M.I.M.I.R - Multi-agent Intelligent Memory & Insight Repository

@echo off REM NornicDB Windows Build Script REM Supports multiple build variants with optional CUDA and embedded models REM REM Variants: REM cpu CPU-only, no embeddings (smallest) REM cpu-localllm CPU with llama.cpp embeddings (BYOM) REM cpu-bge CPU with llama.cpp + BGE model embedded REM cuda CUDA with llama.cpp embeddings (BYOM) REM cuda-bge CUDA with llama.cpp + BGE model embedded REM REM Usage: REM build.bat cpu Build CPU-only (no embeddings) REM build.bat cpu-localllm Build CPU with local embeddings REM build.bat cpu-bge Build CPU with embedded BGE model REM build.bat cuda Build CUDA with local embeddings REM build.bat cuda-bge Build CUDA with embedded BGE model REM build.bat download-libs Download pre-built llama.cpp libs REM build.bat download-model Download BGE model REM build.bat all Build all variants setlocal EnableDelayedExpansion set "SCRIPT_DIR=%~dp0" set "LIB_DIR=%SCRIPT_DIR%lib\llama" set "MODELS_DIR=%SCRIPT_DIR%models" set "BIN_DIR=%SCRIPT_DIR%bin" set "LLAMA_VERSION=b4785" set "RELEASE_URL=https://github.com/timothyswt/nornicdb/releases/download/libs-v1" set "MODEL_URL=https://huggingface.co/BAAI/bge-m3-gguf/resolve/main/bge-m3-Q4_K_M.gguf" REM Check arguments if "%1"=="" goto :help if "%1"=="help" goto :help if "%1"=="-h" goto :help if "%1"=="--help" goto :help if "%1"=="download-libs" goto :download_libs if "%1"=="download-model" goto :download_model if "%1"=="clean" goto :clean if "%1"=="all" goto :build_all REM Validate variant set "VARIANT=%1" set "HEADLESS=" if "%2"=="headless" set "HEADLESS=1" if "%VARIANT%"=="cpu" goto :build_cpu if "%VARIANT%"=="cpu-localllm" goto :build_cpu_localllm if "%VARIANT%"=="cpu-bge" goto :build_cpu_bge if "%VARIANT%"=="cuda" goto :build_cuda if "%VARIANT%"=="cuda-bge" goto :build_cuda_bge echo ERROR: Unknown variant '%VARIANT%' echo Run 'build.bat help' for usage goto :eof REM ============================================================================== REM Build Variants REM ============================================================================== :build_cpu echo =============================================================== echo Building: NornicDB CPU-only (no embeddings) echo Output: %BIN_DIR%\nornicdb-cpu.exe echo =============================================================== call :check_go if not exist "%BIN_DIR%" mkdir "%BIN_DIR%" set "BUILD_TAGS=" if defined HEADLESS set "BUILD_TAGS=noui" set CGO_ENABLED=0 set GOOS=windows set GOARCH=amd64 if defined HEADLESS ( go build -tags="%BUILD_TAGS%" -ldflags="-s -w" -o "%BIN_DIR%\nornicdb-cpu-headless.exe" .\cmd\nornicdb ) else ( go build -tags="%BUILD_TAGS%" -ldflags="-s -w" -o "%BIN_DIR%\nornicdb-cpu.exe" .\cmd\nornicdb ) if errorlevel 1 goto :build_failed echo. echo ✓ Build successful! echo This variant has NO embedding support - use external embedding service goto :eof :build_cpu_localllm echo =============================================================== echo Building: NornicDB CPU + Local Embeddings (BYOM) echo Output: %BIN_DIR%\nornicdb-cpu-localllm.exe echo =============================================================== call :check_go call :check_llama_libs_cpu if not exist "%BIN_DIR%" mkdir "%BIN_DIR%" set "BUILD_TAGS=localllm" if defined HEADLESS set "BUILD_TAGS=localllm noui" set CGO_ENABLED=1 set GOOS=windows set GOARCH=amd64 if defined HEADLESS ( go build -tags="%BUILD_TAGS%" -ldflags="-s -w" -o "%BIN_DIR%\nornicdb-cpu-localllm-headless.exe" .\cmd\nornicdb ) else ( go build -tags="%BUILD_TAGS%" -ldflags="-s -w" -o "%BIN_DIR%\nornicdb-cpu-localllm.exe" .\cmd\nornicdb ) if errorlevel 1 goto :build_failed echo. echo ✓ Build successful! echo Place your .gguf model in: %MODELS_DIR% echo Set: NORNICDB_EMBEDDING_PROVIDER=local goto :eof :build_cpu_bge echo =============================================================== echo Building: NornicDB CPU + Local Embeddings + BGE Model echo Output: %BIN_DIR%\nornicdb-cpu-bge.exe echo =============================================================== call :check_go call :check_llama_libs_cpu call :check_model if not exist "%BIN_DIR%" mkdir "%BIN_DIR%" set "BUILD_TAGS=localllm" if defined HEADLESS set "BUILD_TAGS=localllm noui" set CGO_ENABLED=1 set GOOS=windows set GOARCH=amd64 if defined HEADLESS ( go build -tags="%BUILD_TAGS%" -ldflags="-s -w" -o "%BIN_DIR%\nornicdb-cpu-bge-headless.exe" .