Add multithreaded logging infrastructure for preparation for future designs (#225)

* Update class diagrams

* Implement BoundedChannel and multithreaded logging infra

* Integrate logging channel system

* Update string concatenations to use std::format

* Add pretty print log

.gitattributes

@@ -1 +1 @@
-archive/* linguist-vendored
+archive/** linguist-vendored

docs/pipeline/README.md

@@ -18,6 +18,7 @@ descriptions via a local GGUF model or a deterministic mock.
   - [Build](#build)
   - [Model](#model)
   - [Run](#run)
+- [Docker / RunPod](#docker--runpod)
 - [Architecture](#architecture)
   - [Pipeline Stages](#pipeline-stages)
   - [Key Components](#key-components)
@@ -51,7 +52,7 @@ step.
 
 ### Build
 
-Requirements: C++20 compiler, CMake 3.24+, libcurl, Boost (JSON and
+Requirements: C++20 compiler, CMake 3.31+, OpenSSL, Boost (JSON and
 ProgramOptions). SQLite is fetched from the upstream amalgamation, so no system
 SQLite package is required.
 
@@ -60,6 +61,16 @@ cmake -S . -B build
 cmake --build build
 ```
 
+CMake automatically detects whether a compatible llama.cpp installation is
+present on the system (`libllama`, `libggml`, `libggml-base`, and `llama.h`
+visible on the default search paths). If found, it links against those
+libraries and skips the FetchContent build. If not found, it fetches and builds
+llama.cpp from source at tag `b9012`. No additional flags are required in
+either case.
+
+Metal is enabled automatically on Apple Silicon. CUDA or HIP/ROCm is detected
+automatically on Linux when the relevant toolkit is present.
+
 ### Model
 
 > Skip this step if you only need `--mocked`.
@@ -74,33 +85,124 @@ curl -L \
 ### Run
 
 Run from `build/` so the copied `locations.json` and `prompts/` are available.
-Each run also writes a fresh dated SQLite file such as
+Each run writes a fresh dated SQLite file such as
 `biergarten_seed_2026-04-19T15-30-45.123456Z.sqlite` into the working directory.
 
 ```bash
 ./biergarten-pipeline --mocked
-./biergarten-pipeline --model models/google_gemma-4-E4B-it-Q6_K.gguf --temperature 1.0 --top-p 0.95 --top-k 64 --n-ctx 8192 --seed -1
+
+./biergarten-pipeline \
+  --model ../models/google_gemma-4-E4B-it-Q6_K.gguf \
+  --prompt-dir prompts \
+  --temperature 1.0 --top-p 0.95 --top-k 64 --n-ctx 8192 --seed -1
 ```
 
 #### CLI Flags
 
-| Flag            | Purpose                                                 |
+| Flag            | Purpose                                                                                              |
-| --------------- | ------------------------------------------------------- |
+| --------------- | ---------------------------------------------------------------------------------------------------- |
-| `--mocked`      | Deterministic mock generator, no model required.        |
+| `--mocked`      | Deterministic mock generator, no model required.                                                     |
-| `--model, -m`   | Path to a GGUF file. Required unless `--mocked` is set. |
+| `--model, -m`   | Path to a GGUF file. Required unless `--mocked` is set.                                              |
-| `--temperature` | Sampling temperature. Default: `1.0`.                   |
+| `--prompt-dir`  | Directory containing prompt files (e.g. `BREWERY_GENERATION.md`). Required unless `--mocked` is set. |
-| `--top-p`       | Nucleus sampling. Default: `0.95`.                      |
+| `--output, -o`  | Directory for generated SQLite artifacts. Default: `output`.                                         |
-| `--top-k`       | Top-k sampling. Default: `64`.                          |
+| `--log-path`    | Path for application logs. Default: `pipeline.log`.                                                  |
-| `--n-ctx`       | Context window size. Default: `8192`.                   |
+| `--temperature` | Sampling temperature. Default: `1.0`.                                                                |
-| `--seed`        | Random seed. Default: `-1` (random at runtime).         |
+| `--top-p`       | Nucleus sampling. Default: `0.95`.                                                                   |
-| `--help, -h`    | Print usage and exit.                                   |
+| `--top-k`       | Top-k sampling. Default: `64`.                                                                       |
+| `--n-ctx`       | Context window size. Default: `8192`.                                                                |
+| `--seed`        | Random seed. Default: `-1` (random at runtime).                                                      |
+| `--help, -h`    | Print usage and exit.                                                                                |
 
 `--mocked` and `--model` are mutually exclusive. Omitting both exits with an
 error before the pipeline starts. Sampling flags are ignored when `--mocked` is
 set.
 
 The post-build step copies `prompts/` into `build/prompts/`. Rebuild after
-editing `prompts/system.md`.
+editing any prompt file.
+
+---
+
+## Docker / RunPod
+
+The `tooling/pipeline/runpod/` directory contains a GPU-ready container
+configuration for running the pipeline on RunPod or any Docker host with an
+NVIDIA GPU.
+
+### How it works
+
+The container uses a two-stage build. The first stage pulls prebuilt
+`libllama`, `libggml`, and backend plugin libraries (including `libggml-cuda.so`
+and the CPU variant plugins) from `ghcr.io/ggml-org/llama.cpp:full-cuda`. The
+second stage copies those libraries into `/usr/local/lib` and runs `ldconfig` so
+the dynamic linker and `dlopen` calls from `ggml_backend_load_all()` can resolve
+the CUDA backend plugin at runtime. llama.cpp headers are cloned at the matching
+tag and installed into `/usr/local/include`. CMake auto-detects both and skips
+the FetchContent source build entirely, keeping image build times short.
+
+`GGML_BACKEND_PATH` is set to `/usr/local/lib` so llama.cpp knows where to scan
+for backend plugins.
+
+### Build the image
+
+Run from the `tooling/pipeline/` directory (the CMake project root), not from
+inside `runpod/`, so the `COPY . .` step picks up the full project context.
+
+```bash
+docker build -t biergarten-pipeline:latest -f runpod/Dockerfile .
+```
+
+To monitor the full build output and confirm CMake selects the system llama.cpp:
+
+```bash
+docker build \
+  --progress=plain \
+  --no-cache \
+  -t biergarten-pipeline:latest \
+  -f runpod/Dockerfile \
+  . 2>&1 | tee build.log
+```
+
+Look for `[biergarten] Found system llama.cpp — skipping FetchContent` in the
+output to confirm the fast path was taken.
+
+### Run in mocked mode
+
+No model or GPU required. Useful for validating the pipeline logic and SQLite
+export path.
+
+```bash
+docker run --rm \
+  -e BIERGARTEN_MODE=mocked \
+  -v "$PWD/output:/workspace/output" \
+  -v "$PWD/logs:/workspace/logs" \
+  biergarten-pipeline:latest
+```
+
+### Run in live mode
+
+Mount your GGUF model before starting. The container validates the model path
+before launching the binary.
+
+```bash
+docker run --rm \
+  --runtime=nvidia \
+  -e BIERGARTEN_MODE=live \
+  -e GGML_BACKEND_PATH="/usr/local/lib/libggml-cuda.so" \
+  -v "$PWD/models:/workspace/models" \
+  -v "$PWD/output:/workspace/output" \
+  -v "$PWD/logs:/workspace/logs" \
+  biergarten-pipeline:latest
+```
+
+The model must be present at `./models/google_gemma-4-E4B-it-Q6_K.gguf` on the
+host. See [Model](#model) above for the download command.
+
+### RunPod deployment
+
+Use a GPU pod template. Mount persistent storage for `/workspace/models`,
+`/workspace/output`, and `/workspace/logs`. Set `BIERGARTEN_MODE=live` in the
+template environment. See `tooling/pipeline/runpod/pod-template.yaml` for a
+starter template.
 
 ---
 
@@ -197,16 +299,18 @@ code, latitude, and longitude for each entry.
 ## Tech Stack
 
 - C++20
-- CMake 3.24+
+- CMake 3.31+
 - Boost.JSON, Boost.ProgramOptions, Boost.DI
 - spdlog
-- libcurl
+- cpp-httplib (with OpenSSL)
 - SQLite amalgamation fetched and compiled via CMake FetchContent
-- llama.cpp
+- llama.cpp (auto-detected from system install or fetched via FetchContent)
+- Docker with NVIDIA CUDA 12.6 base image for GPU container builds
+- RunPod for cloud GPU inference
 
-The build fetches Boost.DI, spdlog, llama.cpp, and SQLite via CMake. Metal is
+The build fetches Boost.DI, spdlog, and SQLite via CMake. llama.cpp is fetched
-enabled on Apple Silicon; CUDA or HIP/ROCm is detected on Linux when the toolkit
+only when a system installation is not detected. Metal is enabled on Apple
-is present.
+Silicon; CUDA or HIP/ROCm is detected on Linux when the toolkit is present.
 
 > **Code Style:** Modern C++20 throughout — RAII for ownership,
 > `std::unique_ptr` for injected dependencies, `std::optional` for parse
@@ -218,7 +322,7 @@ is present.
