error_occurred=1
+fi
+
+if [[ "$response5" -ne 200 ]]; then
+    echo "The embedding curl command failed with status code: $response5"
+    cat /tmp/embedding-res.log
     error_occurred=1
 fi
 
@@ -99,11 +154,23 @@ fi
 
 echo "----------------------"
 echo "Log load model:"
-cat /tmp/response1.log
+cat /tmp/load-llm-model-res.log
+
+echo "----------------------"
+echo "Log run test:"
+cat /tmp/completion-res.log
+
+echo "----------------------"
+echo "Log run test:"
+cat /tmp/unload-model-res.log
+
+echo "----------------------"
+echo "Log run test:"
+cat /tmp/load-embedding-model-res.log
 
 echo "----------------------"
 echo "Log run test:"
-cat /tmp/response2.log
+cat /tmp/embedding-res.log
 
 echo "Nitro test run successfully!"
 
diff --git a/.github/scripts/e2e-test-llama-windows.bat b/.github/scripts/e2e-test-llama-windows.bat
index a6526f358..cddca1e0b 100644
--- a/.github/scripts/e2e-test-llama-windows.bat
+++ b/.github/scripts/e2e-test-llama-windows.bat
@@ -1,16 +1,18 @@
 @echo off
 
 set "TEMP=C:\Users\%UserName%\AppData\Local\Temp"
-set "MODEL_PATH=%TEMP%\testllm"
+set "MODEL_LLM_PATH=%TEMP%\testllm"
+set "MODEL_EMBEDDING_PATH=%TEMP%\test-embedding"
 
 rem Check for required arguments
-if "%~2"=="" (
-    echo Usage: %~0 ^<path_to_binary^> ^<url_to_download^>
+if "%~3"=="" (
+    echo Usage: %~0 ^<path_to_binary^> ^<url_to_download_llm^> ^<url_to_download_embedding^>
     exit /b 1
 )
 
 set "BINARY_PATH=%~1"
-set "DOWNLOAD_URL=%~2"
+set "DOWNLOAD_LLM_URL=%~2"
+set "DOWNLOAD_EMBEDDING_URL=%~3"
 
 for %%i in ("%BINARY_PATH%") do set "BINARY_NAME=%%~nxi"
 
@@ -18,6 +20,9 @@ echo BINARY_NAME=%BINARY_NAME%
 
 del %TEMP%\response1.log 2>nul
 del %TEMP%\response2.log 2>nul
+del %TEMP%\response3.log 2>nul
+del %TEMP%\response4.log 2>nul
+del %TEMP%\response5.log 2>nul
 del %TEMP%\nitro.log 2>nul
 
 set /a min=9999
@@ -46,33 +51,53 @@ if not defined pid (
 rem Wait for a few seconds to let the server start
 
 rem Check if %TEMP%\testmodel exists, if not, download it
-if not exist "%MODEL_PATH%" (
-    curl.exe --connect-timeout 300 %DOWNLOAD_URL% --output "%MODEL_PATH%"
+if not exist "%MODEL_LLM_PATH%" (
+    curl.exe --connect-timeout 300 %DOWNLOAD_LLM_URL% --output "%MODEL_LLM_PATH%"
 )
 
-rem Define JSON strings for curl data
-call set "MODEL_PATH_STRING=%%MODEL_PATH:\=\\%%"
-set "curl_data1={\"llama_model_path\":\"%MODEL_PATH_STRING%\"}"
-set "curl_data2={\"messages\":[{\"content\":\"Hello there\",\"role\":\"assistant\"},{\"content\":\"Write a long and sad story for me\",\"role\":\"user\"}],\"stream\":true,\"model\":\"gpt-3.5-turbo\",\"max_tokens\":50,\"stop\":[\"hello\"],\"frequency_penalty\":0,\"presence_penalty\":0,\"temperature\":0.1}"
+if not exist "%MODEL_EMBEDDING_PATH%" (
+    curl.exe --connect-timeout 300 %DOWNLOAD_EMBEDDING_URL% --output "%MODEL_EMBEDDING_PATH%"
+)
 
