chores: fix dockerfile

Author: RobinQu

CHANGELOG.md

@@ -7,8 +7,9 @@
 * Features
   * `instinct-transformer`: New bge-m3 embedding model. Generally speaking, bge-reranker and bge-embedding are still in preview as they are not fast enough for production.
   * `instinct-llm`: New `JinaRerankerModel` for Reranker model API from Jina.ai.
-  * `instinct-retrieval`: New `DuckDBBM25Retriever` for BM25 keyword based retriever using DuckDb's built-in function. 
+  * `instinct-retrieval`: New `DuckDBBM25Retriever` for BM25 keyword based retriever using DuckDB's built-in function. 
 * Improvements
+  * Move example code to standalone repository: [instinct-cpp-examples](https://github.com/RobinQu/instinct-cpp-examples). 
   * Rename for all files for camel-case naming conventions 
   * Build system:
     * Fix include paths for internal header files. Now all files are referenced using angle bracket pattern like `#include <instinct/...>`. 

conanfile.py

@@ -12,6 +12,7 @@ class InstinctCppRecipe(ConanFile):
     settings = "os", "compiler", "build_type", "arch"
     no_copy_source = True
     options = {
+        "shared": [True, False],
         # target-level switches
         "with_duckdb": [True, False],
         "with_exprtk": [True, False],
@@ -23,7 +24,8 @@ class InstinctCppRecipe(ConanFile):
         "with_duckdb": True,
         "with_exprtk": True,
         "with_pdfium": True,
-        "with_duckx": True
+        "with_duckx": True,
+        "shared": True
     }
 
     exports_sources = ["CMakeLists.txt", "modules/*", "cmake/*", "LICENSE"]
@@ -88,6 +90,7 @@ def generate(self):
         tc.variables["WITH_EXPRTK"] = self.options.with_exprtk
         tc.variables["WITH_PDFIUM"] = self.options.with_pdfium
         tc.variables["WITH_DUCKX"] = self.options.with_duckx
+        tc.variables["BUILD_SHARED_LIBS"] = self.options.shared
         tc.generate()
 
     def build(self):

dockerfile/mini-assistant.dockerfile

@@ -17,14 +17,12 @@ RUN --mount=type=cache,id=id=conan-$TARGETARCH$TARGETVARIANT,target=/root/.conan
 RUN --mount=type=cache,id=id=conan-$TARGETARCH$TARGETVARIANT,target=/root/.conan2,sharing=locked \
     conan install . --build=missing --output-folder=build
 RUN --mount=type=cache,id=id=conan-$TARGETARCH$TARGETVARIANT,target=/root/.conan2,sharing=locked \
-    cd build && \
-    cmake .. -DCMAKE_TOOLCHAIN_FILE=conan_toolchain.cmake -DCMAKE_BUILD_TYPE=Release -DBUILD_SHARED_LIBS=ON && \
-    cmake --build . -j $(nproc)
+    conan build -c tools.build:skip_test=True -o shared=False .
 
 
 FROM ubuntu:$UBUNTU_VERSION
 
 WORKDIR /app
-COPY --from=builder /src/build/modules/instinct-examples/mini-assistant/mini-assistant /app/
+COPY --from=builder /src/build/Release/modules/instinct-apps/mini-assistant/mini-assistant /app/
 ENTRYPOINT ["/app/mini-assistant"]
 

docs/library.md

@@ -12,60 +12,46 @@ Build requirements:
 * Compiler that supports C++ 20: GCC 12+ or Clang 15+
 * Conan 2+ (Optional)
 
-### Option 1. Using `conan`
-If you are using conan to resolve all dependencies, simple run the `install` command:
+Simple run the `install` command:
 
 ```shell
 conan install conanfile.py --build=missing --output-folder=build
 ```
 
-To build and install to `/tmp/instinct.cpp`:
+As I am still working on publishing this package to conan center, you have to install to local `conan` cache:
 
-```shell
-
-cd build
-cmake .. -DCMAKE_TOOLCHAIN_FILE=conan_toolchain.cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=/tmp/instinct.cpp
-cmake --build . -j $(nproc) --target install
 ```
-
-### Option 2. Using `CMake` only
-
-Some dependencies are expected to be installed in system, and you should make sure they can be found by Cmake's `find_package` command. These libraries include:
-
-* protobuf >= 5.27.0, which brings in `absl` and `googletest` transitively.
-* ICU >= 74.1
-
-For example, on macOS you can install them using `brew`: 
-
-```shell
-brew install protobuf icu4c
+conan create . 
 ```
 
-If you prefer `conda`, which is available in Linux, macOS and Windows:
-
-```shell
-conda install -c conda-forge icu protobuf libuuid
-```
+## Quick start
 
+In your project's `conanfile.py`, add `instinct-cpp` as requirement. A working example is hosted [here](https://github.com/RobinQu/instinct-cpp-examples/tree/master/quick_start_simple).
 
-However, It's recommended to build and install these libraries from source because of following reasons:
+```py
+from conan import ConanFile
+from conan.tools.build import check_min_cppstd
 
-1. You can make sure both static and dynamic libraries are available. Some package managers like `brew` only ship dynamic libraries of protobuf so that it may prevent you from linking statically.
-2. Mainstream providers like `apt` and `yum` are shipping very old versions of these libraries in "stable" channel. 
+class YourRecipe(ConanFile):
+    settings = "os", "compiler", "build_type", "arch"
+    generators = "CMakeToolchain", "CMakeDeps"
+    
+    def validate(self):
+        check_min_cppstd(self, 20)
 
-I know it's possible to build a standalone `protobuf` using Cmake's `FetchContent`. But I haven't figured a proper way to make every pieces working together.  
+    def requirements(self):
+        # toggle off all switches
+        # you can turn on for further experiments
+        self.requires("instinct-cpp/0.1.5")
+```
 
-And then just build and install in common practices. Example to build install to `/tmp/instinct.cpp` with default project settings.
+And prepare dependencies:
 
 ```shell
-cd build
-cmake -S . -B build -DCMAKE_INSTALL_PREFIX=/tmp/instinct.cpp
-cmake --build build -j $(nproc) --config Release --target install
+conan install conanfile.py --build=missing
 ```
 
-## Quick start
-
-Let's build a simple text completion task using Ollama API.
+Once installed, let's build a simple text completion task using Ollama API.
 
 ```c++
 #include <instinct/chain/message_chain.hpp>