 
 ## Tested Hardware
 
-### ARM macOS - M1 Pro
+### ARM macOS — M1 Pro
 
 |           |                                   |
 | --------- | --------------------------------- |
@@ -229,7 +333,7 @@ is present.
 | Model     | Gemma 4 E4B                       |
 | Inference | llama.cpp with Metal              |
 
-### x86_64 Linux - NVIDIA RTX 2000
+### x86_64 Linux — NVIDIA RTX 2000
 
 |           |                                |
 | --------- | ------------------------------ |
@@ -240,6 +344,15 @@ is present.
 | Model     | Gemma 4 E4B                    |
 | Inference | llama.cpp with CUDA 12.x       |
 
+### x86_64 Linux — Docker / RunPod (NVIDIA CUDA)
+
+|           |                                             |
+| --------- | ------------------------------------------- |
+| Host      | RunPod GPU pod                              |
+| Base      | nvidia/cuda:12.6.3-devel-ubuntu24.04        |
+| Model     | Gemma 4 E4B Q6_K                            |
+| Inference | llama.cpp prebuilt CUDA backends via dlopen |
+
 ---
 
 ## Fixture Strategy
@@ -260,8 +373,9 @@ is present.
 | `includes/`                  | Public headers and shared models.                  |
 | `src/`                       | Implementation files.                              |
 | `locations.json`             | Curated city input copied into the build tree.     |
-| `prompts/`                   | System prompt used by the model-backed path.       |
+| `prompts/`                   | System prompts used by the model-backed path.      |
 | `diagrams/`                  | Architecture and pipeline diagrams.                |
+| `tooling/pipeline/runpod/`   | Dockerfile, launcher, and RunPod pod template.     |
 | `ETHICS-AND-KNOWN-ISSUES.md` | Ethics, bias, hallucination analysis, mitigations. |
 
 ---
@@ -276,6 +390,7 @@ is present.
 - `src/data_generation/llama/` — local inference, prompt loading, output
   validation.
 - `src/data_generation/mock/` — deterministic fallback.
+- `tooling/pipeline/runpod/` — container build and runtime launcher.
 
 ---
 

docs/pipeline/diagrams/current/activity.puml

@@ -29,7 +29,7 @@ if (Are arguments valid?) then (no)
 else (yes)
 endif
 
-:Init CurlGlobalState & LlamaBackendState;
+:Init OpenSSL global state & LlamaBackendState;
 :di::make_injector(...);
 :injector.create<std::unique_ptr<BiergartenDataGenerator>>();
 :BiergartenDataGenerator::Run();

docs/pipeline/diagrams/current/class.puml

@@ -26,6 +26,7 @@ skinparam note {
 title The Biergarten Data Pipeline - Class Diagram
 
 class BiergartenDataGenerator {
+  - logger_ : std::shared_ptr<ILogger>
   - context_service_ : std::unique_ptr<IEnrichmentService>
   - generator_ : std::unique_ptr<DataGenerator>
   - exporter_ : std::unique_ptr<IExportService>
@@ -36,6 +37,46 @@ class BiergartenDataGenerator {
   - LogResults() : void
 }
 
+class LogLevel <<enumeration>> {
+  Debug
+  Info
+  Warn
+  Error
+}
+
+class PipelinePhase <<enumeration>> {
+  Startup
+  UserGeneration
+  BreweryAndBeerGeneration
+  CheckinGeneration
+  RatingGeneration
+  FollowGeneration
+  Teardown
+}
+
+struct LogEntry {
+  + timestamp : std::chrono::system_clock::time_point
+  + level : LogLevel
+  + phase : PipelinePhase
+  + message : std::string
+  + worker : std::optional<std::string>
+}
+
+interface ILogger <<interface>> {
+  + Log(entry : const LogEntry&) : void
+}
+
+class LogProducer {
+  - channel_ : BoundedChannel<LogEntry>&
+  + Log(entry : const LogEntry&) : void
+}
+
+class LogDispatcher {
+  - channel_ : BoundedChannel<LogEntry>&
+  + Run() : void
+  - ToSpdlogLevel(level) : spdlog::level::level_enum
+}
+
 interface IEnrichmentService <<interface>> {
   + GetLocationContext(loc : const Location&) : std::string
 }
@@ -52,7 +93,7 @@ interface WebClient <<interface>> {
   + UrlEncode(value : const std::string&) : std::string
 }
 
-class CURLWebClient {
+class HttpWebClient {
   + Get(url : const std::string&) : std::string
   + UrlEncode(value : const std::string&) : std::string
 }
@@ -123,14 +164,21 @@ class SystemDateTimeProvider {
 }
 
 ' Structural Relationships / Dependency Injection
+BiergartenDataGenerator *-- ILogger : owns
 BiergartenDataGenerator *-- IEnrichmentService : owns
 BiergartenDataGenerator *-- DataGenerator : owns
 BiergartenDataGenerator *-- IExportService : owns
 
+LogEntry *-- LogLevel
+LogEntry *-- PipelinePhase
+ILogger <|.. LogProducer : implements
+LogProducer ..> LogEntry : emits
+LogDispatcher ..> LogEntry : consumes
+
 IEnrichmentService <|.. WikipediaService : implements
 WikipediaService *-- WebClient : owns
 
-WebClient <|.. CURLWebClient : implements
+WebClient <|.. HttpWebClient : implements
 
 DataGenerator <|.. MockGenerator : implements
 DataGenerator <|.. LlamaGenerator : implements

docs/pipeline/diagrams/planned/activity.puml

@@ -13,7 +13,7 @@ if (Invalid args?) then (yes)
   stop
 else (no)
 endif
-:Init CurlGlobalState & LlamaBackendState;
+:Init OpenSSL global state & LlamaBackendState;
 :Build DI injector;
 
 :Initialize SqliteExportService;

docs/pipeline/diagrams/planned/class.puml

@@ -1,4 +1,4 @@
-@startuml
+@startuml class_diagram
 
 ' ==========================================
 ' CONFIGURATION & STYLING
@@ -8,6 +8,8 @@ skinparam classAttributeFontSize 9
 skinparam defaultFontSize 25
 skinparam titleFontSize 30
 
+title Biergarten Data Pipeline — Class Diagram
+
 package "Domain: Models" {
 
   class Location {
@@ -141,7 +143,7 @@ package "Domain: Models" {
 
   LocationContext *-- Completeness
 }
-@startuml
+
 package "Domain: Application Configuration" {
   class SamplingOptions {
     + temperature: float = 1.0F
@@ -167,12 +169,10 @@ package "Domain: Application Configuration" {
     + pipeline: PipelineOptions
   }
 
-  ' --- Domain Model Relationships ---
   ApplicationOptions *-- GeneratorOptions
   ApplicationOptions *-- PipelineOptions
   GeneratorOptions o-- SamplingOptions
 }
-@endum
 
 package "Domain: Policy" {
 
@@ -275,33 +275,29 @@ package "Infrastructure: Logging" {
     + level : LogLevel
     + phase : PipelinePhase
     + message : std::string
-    + city : std::optional<std::string>
-    + entity_id : std::optional<std::string>
     + worker : std::optional<std::string>
   }
 
-  interface Logger <<interface>> {
+  interface ILogger <<interface>> {
-    + Log(level, phase, message,\n  city, entity_id, worker) : void
+    + Log(entry : const LogEntry&) : void
   }
 
-  class PipelineLogger {
+  class LogProducer {
-    - log_ch_ : BoundedChannel<LogEntry>&
+    - channel_ : BoundedChannel<LogEntry>&
-    + Log(level, phase, message,\n  city, entity_id, worker) : void
+    + Log(entry : const LogEntry&) : void
   }
 
-  class LogWorker {
+  class LogDispatcher {
-    - log_ch_ : BoundedChannel<LogEntry>&
+    - channel_ : BoundedChannel<LogEntry>&
     + Run() : void
-    - FormatTimestamp(tp) : std::string
     - ToSpdlogLevel(level) : spdlog::level::level_enum
-    - ToString(phase) : std::string
   }
 
-  ' --- Logging Relationships ---
   LogEntry       *--  LogLevel
   LogEntry       *--  PipelinePhase
-  PipelineLogger ..>  LogEntry       : emits
+  ILogger        <|.. LogProducer
-  LogWorker      ..>  LogEntry       : consumes
+  LogProducer    ..>  LogEntry       : emits
+  LogDispatcher  ..>  LogEntry       : consumes
 }
 
 package "Infrastructure: Pipeline Channel" {
@@ -356,13 +352,29 @@ package "Infrastructure: Enrichment" {
     + UrlEncode(value : const std::string&) : std::string
   }
 
-  class CURLWebClient {
+  class HttpWebClient {
     + Get(url : const std::string&) : std::string
     + UrlEncode(value : const std::string&) : std::string
   }
 
 }
 
+package "Infrastructure: Prompting" {
+
+  interface IPromptDirectory <<interface>> {
+    + Load(key : std::string_view) : std::string
+  }
+
+  class PromptDirectory {
+    - prompt_dir_ : std::filesystem::path
+    - cache_ : std::unordered_map<std::string, std::string>
+    + PromptDirectory(prompt_dir : const std::filesystem::path&)
+    + Load(key : std::string_view) : std::string
+  }
+
+  IPromptDirectory <|.. PromptDirectory
+}
+
 package "Infrastructure: Data Generation" {
 
   interface DataGenerator <<interface>> {
@@ -386,6 +398,7 @@ package "Infrastructure: Data Generation" {
     - model_ : ModelHandle
     - context_ : ContextHandle
     - prompt_formatter_ : std::unique_ptr<PromptFormatter>
+    - prompt_directory_ : std::unique_ptr<IPromptDirectory>
     - rng_ : std::mt19937
     + GenerateBrewery(...) : BreweryResult
     + GenerateBeer(...) : BeerResult
@@ -459,8 +472,6 @@ package "Infrastructure: Data Export" {
 
 }
 
-
-
 class BiergartenPipelineOrchestrator {
   - preloader_ : std::unique_ptr<DataPreloader>
   - enrichment_service_ : std::unique_ptr<EnrichmentService>
@@ -520,7 +531,7 @@ CheckinDistributionStrategy    <|.. RandomCheckinStrategy
 FollowGenerationStrategy       <|.. RandomFollowStrategy
 FollowGenerationStrategy       <|.. ActivityWeightedFollowStrategy
 EnrichmentService              <|.. WikipediaService
-WebClient                      <|.. CURLWebClient
+WebClient                      <|.. HttpWebClient
 DataGenerator                  <|.. MockGenerator
 DataGenerator                  <|.. LlamaGenerator
 PromptFormatter                <|.. Gemma4JinjaPromptFormatter
@@ -531,6 +542,7 @@ DateTimeProvider               <|.. SystemDateTimeProvider
 WikipediaService               *--  WebClient
 WikipediaService               ..>  ContextStrategy
 LlamaGenerator                 *--  PromptFormatter
+LlamaGenerator                 *--  IPromptDirectory
 LlamaGenerator                 ..>  GeneratorOptions
 SqliteExportService            *--  DateTimeProvider
 

tooling/pipeline/.dockerignore

@@ -0,0 +1,9 @@
+build/
+cmake-build-debug/
+.git/
+.idea/
+**/*.sqlite
+**/*.log
+**/*.sqlite3
+**/*.db
+

tooling/pipeline/CMakeLists.txt

@@ -1,41 +1,45 @@
 cmake_minimum_required(VERSION 3.31)
 project(biergarten-pipeline)
 
+# Set policy to allow FetchContent_Populate for header-only libraries
+# that have outdated CMakeLists.txt files
+cmake_policy(SET CMP0169 OLD)
+
 # 1. Build Options
 
 option(BIERGARTEN_MOCK_ONLY "Build with mock data generators only — skips llama.cpp" OFF)
-if (BIERGARTEN_MOCK_ONLY)
+if(BIERGARTEN_MOCK_ONLY)
-    message(STATUS "[biergarten] MOCK_ONLY build — llama.cpp will not be compiled.")