-rem Print the values of curl_data1 and curl_data2 for debugging
+rem Define JSON strings for curl data
+call set "MODEL_LLM_PATH_STRING=%%MODEL_LLM_PATH:\=\\%%"
+call set "MODEL_EMBEDDING_PATH_STRING=%%MODEL_EMBEDDING_PATH:\=\\%%"
+set "curl_data1={\"llama_model_path\":\"%MODEL_LLM_PATH_STRING%\"}"
+set "curl_data2={\"messages\":[{\"content\":\"Hello there\",\"role\":\"assistant\"},{\"content\":\"Write a long and sad story for me\",\"role\":\"user\"}],\"stream\":false,\"model\":\"gpt-3.5-turbo\",\"max_tokens\":50,\"stop\":[\"hello\"],\"frequency_penalty\":0,\"presence_penalty\":0,\"temperature\":0.1}"
+set "curl_data3={\"llama_model_path\":\"%MODEL_LLM_PATH_STRING%\"}"
+set "curl_data4={\"llama_model_path\":\"%MODEL_EMBEDDING_PATH_STRING%\", \"embedding\": true, \"model_type\": \"embedding\"}"
+set "curl_data5={\"input\": \"Hello\", \"model\": \"test-embedding\", \"encoding_format\": \"float\"}"
+
+rem Print the values of curl_data for debugging
 echo curl_data1=%curl_data1%
 echo curl_data2=%curl_data2%
+echo curl_data3=%curl_data3%
+echo curl_data4=%curl_data4%
+echo curl_data5=%curl_data5%
 
 rem Run the curl commands and capture the status code
 curl.exe --connect-timeout 60 -o "%TEMP%\response1.log" -s -w "%%{http_code}" --location "http://127.0.0.1:%PORT%/inferences/llamacpp/loadModel" --header "Content-Type: application/json" --data "%curl_data1%" > %TEMP%\response1.log 2>&1
 
 curl.exe --connect-timeout 60 -o "%TEMP%\response2.log" -s -w "%%{http_code}" --location "http://127.0.0.1:%PORT%/inferences/llamacpp/chat_completion" ^
 --header "Content-Type: application/json" ^
---header "Accept: text/event-stream" ^
---header "Access-Control-Allow-Origin: *" ^
 --data "%curl_data2%" > %TEMP%\response2.log 2>&1
 
+curl.exe --connect-timeout 60 -o "%TEMP%\response3.log" --request GET -s -w "%%{http_code}" --location "http://127.0.0.1:%PORT%/inferences/llamacpp/unloadModel" --header "Content-Type: application/json" --data "%curl_data3%" > %TEMP%\response3.log 2>&1
+
+curl.exe --connect-timeout 60 -o "%TEMP%\response4.log" --request POST -s -w "%%{http_code}" --location "http://127.0.0.1:%PORT%/inferences/llamacpp/loadModel" --header "Content-Type: application/json" --data "%curl_data4%" > %TEMP%\response4.log 2>&1
+
+curl.exe --connect-timeout 60 -o "%TEMP%\response5.log" -s -w "%%{http_code}" --location "http://127.0.0.1:%PORT%/v1/embeddings" ^
+--header "Content-Type: application/json" ^
+--data "%curl_data5%" > %TEMP%\response5.log 2>&1
+
 set "error_occurred=0"
 
 rem Read the status codes from the log files
 for /f %%a in (%TEMP%\response1.log) do set "response1=%%a"
 for /f %%a in (%TEMP%\response2.log) do set "response2=%%a"
+for /f %%a in (%TEMP%\response3.log) do set "response3=%%a"
+for /f %%a in (%TEMP%\response4.log) do set "response4=%%a"
+for /f %%a in (%TEMP%\response5.log) do set "response5=%%a"
 
 if "%response1%" neq "200" (
     echo The first curl command failed with status code: %response1%
@@ -86,6 +111,24 @@ if "%response2%" neq "200" (
     set "error_occurred=1"
 )
 