\cmd\nornicdb ) else ( go build -tags="%BUILD_TAGS%" -ldflags="-s -w" -o "%BIN_DIR%\nornicdb-cpu-bge.exe" .\cmd\nornicdb ) if errorlevel 1 goto :build_failed REM Copy model alongside binary copy "%MODELS_DIR%\bge-m3.gguf" "%BIN_DIR%\" >nul 2>&1 echo. echo ✓ Build successful! echo Model embedded: %BIN_DIR%\bge-m3.gguf echo Ready to run - no additional setup needed! goto :eof :build_cuda echo =============================================================== echo Building: NornicDB CUDA + Local Embeddings (BYOM) echo Output: %BIN_DIR%\nornicdb-cuda.exe echo =============================================================== call :check_go call :check_cuda call :check_llama_libs_cuda if not exist "%BIN_DIR%" mkdir "%BIN_DIR%" set "BUILD_TAGS=cuda localllm" if defined HEADLESS set "BUILD_TAGS=cuda localllm noui" set CGO_ENABLED=1 set GOOS=windows set GOARCH=amd64 if defined HEADLESS ( go build -tags="%BUILD_TAGS%" -ldflags="-s -w" -o "%BIN_DIR%\nornicdb-cuda-headless.exe" .\cmd\nornicdb ) else ( go build -tags="%BUILD_TAGS%" -ldflags="-s -w" -o "%BIN_DIR%\nornicdb-cuda.exe" .\cmd\nornicdb ) if errorlevel 1 goto :build_failed echo. echo ✓ Build successful! echo Place your .gguf model in: %MODELS_DIR% echo Set: NORNICDB_EMBEDDING_PROVIDER=local echo Set: NORNICDB_EMBEDDING_GPU_LAYERS=-1 (all layers on GPU) goto :eof :build_cuda_bge echo =============================================================== echo Building: NornicDB CUDA + Local Embeddings + BGE Model echo Output: %BIN_DIR%\nornicdb-cuda-bge.exe echo =============================================================== call :check_go call :check_cuda call :check_llama_libs_cuda call :check_model if not exist "%BIN_DIR%" mkdir "%BIN_DIR%" set "BUILD_TAGS=cuda localllm" if defined HEADLESS set "BUILD_TAGS=cuda localllm noui" set CGO_ENABLED=1 set GOOS=windows set GOARCH=amd64 if defined HEADLESS ( go build -tags="%BUILD_TAGS%" -ldflags="-s -w" -o "%BIN_DIR%\nornicdb-cuda-bge-headless.exe" .\cmd\nornicdb ) else ( go build -tags="%BUILD_TAGS%" -ldflags="-s -w" -o "%BIN_DIR%\nornicdb-cuda-bge.exe" .\cmd\nornicdb ) if errorlevel 1 goto :build_failed REM Copy model alongside binary copy "%MODELS_DIR%\bge-m3.gguf" "%BIN_DIR%\" >nul 2>&1 echo. echo ✓ Build successful! echo Model embedded: %BIN_DIR%\bge-m3.gguf echo GPU acceleration enabled - ready to run! goto :eof :build_all echo Building all variants... echo. call :build_cpu call :build_cpu_localllm call :build_cpu_bge call :build_cuda call :build_cuda_bge echo. echo =============================================================== echo All variants built! echo =============================================================== dir /b "%BIN_DIR%\nornicdb*.exe" goto :eof REM ============================================================================== REM Checks REM ============================================================================== :check_go where go >nul 2>&1 if errorlevel 1 ( echo ERROR: Go not found in PATH. Please install Go 1.23+ echo https://go.dev/dl/ exit /b 1 ) goto :eof :check_cuda where nvcc >nul 2>&1 if errorlevel 1 ( REM Try common CUDA paths for %%v in (13.0 12.6 12.4 12.2 12.0) do ( if exist "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v%%v\bin\nvcc.exe" ( set "PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v%%v\bin;%PATH%" echo Found CUDA v%%v goto :eof ) ) echo ERROR: CUDA Toolkit not found. Please install CUDA Toolkit 12.x echo https://developer.nvidia.com/cuda-downloads exit /b 1 ) goto :eof :check_llama_libs_cpu if exist "%LIB_DIR%\libllama_windows_amd64.lib" goto :eof if exist "%LIB_DIR%\libllama_windows_amd64_cpu.lib" goto :eof echo ERROR: llama.cpp CPU libraries not found! echo. echo Option 1: Build locally echo powershell -ExecutionPolicy Bypass -File scripts\build-llama.ps1 echo. echo Option 2: Download pre-built echo build.bat download-libs echo. exit /b 1 :check_llama_libs_cuda if exist "%LIB_DIR%\libllama_windows_amd64.lib" goto :eof if exist "%LIB_DIR%\libllama_windows_amd64_cuda.lib" goto :eof echo ERROR: llama.cpp CUDA libraries not found! echo. echo Option 1: Build locally (requires VS2022 + CUDA Toolkit) echo powershell -ExecutionPolicy Bypass -File scripts\build-llama-cuda.ps1 echo. echo Option 2: Download pre-built echo build.bat download-libs echo. exit /b 1 :check_model if exist "%MODELS_DIR%\bge-m3.gguf" goto :eof echo ERROR: BGE model not found at %MODELS_DIR%\bge-m3.gguf echo. echo Download with: echo build.bat download-model echo. exit /b 1 :build_failed echo. echo ERROR: Build failed! exit /b 1 REM ============================================================================== REM Downloads REM ============================================================================== :download_libs echo =============================================================== echo Downloading pre-built llama.cpp libraries echo =============================================================== echo. if not exist "%LIB_DIR%" mkdir "%LIB_DIR%" echo Downloading CUDA library... powershell -Command "Invoke-WebRequest -Uri '%RELEASE_URL%/libllama_windows_amd64.lib' -OutFile '%LIB_DIR%\libllama_windows_amd64.lib'" if errorlevel 1 ( echo ERROR: Download failed. Build locally instead: echo powershell -ExecutionPolicy Bypass -File scripts\build-llama-cuda.ps1 goto :eof ) echo Downloading headers... for %%h in (llama.h ggml.h ggml-alloc.h ggml-backend.h ggml-cuda.h) do ( powershell -Command "Invoke-WebRequest -Uri '%RELEASE_URL%/%%h' -OutFile '%LIB_DIR%\%%h'" 2>nul ) echo. echo ✓ Libraries downloaded to %LIB_DIR% goto :eof :download_model echo =============================================================== echo Downloading BGE-M3 embedding model (~400MB) echo =============================================================== echo. if not exist "%MODELS_DIR%" mkdir "%MODELS_DIR%" echo Downloading bge-m3-Q4_K_M.gguf... powershell -Command "Invoke-WebRequest -Uri '%MODEL_URL%' -OutFile '%MODELS_DIR%\bge-m3.gguf'" if errorlevel 1 ( echo ERROR: Download failed. Download manually from: echo https://huggingface.co/BAAI/bge-m3-gguf goto :eof ) echo. echo ✓ Model downloaded to %MODELS_DIR%\bge-m3.gguf goto :eof REM ============================================================================== REM Cleanup REM ============================================================================== :clean echo Cleaning build artifacts... del /q "%BIN_DIR%\nornicdb*.exe" 2>nul echo Done. goto :eof REM ============================================================================== REM Help REM ============================================================================== :help echo =============================================================== echo NornicDB Windows Build Script echo =============================================================== echo. echo USAGE: build.bat ^<variant^> [headless] echo. echo VARIANTS: echo cpu CPU-only, no embeddings (smallest, ~15MB) echo Use with external embedding service (Ollama, OpenAI) echo. echo cpu-localllm CPU + llama.cpp embeddings, BYOM (~25MB) echo Bring Your Own Model - place .gguf in models/ echo. echo cpu-bge CPU + llama.cpp + BGE model (~425MB) echo Ready to run - no setup needed echo. echo cuda CUDA + llama.cpp embeddings, BYOM (~30MB) echo GPU-accelerated, requires NVIDIA GPU + CUDA echo. echo cuda-bge CUDA + llama.cpp + BGE model (~430MB) echo GPU-accelerated with embedded model echo. echo OPTIONS: echo headless Build without web UI (add as second argument) echo. echo COMMANDS: echo download-libs Download pre-built llama.cpp libraries echo download-model Download BGE-M3 embedding model echo all Build all variants echo clean Remove build artifacts echo help Show this help echo. echo EXAMPLES: echo build.bat cpu Build smallest variant echo build.bat cuda-bge Build GPU version with model echo build.bat cuda headless Build GPU version without UI echo build.bat download-libs Get pre-built libs first echo build.bat download-model Get BGE model echo. echo PREREQUISITES: echo All variants: Go 1.23+ (https://go.dev/dl/) echo localllm/bge: Pre-built llama.cpp libs (build.bat download-libs) echo cuda variants: CUDA Toolkit 12.x + NVIDIA GPU echo bge variants: BGE model file (build.bat download-model) echo. goto :eof