+        message(STATUS "[biergarten] MOCK_ONLY build — llama.cpp will not be compiled.")
-endif ()
+endif()
 
 # 2. Platform & GPU Detection
-if (NOT UNIX)
+if(NOT UNIX)
-    message(FATAL_ERROR "[biergarten] Windows is not supported. Please use Linux (Fedora 43) or macOS (M1 Pro).")
+        message(FATAL_ERROR "[biergarten] Windows is not supported. Please use Linux (Fedora 43) or macOS (M1 Pro).")
-endif ()
+endif()
 
-if (APPLE)
+if(APPLE)
-    if (CMAKE_SYSTEM_PROCESSOR MATCHES "arm64")
+        if(CMAKE_SYSTEM_PROCESSOR MATCHES "arm64")
-        message(STATUS "[biergarten] Apple Silicon detected — enabling Metal acceleration.")
+                message(STATUS "[biergarten] Apple Silicon detected — enabling Metal acceleration.")
-        set(GGML_METAL ON CACHE BOOL "Enable Metal for Apple Silicon" FORCE)
+                set(GGML_METAL ON CACHE BOOL "Enable Metal for Apple Silicon" FORCE)
-    else ()
+        else()
-        message(STATUS "[biergarten] Intel Mac detected — using CPU / Accelerate framework.")
+                message(STATUS "[biergarten] Intel Mac detected — using CPU / Accelerate framework.")
-        set(GGML_METAL OFF CACHE BOOL "Disable Metal for Intel Macs" FORCE)
+                set(GGML_METAL OFF CACHE BOOL "Disable Metal for Intel Macs" FORCE)
-    endif ()
+        endif()
-else ()
+else()
-    find_package(CUDAToolkit QUIET)
+        find_package(CUDAToolkit QUIET)
-    find_package(hip CONFIG QUIET)
+        find_package(hip CONFIG QUIET)
 
-    if (CUDAToolkit_FOUND)
+        if(CUDAToolkit_FOUND)
-        message(STATUS "[biergarten] NVIDIA GPU detected — enabling CUDA acceleration.")
+                message(STATUS "[biergarten] NVIDIA GPU detected — enabling CUDA acceleration.")
-        set(GGML_CUDA ON CACHE BOOL "Enable CUDA for NVIDIA GPUs" FORCE)
+                set(GGML_CUDA ON CACHE BOOL "Enable CUDA for NVIDIA GPUs" FORCE)
-        set(CMAKE_CUDA_ARCHITECTURES native)
+                set(CMAKE_CUDA_ARCHITECTURES native)
-    elseif (hip_FOUND OR DEFINED ENV{ROCM_PATH} OR EXISTS "/opt/rocm")
+        elseif(hip_FOUND OR DEFINED ENV{ROCM_PATH} OR EXISTS "/opt/rocm")
-        message(STATUS "[biergarten] AMD GPU detected — enabling HIP/ROCm acceleration.")
+                message(STATUS "[biergarten] AMD GPU detected — enabling HIP/ROCm acceleration.")
-        set(GGML_HIPBLAS ON CACHE BOOL "Enable HIP for AMD GPUs" FORCE)
+                set(GGML_HIPBLAS ON CACHE BOOL "Enable HIP for AMD GPUs" FORCE)
-    else ()
+        else()
-        message(STATUS "[biergarten] No NVIDIA or AMD GPU found — falling back to CPU.")
+                message(STATUS "[biergarten] No NVIDIA or AMD GPU found — falling back to CPU.")
-    endif ()
+        endif()
-endif ()
+endif()
 
 # 3. Project-wide Settings
 set(CMAKE_CXX_STANDARD 20)
@@ -51,16 +55,23 @@ include(FetchContent)
 find_package(Boost REQUIRED COMPONENTS json program_options)
 
 # Boost.DI (unofficial Boost extension, must declare separately from main Boost dependency)
+# Header-only library, so we only fetch without invoking its CMakeLists.txt
 FetchContent_Declare(
         boost-di
         GIT_REPOSITORY https://github.com/boost-ext/di.git
         GIT_TAG v1.3.0
+        GIT_SHALLOW TRUE
 )
-FetchContent_MakeAvailable(boost-di)
+FetchContent_GetProperties(boost-di)
-if (TARGET Boost.DI AND NOT TARGET boost::di)
+if(NOT boost-di_POPULATED)
-    add_library(boost::di ALIAS Boost.DI)
+        FetchContent_Populate(boost-di)
-endif ()
+endif()
 
+add_library(boost_di INTERFACE)
+add_library(boost::di ALIAS boost_di)
+target_include_directories(boost_di INTERFACE
+        $<BUILD_INTERFACE:${boost-di_SOURCE_DIR}/include>
+)
 # SQLite amalgamation
 FetchContent_Declare(
         sqlite_amalgamation
@@ -69,21 +80,38 @@ FetchContent_Declare(
         EXCLUDE_FROM_ALL
 )
 FetchContent_MakeAvailable(sqlite_amalgamation)
-if (NOT TARGET sqlite3)
+if(NOT TARGET sqlite3)
-    add_library(sqlite3 STATIC ${sqlite_amalgamation_SOURCE_DIR}/sqlite3.c)
+        add_library(sqlite3 STATIC ${sqlite_amalgamation_SOURCE_DIR}/sqlite3.c)
-    target_include_directories(sqlite3 PUBLIC ${sqlite_amalgamation_SOURCE_DIR})
+        target_include_directories(sqlite3 PUBLIC ${sqlite_amalgamation_SOURCE_DIR})
-    target_compile_definitions(sqlite3 PUBLIC SQLITE_THREADSAFE=1)
+        target_compile_definitions(sqlite3 PUBLIC SQLITE_THREADSAFE=1)
-endif ()
+endif()
 
 # llama.cpp — skipped for mock-only builds
-if (NOT BIERGARTEN_MOCK_ONLY)
+if(NOT BIERGARTEN_MOCK_ONLY)
-    FetchContent_Declare(
+        find_library(LLAMA_LIB NAMES llama)
-            llama-cpp
+        find_library(GGML_LIB NAMES ggml)
-            GIT_REPOSITORY https://github.com/ggml-org/llama.cpp.git
+        find_library(GGML_BASE_LIB NAMES ggml-base)
-            GIT_TAG b8742
+        find_path(LLAMA_INC_DIR NAMES llama.h PATH_SUFFIXES include)
-    )
+
-    FetchContent_MakeAvailable(llama-cpp)
+        if(LLAMA_LIB AND GGML_LIB AND GGML_BASE_LIB AND LLAMA_INC_DIR)
-endif ()
+                message(STATUS "[biergarten] Found system llama.cpp — skipping FetchContent")
+
+                add_library(llama SHARED IMPORTED)
+                set_target_properties(llama PROPERTIES
+                        IMPORTED_LOCATION "${LLAMA_LIB}"
+                        INTERFACE_INCLUDE_DIRECTORIES "${LLAMA_INC_DIR}"
+                        INTERFACE_LINK_LIBRARIES "${GGML_LIB};${GGML_BASE_LIB}"
+                )
+        else()
+                message(STATUS "[biergarten] System llama.cpp not found — fetching via FetchContent")
+                FetchContent_Declare(
+                        llama-cpp
+                        GIT_REPOSITORY https://github.com/ggml-org/llama.cpp.git
+                        GIT_TAG b9012
+                )
+                FetchContent_MakeAvailable(llama-cpp)
+        endif()
+endif()
 
 # spdlog
 FetchContent_Declare(
@@ -109,7 +137,9 @@ set(HTTPLIB_REQUIRE_OPENSSL ON CACHE BOOL "Require OpenSSL for cpp-httplib" FORC
 FetchContent_MakeAvailable(cpp-httplib)
 
 # 5. Executable & Sources
-add_executable(${PROJECT_NAME})
+add_executable(${PROJECT_NAME}
+        includes/services/enrichment/mock_enrichment.h
+        includes/json_handling/pretty_print.h)
 
 # --- Entry point ---
 target_sources(${PROJECT_NAME} PRIVATE
@@ -126,13 +156,13 @@ target_sources(${PROJECT_NAME} PRIVATE
         src/application_options/parse_arguments.cc
 )
 
-# --- biergarten_data_generator ---
+# --- biergarten_pipeline_orchestrator ---
 target_sources(${PROJECT_NAME} PRIVATE
-        src/biergarten_data_generator/log_results.cc
+        src/biergarten_pipeline_orchestrator/log_results.cc
-        src/biergarten_data_generator/biergarten_data_generator.cc
+        src/biergarten_pipeline_orchestrator/biergarten_pipeline_orchestrator.cc
-        src/biergarten_data_generator/generate_breweries.cc