+if "%response3%" neq "200" (
+    echo The third curl command failed with status code: %response3%
+    type %TEMP%\response3.log
+    set "error_occurred=1"
+)
+
+if "%response4%" neq "200" (
+    echo The fourth curl command failed with status code: %response4%
+    type %TEMP%\response4.log
+    set "error_occurred=1"
+)
+
+if "%response5%" neq "200" (
+    echo The fifth curl command failed with status code: %response5%
+    type %TEMP%\response5.log
+    set "error_occurred=1"
+)
+
 if "%error_occurred%"=="1" (
     echo Nitro test run failed!!!!!!!!!!!!!!!!!!!!!!
     echo Nitro Error Logs:
@@ -96,13 +139,25 @@ if "%error_occurred%"=="1" (
 
 
 echo ----------------------
-echo Log load model:
+echo Log load llm model:
 type %TEMP%\response1.log
 
 echo ----------------------
-echo "Log run test:"
+echo Log run test:
 type %TEMP%\response2.log
 
+echo ----------------------
+echo Log unload model:
+type %TEMP%\response3.log
+
+echo ----------------------
+echo Log load embedding model:
+type %TEMP%\response3.log
+
+echo ----------------------
+echo Log run embedding test:
+type %TEMP%\response5.log
+
 echo Nitro test run successfully!
 
 rem Kill the server process
diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index 5d6f44115..2705ba701 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -49,6 +49,7 @@ env:
   BRANCH_NAME: ${{ github.head_ref || github.ref_name }}
   LLM_MODEL_URL: https://delta.jan.ai/tinyllama-1.1b-chat-v0.3.Q2_K.gguf
   WHISPER_MODEL_URL: https://delta.jan.ai/ggml-tiny-q5_1.bin
+  EMBEDDING_MODEL_URL: https://catalog.jan.ai/dist/models/embeds/nomic-embed-text-v1.5.f16.gguf
 
 jobs:
   create-draft-release:
@@ -186,7 +187,7 @@ jobs:
         run: |
           # run e2e testing
           cd nitro
-          chmod +x ../.github/scripts/e2e-test-llama-linux-and-mac.sh && ../.github/scripts/e2e-test-llama-linux-and-mac.sh ./nitro ${{ env.LLM_MODEL_URL }}
+          chmod +x ../.github/scripts/e2e-test-llama-linux-and-mac.sh && ../.github/scripts/e2e-test-llama-linux-and-mac.sh ./nitro ${{ env.LLM_MODEL_URL }} ${{ env.EMBEDDING_MODEL_URL }}
           rm -rf uploads/
 
       - name: Run e2e testing - Whisper.CPP
@@ -307,7 +308,7 @@ jobs:
         run: |
           # run e2e testing
           cd nitro/
-          chmod +x ../.github/scripts/e2e-test-llama-linux-and-mac.sh && ../.github/scripts/e2e-test-llama-linux-and-mac.sh ./nitro ${{ env.LLM_MODEL_URL }}
+          chmod +x ../.github/scripts/e2e-test-llama-linux-and-mac.sh && ../.github/scripts/e2e-test-llama-linux-and-mac.sh ./nitro ${{ env.LLM_MODEL_URL }} ${{ env.EMBEDDING_MODEL_URL }}
           rm -rf uploads/
 
       - name: Run e2e testing - Whisper.CPP
@@ -373,7 +374,7 @@ jobs:
         run: |
           # run e2e testing
           cd nitro
-          chmod +x ../.github/scripts/e2e-test-llama-linux-and-mac.sh && ../.github/scripts/e2e-test-llama-linux-and-mac.sh ./nitro ${{ env.LLM_MODEL_URL }}
+          chmod +x ../.github/scripts/e2e-test-llama-linux-and-mac.sh && ../.github/scripts/e2e-test-llama-linux-and-mac.sh ./nitro ${{ env.LLM_MODEL_URL }} ${{ env.EMBEDDING_MODEL_URL }}
           rm -rf uploads/
 
       - name: Run e2e testing - Whisper.CPP
@@ -519,7 +520,7 @@ jobs:
         if: ${{ matrix.build != 'arm64' && matrix.build != 'amd64-vulkan' && matrix.build != 'amd64-avx512' }}
         run: |
           cd build\Release
-          ..\..\.github\scripts\e2e-test-llama-windows.bat nitro.exe ${{ env.LLM_MODEL_URL }}
+          ..\..\.github\scripts\e2e-test-llama-windows.bat nitro.exe ${{ env.LLM_MODEL_URL }} ${{ env.EMBEDDING_MODEL_URL }}
           rmdir /S /Q .\build\Release\uploads
 
       - name: Run e2e testing - Whisper.cpp