+        src/biergarten_pipeline_orchestrator/generate_breweries.cc
-        src/biergarten_data_generator/run.cc
+        src/biergarten_pipeline_orchestrator/run.cc
-        src/biergarten_data_generator/query_cities_with_countries.cc
+        src/biergarten_pipeline_orchestrator/query_cities_with_countries.cc
 )
 
 # --- web_client ---
@@ -153,44 +183,48 @@ target_sources(${PROJECT_NAME} PRIVATE
 )
 
 # --- data_generation: llama (skipped for mock-only builds) ---
-if (NOT BIERGARTEN_MOCK_ONLY)
+if(NOT BIERGARTEN_MOCK_ONLY)
-    target_sources(${PROJECT_NAME} PRIVATE
+        target_sources(${PROJECT_NAME} PRIVATE
-            src/data_generation/llama/load.cc
+                src/data_generation/llama/load.cc
-            src/data_generation/llama/helpers.cc
+                src/data_generation/llama/helpers.cc
-            src/data_generation/llama/generate_brewery.cc
+                src/data_generation/llama/generate_brewery.cc
-            src/data_generation/llama/infer.cc
+                src/data_generation/llama/infer.cc
-            src/data_generation/llama/llama_generator.cc
+                src/data_generation/llama/llama_generator.cc
-            src/data_generation/llama/generate_user.cc
+                src/data_generation/llama/generate_user.cc
-    )
+        )
-endif ()
+endif()
 
 # --- services: wikipedia ---
 target_sources(${PROJECT_NAME} PRIVATE
-        src/services/wikipedia/wikipedia_service.cc
+        src/services/enrichment/wikipedia/wikipedia_service.cc
-        src/services/wikipedia/fetch_extract.cc
+        src/services/enrichment/wikipedia/fetch_extract.cc
-        src/services/wikipedia/get_summary.cc
+        src/services/enrichment/wikipedia/get_summary.cc
 )
 
 # --- services: sqlite ---
 target_sources(${PROJECT_NAME} PRIVATE
-        src/services/sqlite/process_record.cc
+	src/services/sqlite/process_record.cc
-        src/services/sqlite/sqlite_export_service.cc
+	src/services/sqlite/sqlite_export_service.cc
-        src/services/sqlite/finalize.cc
+	src/services/sqlite/finalize.cc
-        src/services/sqlite/initialize.cc
+	src/services/sqlite/initialize.cc
-        src/services/sqlite/helpers/sqlite_connection_helpers.cc
+	src/services/sqlite/helpers/sqlite_connection_helpers.cc
-        src/services/sqlite/helpers/sqlite_statement_helpers.cc
+	src/services/sqlite/helpers/sqlite_statement_helpers.cc
+)
+
+# --- services: logging ---
+target_sources(${PROJECT_NAME} PRIVATE
+	"src/services/logging/log_producer.cc"
+        src/services/logging/log_dispatcher.cc
 )
 
 # --- services (top-level) ---
 target_sources(${PROJECT_NAME} PRIVATE
-        src/services/prompt_directory.cc
+	src/services/prompt_directory.cc
 )
 
 # 6. Include Directories, Link Libraries & Compile Definitions
 target_include_directories(${PROJECT_NAME} PRIVATE
         includes
-        $<$<NOT:$<BOOL:${BIERGARTEN_MOCK_ONLY}>>:${llama-cpp_SOURCE_DIR}/include>
-        $<$<NOT:$<BOOL:${BIERGARTEN_MOCK_ONLY}>>:${llama-cpp_SOURCE_DIR}/common>
 )
 
 target_link_libraries(${PROJECT_NAME} PRIVATE
@@ -215,6 +249,11 @@ target_compile_definitions(${PROJECT_NAME} PRIVATE
         $<$<CONFIG:Debug>:DEBUG>
 )
 
+target_compile_options(biergarten-pipeline PRIVATE
+   -fmacro-prefix-map=${CMAKE_SOURCE_DIR}/tooling/pipeline/src/=
+)
+
+
 # 7. Runtime Assets
 configure_file(
         ${CMAKE_SOURCE_DIR}/locations.json
@@ -225,4 +264,5 @@ add_custom_command(TARGET ${PROJECT_NAME} POST_BUILD
         COMMAND ${CMAKE_COMMAND} -E copy_directory
         ${CMAKE_SOURCE_DIR}/prompts
         ${CMAKE_BINARY_DIR}/prompts
-)
+)
+

tooling/pipeline/includes/biergarten_pipeline_orchestrator.h

@@ -3,7 +3,10 @@
 
 /**
  * @file biergarten_data_generator.h
- * @brief Core orchestration class for pipeline data generation.
+ * @brief Orchestration for end-to-end brewery data generation pipeline.
+ *
+ * Intent: Coordinates location loading, enrichment, and generation phases
+ * to produce a complete dataset. Coordinates dependencies via composition root.
  */
 
 #include <memory>
@@ -12,8 +15,10 @@
 
 #include "data_generation/data_generator.h"
 #include "data_model/generated_models.h"
-#include "services/enrichment/enrichment_service.h"
 #include "services/database/export_service.h"
+#include "services/enrichment/enrichment_service.h"
+
+#include "services/logging/logger.h"
 
 /**
  * @brief Main data generator class for the Biergarten pipeline.
@@ -21,18 +26,22 @@
  * This class encapsulates the core logic for generating brewery data.
  * It handles location loading, city enrichment, and brewery generation.
  */
-class BiergartenDataGenerator {
+class BiergartenPipelineOrchestrator {
  public:
-  /**
+/**
-   * @brief Construct a BiergartenDataGenerator with injected dependencies.
+ * @brief Constructs the orchestrator with injected pipeline dependencies.
-   *
+ *
-   * @param context_service Context provider for sampled locations.
+ * @param context_service Provides regional context for locations.
-   * @param generator Brewery and user data generator.
+ * @param generator Implementation (Llama or Mock) for brewery/user generation.
-   * @param exporter Storage backend for generated brewery data.
+ * @param exporter Database backend for persisting generated records.
-   */
+ * @param application_options CLI configuration and paths.
-  BiergartenDataGenerator(std::unique_ptr<IEnrichmentService> context_service,
+ */
-                          std::unique_ptr<DataGenerator> generator,
+  BiergartenPipelineOrchestrator(
-                          std::unique_ptr<IExportService> exporter);
+      std::shared_ptr<ILogger> logger,
+      std::unique_ptr<IEnrichmentService> context_service,
+      std::unique_ptr<DataGenerator> generator,
+      std::unique_ptr<IExportService> exporter,
+      const ApplicationOptions& application_options);
 
   /**
    * @brief Run the data generation pipeline.
@@ -42,26 +51,38 @@ class BiergartenDataGenerator {
    * 2. Resolve context for each city using the injected context service
    * 3. Generate brewery data for sampled cities
    *
+   * @note STRUCTURAL CONCURRENCY REQUIREMENT:
+   * When transitioned to a multithreaded design, this method MUST structurally
+   * enforce that all deployed worker threads are joined before returning (e.g.
+   * by using std::jthread or a structured concurrency primitive). This ensures
+   * workers do not attempt to log to a closed channel during application teardown.
+   *
    * @return true if successful, false if not
    */
   bool Run();
 
  private:
+  /// @brief Logger instance for emitting pipeline messages.
+  std::shared_ptr<ILogger> logger_;
+
   /// @brief Owning context provider dependency.
   std::unique_ptr<IEnrichmentService> context_service_;
 
   /// @brief Generator dependency selected in the composition root.
   std::unique_ptr<DataGenerator> generator_;
 
-  /// @brief Storage backend for generated brewery records.
+   /// @brief Storage backend for generated brewery records.
-  std::unique_ptr<IExportService> exporter_;
+   std::unique_ptr<IExportService> exporter_;
+
+   /// @brief CLI configuration: paths, model settings, generation parameters.
+   ApplicationOptions application_options_;
 
   /**
    * @brief Load locations from JSON and sample cities.
    *
    * @return Vector of sampled locations capped at 50 entries.
    */
-  static std::vector<Location> QueryCitiesWithCountries();
+  std::vector<Location> QueryCitiesWithCountries();
 
   /**
    * @brief Generate breweries for enriched cities.

tooling/pipeline/includes/concurrency/bounded_channel.h

@@ -0,0 +1,73 @@
+#ifndef BIERGARTEN_PIPELINE_INCLUDES_CONCURRENCY_BOUNDED_CHANNEL_H_
+#define BIERGARTEN_PIPELINE_INCLUDES_CONCURRENCY_BOUNDED_CHANNEL_H_
+
+#include <condition_variable>
+#include <cstddef>
+#include <mutex>
+#include <optional>
+#include <queue>
+
+/**
+ * @file bounded_channel.h
+ * @brief Thread-safe, bounded multi-producer/multi-consumer synchronous channel.
+ *
+ * Intent: Enables asynchronous inter-thread communication with backpressure.
+ * Models a synchronous channel where producers/consumers block on capacity limits.
+ */
+
+/**
+ * @class BoundedChannel
+ * @brief MPMC channel with fixed capacity and blocking semantics.
+ *
+ * Producers block when buffer is full; consumers block when empty.
+ * Close() unblocks all waiters and signals channel exhaustion.
+ */
+template <typename T>
+class BoundedChannel {
+  // -------------------------------------------------------------------------
+  // Internal state — all access must be guarded by mutex_.
+  // -------------------------------------------------------------------------
+
+  std::queue<T> queue_;
+
+  std::mutex mutex_;
+
+  std::condition_variable not_full_;
+
+  std::condition_variable not_empty_;
+
+  std::size_t capacity_;
+
+  bool closed_ = false;
+
+ public:
+  /**
+   * @brief Construct a bounded channel with the given capacity.
+   * @param capacity Maximum number of items the channel may hold.
+   */
+  explicit BoundedChannel(std::size_t capacity) : capacity_(capacity) {}
+
+  /**
+   * @brief Send an item into the channel. Blocks when the channel is full.
+   * @param item Move-only item to enqueue.
+   */
+  void Send(T item);
+
+  /**
+   * @brief Receive an item from the channel. Blocks when the channel is
+   * empty.
+   * @return std::optional<T> containing the item, or std::nullopt when the
+   * channel is closed and drained.
+   */
+  std::optional<T> Receive();
+
+  /**
+   * @brief Close the channel and unblock all waiting threads. Idempotent.
+   */
+  void Close();
+};
+
+// Include the template implementation
+#include "bounded_channel.tcc"
+
+#endif  // BIERGARTEN_PIPELINE_INCLUDES_CONCURRENCY_BOUNDED_CHANNEL_H_

tooling/pipeline/includes/concurrency/bounded_channel.tcc

@@ -0,0 +1,57 @@
+#include "bounded_channel.h"
+
+template <typename T>
+void BoundedChannel<T>::Send(T item) {
+  // Acquire exclusive ownership of the mutex; released automatically on scope exit.
+  std::unique_lock lock(mutex_);
+
+  // Block until there is space in the queue or the channel has been closed.
+  // The predicate guards against spurious wakeups.
+  not_full_.wait(lock, [&] { return queue_.size() < capacity_ || closed_; });
+
+  // If the channel was closed while waiting, discard the item and return.
+  if (closed_) return;
+
+  // Move the item into the queue to avoid an unnecessary copy.
+  queue_.push(std::move(item));
+
+  // Wake one blocked Receive() call to signal that data is now available.
+  not_empty_.notify_one();
+}
+
+template <typename T>
+std::optional<T> BoundedChannel<T>::Receive() {
+  // Acquire exclusive ownership of the mutex.
+  std::unique_lock lock(mutex_);
+
+  // Block until the queue is non-empty or the channel has been closed.
+  // The predicate guards against spurious wakeups.
+  not_empty_.wait(lock, [&] { return !queue_.empty() || closed_; });
+
+  // If woken due to closure and no items remain, signal exhaustion via nullopt.
+  if (queue_.empty()) return std::nullopt;
+
+  // Move the front item out of the queue to avoid an unnecessary copy.
+  T item = std::move(queue_.front());
+  queue_.pop();
+
+  // Wake one blocked Send() call to signal that a slot has opened.
+  not_full_.notify_one();
+
+  return item;
+}
+
+template <typename T>
+void BoundedChannel<T>::Close() {
+  // Acquire exclusive ownership of the mutex to ensure visibility of the flag.
+  std::unique_lock lock(mutex_);
+
+  // Mark the channel as closed; subsequent Send() calls will be dropped.
+  closed_ = true;
+
+  // Wake all blocked Send() callers so they can observe the closed flag and exit.
+  not_full_.notify_all();
+
+  // Wake all blocked Receive() callers so they can drain remaining items or return nullopt.
+  not_empty_.notify_all();
+}

tooling/pipeline/includes/data_generation/llama_generator.h

@@ -14,10 +14,11 @@
 #include <string>
 #include <string_view>
 
+#include "../services/prompting/prompt_directory.h"
 #include "data_generation/data_generator.h"
 #include "data_generation/prompt_formatting/prompt_formatter.h"
 #include "data_model/models.h"
-#include "../services/prompting/prompt_directory.h"
+#include "services/logging/logger.h"
 
 struct llama_model;
 struct llama_context;
@@ -37,7 +38,7 @@ class LlamaGenerator final : public DataGenerator {
    * @param prompt_directory Directory service for loading named prompt files.
    */
   LlamaGenerator(const ApplicationOptions& options,
-                 const std::string& model_path,
+                 const std::string& model_path, std::shared_ptr<ILogger> logger,
                  std::unique_ptr<IPromptFormatter> prompt_formatter,
                  std::unique_ptr<IPromptDirectory> prompt_directory);
 
@@ -129,6 +130,8 @@ class LlamaGenerator final : public DataGenerator {
   uint32_t sampling_top_k_ = kDefaultSamplingTopK;
   std::mt19937 rng_;
   uint32_t n_ctx_ = kDefaultContextSize;
+  int n_gpu_layers_ = 0;
+  std::shared_ptr<ILogger> logger_;
   std::unique_ptr<IPromptFormatter> prompt_formatter_;
   std::unique_ptr<IPromptDirectory> prompt_directory_;
 };

tooling/pipeline/includes/data_model/models.h

@@ -3,17 +3,21 @@
 
 /**
  * @file data_model/models.h
- * @brief Core data models: locations, application configuration, and generation inputs.
+ * @brief Core data models: locations, application configuration, and generation
+ * inputs.
  */
 
 #include <boost/program_options.hpp>
 #include <cstdint>
 #include <filesystem>
+#include <memory>
 #include <optional>
 #include <string>
 #include <string_view>
 #include <vector>
 
+class ILogger;
+
 namespace prog_opts = boost::program_options;
 
 // ============================================================================
@@ -82,6 +86,9 @@ struct SamplingOptions {
 
   /// @brief Random seed (-1 for random, otherwise non-negative).
   int seed = -1;
+
+  /// @brief Number of layers to offload to GPU.
+  int n_gpu_layers = 0;
 };
 
 /**
@@ -94,6 +101,8 @@ struct GeneratorOptions {
   /// @brief Use mocked generator instead of actual LLM inference.
   bool use_mocked = false;
 
+
+
   /// @brief Specific sampling parameters for this generator.
   /// If nullopt, the application should use global defaults.
   std::optional<SamplingOptions> sampling;
@@ -112,6 +121,10 @@ struct PipelineOptions {
 
   /// @brief Path for application logs.
   std::filesystem::path log_path;
+
+  /// @brief Number of locations to sample from the dataset
+  /// More locations -> more users/more breweries
+  uint32_t location_count;
 };
 
 /**
@@ -126,6 +139,7 @@ struct ApplicationOptions {
 // Function Declarations
 // ============================================================================
 
-std::optional<ApplicationOptions> ParseArguments(const int argc, char** argv);
+std::optional<ApplicationOptions> ParseArguments(const int argc, char** argv,
+                                                std::shared_ptr<ILogger> logger = nullptr);
 
 #endif  // BIERGARTEN_PIPELINE_INCLUDES_DATA_MODEL_MODELS_H_

tooling/pipeline/includes/json_handling/json_loader.h

@@ -7,16 +7,19 @@
  */
 
 #include <filesystem>
+#include <memory>
 #include <vector>
 
 #include "data_model/models.h"
+#include "services/logging/logger.h"
 
 /// @brief Loads curated world locations from a JSON file into memory.
 class JsonLoader {
  public:
   /// @brief Parses a JSON array file and returns all location records.
   static std::vector<Location> LoadLocations(
-      const std::filesystem::path& filepath);
+      const std::filesystem::path& filepath,
+      std::shared_ptr<ILogger> logger = nullptr);
 };
 
 #endif  // BIERGARTEN_PIPELINE_INCLUDES_JSON_HANDLING_JSON_LOADER_H_

tooling/pipeline/includes/json_handling/pretty_print.h

@@ -0,0 +1,109 @@
+#ifndef BIERGARTEN_PIPELINE_INCLUDES_JSON_HANDLING_PRETTY_PRINT_H_
+#define BIERGARTEN_PIPELINE_INCLUDES_JSON_HANDLING_PRETTY_PRINT_H_
+
+/**
+ * @file json_handling/pretty_print.h
+ * @brief Pretty-printing utilities for JSON values.
+ *
+ * Provides formatting capability for boost::json::value with indentation and
+ * readable output. Adapted from Boost JSON library examples.
+ */
+
+#include <boost/json.hpp>
+#include <ostream>
+#include <string>
+
+/**
+ * @brief Pretty-prints a JSON value to an output stream with indentation.
+ *
+ * Recursively formats JSON objects and arrays with consistent 4-space
+ * indentation. Adapted from:
+ * https://raw.githubusercontent.com/boostorg/json/refs/heads/develop/example/pretty.cpp
+ *
+ * @param outstream Output stream to write formatted JSON.
+ * @param json_val JSON value to format.
+ * @param indent Optional indentation string (managed internally on first call).
+ */
+inline void PrettyPrint(std::ostream& outstream,
+                        boost::json::value const& json_val,
+                        std::string* indent = nullptr) {
+  std::string str;
+  if (indent == nullptr) {
+    indent = &str;
+  }
+  switch (json_val.kind()) {
+    case boost::json::kind::object: {
+      outstream << "{\n";
+      indent->append(4, ' ');
+      auto const& obj = json_val.get_object();
+      if (!obj.empty()) {
+        const auto* iter = obj.begin();
+        for (;;) {
+          outstream << *indent << boost::json::serialize(iter->key()) << " : ";
+          PrettyPrint(outstream, iter->value(), indent);
+          iter = std::next(iter);
+          if (iter == obj.end()) {
+            break;
+          }
+
+          outstream << ",\n";
+        }
+      }
+      outstream << "\n";
+      indent->resize(indent->size() - 4);
+      outstream << *indent << "}";
+      break;
+    }
+
+    case boost::json::kind::array: {
+      outstream << "[\n";
+      indent->append(4, ' ');
+      auto const& arr = json_val.get_array();
+      if (!arr.empty()) {
+        const auto* iter = arr.begin();
+        for (;;) {
+          outstream << *indent;
+          PrettyPrint(outstream, *iter, indent);
+          iter = std::next(iter);
+          if (iter == arr.end()) {
+            break;
+          }
+          outstream << ",\n";
+        }
+      }
+      outstream << "\n";
+      indent->resize(indent->size() - 4);
+      outstream << *indent << "]";
+      break;
+    }
+
+    case boost::json::kind::string: {
+      outstream << serialize(json_val.get_string());
+      break;
+    }
+
+    case boost::json::kind::uint64:
+    case boost::json::kind::int64:
+    case boost::json::kind::double_:
+      outstream << json_val;
+      break;
+
+    case boost::json::kind::bool_:
+      if (json_val.get_bool()) {
+        outstream << "true";
+      } else {
+        outstream << "false";
+      }
+      break;
+
+    case boost::json::kind::null:
+      outstream << "null";
+      break;
+  }
+
+  if (indent->empty()) {
+    outstream << "\n";
+  }
+}
+
+#endif

tooling/pipeline/includes/services/enrichment/mock_enrichment.h

@@ -0,0 +1,17 @@
+//
+// Created by aaronpo on 13/05/2026.
+//
+
+#ifndef BIERGARTEN_PIPELINE_INCLUDES_SERVICES_ENRICHMENT_MOCK_ENRICHMENT_H_
+#define BIERGARTEN_PIPELINE_INCLUDES_SERVICES_ENRICHMENT_MOCK_ENRICHMENT_H_
+#include <string>
+
+#include "enrichment_service.h"
+
+class MockEnrichmentService final : public IEnrichmentService {
+ public:
+  std::string GetLocationContext(const Location& /*loc*/) override {
+    return {};
+  }
+};
+#endif  // BIERGARTEN_PIPELINE_INCLUDES_SERVICES_ENRICHMENT_MOCK_ENRICHMENT_H_

tooling/pipeline/includes/services/enrichment/wikipedia_service.h

@@ -12,13 +12,15 @@
 #include <unordered_map>
 
 #include "enrichment_service.h"
+#include "services/logging/logger.h"
 #include "web_client/web_client.h"
 
 /// @brief Provides Wikipedia summary lookups backed by cached raw extracts.
-class WikipediaService final : public IEnrichmentService {
+class WikipediaEnrichmentService final : public IEnrichmentService {
  public:
   /// @brief Creates a new Wikipedia service with the provided web client.
-  explicit WikipediaService(std::unique_ptr<WebClient> client);
+  explicit WikipediaEnrichmentService(std::unique_ptr<WebClient> client,
+                                      std::shared_ptr<ILogger> logger);
 
   /// @brief Returns the Wikipedia-derived context for a location.
   [[nodiscard]] std::string GetLocationContext(const Location& loc) override;
@@ -26,6 +28,7 @@ class WikipediaService final : public IEnrichmentService {
  private:
   std::string FetchExtract(std::string_view query);
   std::unique_ptr<WebClient> client_;
+  std::shared_ptr<ILogger> logger_;
   /// @brief Canonical cache for raw Wikipedia query extracts.
   std::unordered_map<std::string, std::string> extract_cache_;
 };

tooling/pipeline/includes/services/logging/log_dispatcher.h

@@ -0,0 +1,53 @@
+/**
+ * @file services/logging/log_dispatcher.h
+ * @brief Dedicated log dispatcher for asynchronous pipeline logging.
+ *
+ * The dispatcher drains LogEntry values from a bounded channel and forwards
+ * them to spdlog on a dedicated thread.
+ */
+
+#ifndef BIERGARTEN_PIPELINE_INCLUDES_SERVICES_LOGGING_LOG_DISPATCHER_H_
+#define BIERGARTEN_PIPELINE_INCLUDES_SERVICES_LOGGING_LOG_DISPATCHER_H_
+
+#include <spdlog/spdlog.h>
+
+#include "concurrency/bounded_channel.h"
+#include "services/logging/log_entry.h"
+
+/**
+ * @class LogDispatcher
+ * @brief Consumes log entries from a channel and forwards them to spdlog.
+ *
+ * Non-copyable and non-movable. Intended to run on its own dedicated thread
+ * and exit once the channel has been closed and drained.
+ */
+class LogDispatcher {
+ public:
+  /**
+   * @brief Construct a log dispatcher.
+   *
+   * @param channel Reference to the bounded channel used for log retrieval.
+   */
+  explicit LogDispatcher(BoundedChannel<LogEntry>& channel);
+
+  LogDispatcher(const LogDispatcher&) = delete;
+  LogDispatcher& operator=(const LogDispatcher&) = delete;
+  LogDispatcher(LogDispatcher&&) = delete;
+  LogDispatcher& operator=(LogDispatcher&&) = delete;
+  ~LogDispatcher() = default;
+
+  /**
+   * @brief Drain the channel and forward entries to spdlog.
+   *
+   * Intended to be called once on a dedicated thread. The loop returns after
+   * the channel has been closed and all queued entries have been processed.
+   */
+  void Run();
+
+ private:
+  BoundedChannel<LogEntry>& channel_;
+
+  static spdlog::level::level_enum ToSpdlogLevel(LogLevel level);
+};
+
+#endif  // BIERGARTEN_PIPELINE_INCLUDES_SERVICES_LOGGING_LOG_DISPATCHER_H_

tooling/pipeline/includes/services/logging/log_entry.h

@@ -0,0 +1,88 @@
+/**
+ * @file services/logging/log_entry.h
+ * @brief Structured log record shared by the pipeline logging infra.
+ *
+ * LogEntry is a lightweight value type that can be passed safely between the
+ * logging producer and dispatcher through BoundedChannel<LogEntry>.
+ */
+
+#ifndef BIERGARTEN_PIPELINE_INCLUDES_SERVICES_LOGGING_LOG_ENTRY_H_
+#define BIERGARTEN_PIPELINE_INCLUDES_SERVICES_LOGGING_LOG_ENTRY_H_
+
+#include <chrono>
+#include <source_location>
+#include <string>
+#include <thread>
+#include <vector>
+
+/**
+ * @enum LogLevel
+ * @brief Severity levels supported by the logging infra.
+ */
+enum class LogLevel {
+  Debug,  ///< Development/debugging information.
+  Info,   ///< General informational messages.
+  Warn,   ///< Warning conditions.
+  Error,  ///< Error conditions.
+};
+
+/**
+ * @enum PipelinePhase
+ * @brief Pipeline execution phases used to tag log records.
+ *
+ * The phase tag makes it easier to correlate log output with the part of the
+ * pipeline that emitted it.
+ */
+enum class PipelinePhase {
+  Startup,                   ///< Initialization and validation.
+  UserGeneration,            ///< User profile generation.
+  BreweryAndBeerGeneration,  ///< Brewery and beer data generation.
+  CheckinGeneration,         ///< Checkin (visit) record generation.
+  RatingGeneration,          ///< Rating and review generation.
+  FollowGeneration,          ///< Follow relationship generation.
+  Teardown,                  ///< Finalization and cleanup.
+};
+
+/**
+ * @struct LogDTO
+ * @brief User-provided subset of log fields. Used to capture call-site info transparently.
+ */
+struct LogDTO {
+  LogLevel level;
+  PipelinePhase phase;
+  std::string message;
+};
+
+/**
+ * @struct LogEntry
+ * @brief Single structured log event.
+ *
+ * All fields are value types, which keeps transfer across the bounded channel
+ * simple and avoids shared ownership.
+ *
+ * NOTE: timestamp, thread_id, and origin must be populated by ILogger::Log()
+ * before the entry is dispatched.
+ */
+struct LogEntry {
+  /// @brief Timestamp when the entry was created.
+  std::chrono::system_clock::time_point timestamp{};
+
+  /// @brief Source location where the log call was made.
+  std::source_location origin{};
+
+  /// @brief Thread responsible for emitting the log.
+  std::thread::id thread_id{};
+
+
+  /// @brief Severity level of this entry.
+  LogLevel level;
+
+  /// @brief Pipeline phase associated with the entry.
+  PipelinePhase phase;
+
+  /// @brief Log message text.
+  std::string message;
+
+};
+
+#endif  // BIERGARTEN_PIPELINE_INCLUDES_SERVICES_LOGGING_LOG_ENTRY_H_

tooling/pipeline/includes/services/logging/log_producer.h

@@ -0,0 +1,53 @@
+/**
+ * @file services/logging/log_producer.h
+ * @brief Channel-backed log producer for asynchronous pipeline logging.
+ *
+ * The producer captures log records from application code and forwards them to
+ * a bounded channel for later processing by the dispatcher.
+ */
+
+#ifndef BIERGARTEN_PIPELINE_INCLUDES_SERVICES_LOGGING_CHANNEL_LOGGER_H_
+#define BIERGARTEN_PIPELINE_INCLUDES_SERVICES_LOGGING_CHANNEL_LOGGER_H_
+
+#include <string_view>
+
+#include "concurrency/bounded_channel.h"
+#include "services/logging/log_entry.h"
+#include "services/logging/logger.h"
+
+/**
+ * @class LogProducer
+ * @brief ILogger implementation that forwards entries to a bounded channel.
+ *
+ * Non-copyable and non-movable. The channel reference is non-owning and must
+ * remain valid for the lifetime of the producer.
+ */
+class LogProducer final : public ILogger {
+ public:
+  /**
+   * @brief Construct a channel-backed producer.
+   *
+   * @param channel Reference to the bounded channel used for log transfer.
+   */
+  explicit LogProducer(BoundedChannel<LogEntry>& channel);
+
+  LogProducer(const LogProducer&) = delete;
+  LogProducer& operator=(const LogProducer&) = delete;
+  LogProducer(LogProducer&&) = delete;
+  LogProducer& operator=(LogProducer&&) = delete;
+
+  ~LogProducer() override = default;
+
+  /**
+   * @brief Queue a log message for asynchronous processing.
+   *
+   * Blocks while the channel applies backpressure. This blocking behavior
+   * under heavy load is an accepted trade-off for simplicity.
+   */
+  void DoLog(LogEntry log_entry) override;
+
+ private:
+  BoundedChannel<LogEntry>& channel_;
+};
+
+#endif  // BIERGARTEN_PIPELINE_INCLUDES_SERVICES_LOGGING_CHANNEL_LOGGER_H_

tooling/pipeline/includes/services/logging/logger.h

@@ -0,0 +1,64 @@
+/**
+ * @file services/logging/logger.h
+ * @brief Abstract logging interface used by pipeline components.
+ *
+ * The interface keeps application code independent from the concrete logging
+ * transport, buffering, and formatting implementation.
+ */
+
+#ifndef BIERGARTEN_PIPELINE_INCLUDES_SERVICES_LOGGING_LOGGER_H_
+#define BIERGARTEN_PIPELINE_INCLUDES_SERVICES_LOGGING_LOGGER_H_
+
+#include <source_location>
+#include <string>
+#include <utility>
+
+#include "services/logging/log_entry.h"
+
+/**
+ * @class ILogger
+ * @brief Minimal interface for submitting structured log messages.
+ *
+ * Implementations are non-copyable and non-movable. They are typically owned
+ * by the composition root and injected into services that emit diagnostics.
+ */
+class ILogger {
+ public:
+  ILogger() = default;
+  ILogger(const ILogger&) = delete;
+  ILogger& operator=(const ILogger&) = delete;
+  ILogger(ILogger&&) = delete;
+  ILogger& operator=(ILogger&&) = delete;
+  virtual ~ILogger() = default;
+
+  /**
+   * @brief Submit a log message to the logging subsystem.
+   *
+   * @param payload User-provided log data (level, phase, message).
+   * @param origin Auto-captured source location of the call site.
+   */
+  void Log(LogDTO payload,
+           std::source_location origin = std::source_location::current(),
+           std::chrono::system_clock::time_point timestamp = std::chrono::system_clock::now(),
+           std::thread::id thread_id = std::this_thread::get_id()) {
+    LogEntry entry;
+    entry.timestamp = timestamp;
+    entry.thread_id = thread_id;
+    entry.level = payload.level;
+    entry.phase = payload.phase;
+    entry.message = std::move(payload.message);
+    entry.origin = origin;
+    DoLog(std::move(entry));
+  }
+
+ protected:
+  /**
+   * @brief Underlying implementation to transport the log entry.
+   *
+   * Implementations must be thread-safe as DoLog can be called concurrently
+   * from multiple worker threads.
+   */
+  virtual void DoLog(LogEntry log_entry) = 0;
+};
+
+#endif  // BIERGARTEN_PIPELINE_INCLUDES_SERVICES_LOGGING_LOGGER_H_

tooling/pipeline/includes/services/prompting/prompt_directory.h

@@ -12,11 +12,14 @@
  */
 
 #include <filesystem>
+#include <memory>
 #include <stdexcept>
 #include <string>
 #include <string_view>
 #include <unordered_map>
 
+#include "services/logging/logger.h"
+
 /**
  * @brief Interface for loading named prompt files.
  */
@@ -56,6 +59,8 @@ class PromptDirectory final : public IPromptDirectory {
    *         directory.
    */
   explicit PromptDirectory(const std::filesystem::path& prompt_dir);
+  PromptDirectory(const std::filesystem::path& prompt_dir,
+                  std::shared_ptr<ILogger> logger);
 
   /**
    * @brief Loads the prompt for @p key, caching the result.
@@ -70,6 +75,7 @@ class PromptDirectory final : public IPromptDirectory {
 
  private:
   std::filesystem::path prompt_dir_;
+  std::shared_ptr<ILogger> logger_;
   std::unordered_map<std::string, std::string> cache_;
 };
 

tooling/pipeline/includes/web_client/http_web_client.h

@@ -8,8 +8,11 @@
 
 
 #include "web_client/web_client.h"
+#include "services/logging/logger.h"
 
+#include <memory>
 #include <string>
+#include <utility>
 
 /**
  * @brief WebClient implementation backed by cpp-httplib.
@@ -24,7 +27,8 @@
  */
 class HttpWebClient final : public WebClient {
 public:
-  HttpWebClient() = default;
+  explicit HttpWebClient(std::shared_ptr<ILogger> logger)
+      : logger_(std::move(logger)) {}
   ~HttpWebClient() override = default;
 
   /**
@@ -42,7 +46,10 @@ public:
    * @param value Raw string to encode.
    * @return Percent-encoded string safe for use in a URL.
    */
-  std::string UrlEncode(const std::string& value) override;
+  std::string EncodeURL(const std::string& value) override;
+
+ private:
+  std::shared_ptr<ILogger> logger_;
 };
 
 

tooling/pipeline/includes/web_client/web_client.h

@@ -30,7 +30,7 @@ class WebClient {
    * @param value Raw string value.
    * @return Encoded value safe for URL usage.
    */
-  virtual std::string UrlEncode(const std::string& value) = 0;
+  virtual std::string EncodeURL(const std::string& value) = 0;
 };
 
 #endif  // BIERGARTEN_PIPELINE_INCLUDES_WEB_CLIENT_WEB_CLIENT_H_

tooling/pipeline/runpod/.dockerignore

@@ -0,0 +1,9 @@
+# Ignore model files!
+*.gguf
+*.bin
+models/
+weights/
+
+# Ignore local build folders
+build/
+.git/

tooling/pipeline/runpod/Dockerfile

@@ -0,0 +1,72 @@
+# --- Stage 1: Build Environment (The "Heavy" Stage) ---
+FROM nvidia/cuda:12.6.3-devel-ubuntu24.04 AS builder
+
+ENV DEBIAN_FRONTEND=noninteractive \
+  CMAKE_GENERATOR=Ninja
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+  build-essential ca-certificates curl git libboost-json-dev \
+  libboost-program-options-dev libssl-dev ninja-build pkg-config zlib1g-dev \
+  && rm -rf /var/lib/apt/lists/*
+
+# Install modern CMake
+RUN curl -L https://github.com/Kitware/CMake/releases/download/v3.31.0/cmake-3.31.0-linux-x86_64.sh -o cmake.sh && \
+  sh cmake.sh --skip-license --prefix=/usr/local && rm cmake.sh
+
+# Get headers for C++ build
+RUN curl -L https://github.com/ggml-org/llama.cpp/archive/refs/tags/b9012.tar.gz -o /tmp/llama-src.tar.gz && \
+  tar -xzf /tmp/llama-src.tar.gz -C /tmp && \
+  cp -r /tmp/llama.cpp-b9012/include/* /usr/local/include/ && \
+  cp -r /tmp/llama.cpp-b9012/ggml/include/* /usr/local/include/
+
+# Pull llama.cpp binaries to use during build if needed
+COPY --from=ghcr.io/ggml-org/llama.cpp:full-cuda /app/lib*.so* /usr/local/lib/
+
+WORKDIR /app
+COPY . .
+
+# Build the C++ pipeline
+RUN cmake -S . -B build -G Ninja -DCMAKE_BUILD_TYPE=Release && \
+  cmake --build build -j$(nproc)
+
+# --- Stage 2: Runtime Environment (The "Slim" Stage) ---
+FROM nvidia/cuda:12.6.3-runtime-ubuntu24.04 AS runtime
+
+# Install only necessary runtime shared libraries
+RUN apt-get update && apt-get install -y --no-install-recommends \
+  curl \
+  ca-certificates \
+  libboost-json1.83.0 \
+  libboost-program-options1.83.0 \
+  libgomp1 \
+  libssl3 \
+  zlib1g \
+  && rm -rf /var/lib/apt/lists/*
+
+ENV APP_ROOT=/app \
+  LD_LIBRARY_PATH="/usr/local/lib:${LD_LIBRARY_PATH}"
+
+WORKDIR /app/build
+
+# Copy only the compiled binaries from the builder
+COPY --from=builder /app/build/biergarten-pipeline ./
+
+# Copy required config files
+COPY locations.json /app/build/
+COPY beer-styles.json /app/build/
+
+# Copy prompt templates
+COPY prompts /app/prompts
+
+# Copy only the necessary shared libraries from builder/llama-bin
+COPY --from=ghcr.io/ggml-org/llama.cpp:full-cuda /app/lib*.so* /usr/local/lib/
+
+# Co-locate plugins
+RUN cp /usr/local/lib/libggml-cuda.so . 2>/dev/null || true && \
+  cp /usr/local/lib/libggml-cpu*.so . 2>/dev/null || true
+
+# Setup Start Script
+COPY ./runpod/start.sh /usr/local/bin/biergarten-start
+RUN chmod +x /usr/local/bin/biergarten-start
+
+ENTRYPOINT ["/usr/local/bin/biergarten-start"]

tooling/pipeline/runpod/README.md

@@ -0,0 +1,8 @@
+```bash
+touch runpod/start.sh
+docker build \
+ --progress=plain \
+ -t biergarten-pipeline:latest \
+ -f runpod/Dockerfile \
+ . 2>&1 | tee build.log
+```

tooling/pipeline/runpod/pod-template.yaml

@@ -0,0 +1,22 @@
+name: biergarten-pipeline-live
+imageName: biergarten-pipeline:latest
+category: NVIDIA
+containerDiskInGb: 50
+volumeInGb: 50
+volumeMountPath: /workspace
+dockerEntrypoint:
+  - /usr/local/bin/biergarten-start
+dockerStartCmd: []
+isPublic: false
+isServerless: false
+env:
+  BIERGARTEN_MODE: live
+  BIERGARTEN_MODEL_PATH: /workspace/models/google_gemma-4-E4B-it-Q6_K.gguf
+  BIERGARTEN_PROMPT_DIR: /workspace/app/build/prompts
+  BIERGARTEN_OUTPUT_DIR: /workspace/output
+  BIERGARTEN_LOG_PATH: /workspace/logs/pipeline.log
+  BIERGARTEN_TEMPERATURE: "1.0"
+  BIERGARTEN_TOP_P: "0.95"
+  BIERGARTEN_TOP_K: "64"
+  BIERGARTEN_N_CTX: "8192"
+  BIERGARTEN_SEED: "-1"

tooling/pipeline/runpod/start.sh

@@ -0,0 +1,58 @@
+#!/bin/bash
+set -e
+
+MODEL_PATH="${BIERGARTEN_MODEL_PATH:-/workspace/models/google_gemma-4-E4B-it-Q6_K.gguf}"
+OUTPUT_DIR="${BIERGARTEN_OUTPUT_DIR:-/workspace/output}"
+LOG_PATH="${BIERGARTEN_LOG_PATH:-/workspace/logs/pipeline.log}"
+EXECUTABLE="/app/build/biergarten-pipeline"
+PROMPT_DIR="/app/prompts"
+
+echo "--- Starting Biergarten Pipeline Environment Check ---"
+
+# Ensure directories exist
+mkdir -p "$OUTPUT_DIR"
+mkdir -p "$(dirname "$LOG_PATH")"
+mkdir -p "$(dirname "$MODEL_PATH")"
+
+# Download model if missing
+if [ ! -f "$MODEL_PATH" ]; then
+    echo "Model not found. Downloading (this may take a while)..."
+
+    curl -L -C - \
+      -o "$MODEL_PATH" \
+      "https://huggingface.co/bartowski/google_gemma-4-E4B-it-GGUF/resolve/main/google_gemma-4-E4B-it-Q6_K.gguf?download=true"
+
+    echo "Download complete."
+fi
+
+# Verify model exists
+if [ ! -f "$MODEL_PATH" ]; then
+    echo "ERROR: Model still not found after download attempt."
+    exit 1
+fi
+
+# Default GPU layers
+GL_LAYERS="${BIERGARTEN_GL_LAYERS:-40}"
+
+# Build args
+ARGS=(
+    "--model"      "$MODEL_PATH"
+    "--prompt-dir" "$PROMPT_DIR"
+    "--output"     "$OUTPUT_DIR"
+    "--log-path"   "$LOG_PATH"
+    "--n-gpu-layers" "$GL_LAYERS"
+)
+
+# Optional params
+[[ -n "$BIERGARTEN_TEMPERATURE" ]] && ARGS+=("--temperature"  "$BIERGARTEN_TEMPERATURE")
+[[ -n "$BIERGARTEN_TOP_P" ]]       && ARGS+=("--top-p"        "$BIERGARTEN_TOP_P")
+[[ -n "$BIERGARTEN_TOP_K" ]]       && ARGS+=("--top-k"        "$BIERGARTEN_TOP_K")
+[[ -n "$BIERGARTEN_N_CTX" ]]       && ARGS+=("--n-ctx"        "$BIERGARTEN_N_CTX")
+[[ -n "$BIERGARTEN_SEED" ]]        && ARGS+=("--seed"         "$BIERGARTEN_SEED")
+
+# Extra args
+[[ -n "$BIERGARTEN_EXTRA_ARGS" ]] && ARGS+=($BIERGARTEN_EXTRA_ARGS)
+
+echo "--- Executing: $EXECUTABLE ${ARGS[*]} ---"
+
+exec "$EXECUTABLE" "${ARGS[@]}"

tooling/pipeline/src/application_options/parse_arguments.cc

@@ -1,12 +1,15 @@
-#include <spdlog/spdlog.h>
+#include <chrono>
-
+#include <format>
+#include <iostream>
 #include <optional>
 #include <sstream>
 #include <string>
 
 #include "data_model/models.h"
+#include "services/logging/logger.h"
 
-std::optional<ApplicationOptions> ParseArguments(const int argc, char** argv) {
+std::optional<ApplicationOptions> ParseArguments(
+    const int argc, char** argv, std::shared_ptr<ILogger> logger) {
   prog_opts::options_description desc("Pipeline Options");
   auto opt = desc.add_options();
 
@@ -30,6 +33,8 @@ std::optional<ApplicationOptions> ParseArguments(const int argc, char** argv) {
         "Context window size in tokens");
     opt("seed", prog_opts::value<int>()->default_value(sampling_defaults.seed),
         "Sampler seed: -1 for random, otherwise non-negative integer");
+    opt("n-gpu-layers", prog_opts::value<int>()->default_value(0),
+        "Number of layers to offload to GPU");
   };
 
   // --mocked and --model are mutually exclusive; validation is enforced below
@@ -50,6 +55,7 @@ std::optional<ApplicationOptions> ParseArguments(const int argc, char** argv) {
     opt("prompt-dir", prog_opts::value<std::string>()->default_value(""),
         "Directory containing named prompt files (e.g. BREWERY_GENERATION.md)."
         " Required when not using --mocked.");
+    opt("location-count", prog_opts::value<uint32_t>()->default_value(10));
   };
 
   add_sampling_options();
@@ -58,10 +64,20 @@ std::optional<ApplicationOptions> ParseArguments(const int argc, char** argv) {
 
   // No flags provided — treat as a help request rather than an error.
   if (argc == 1) {
-    spdlog::info("Biergarten Pipeline");
+    const std::string title = "Biergarten Pipeline";
-    std::stringstream usage_stream;
+    const std::string usage = ([&] {
-    usage_stream << "\nUsage: biergarten-pipeline [options]\n\n" << desc;
+      std::stringstream usage_stream;
-    spdlog::info(usage_stream.str());
+      usage_stream << "\nUsage: biergarten-pipeline [options]\n\n" << desc;
+      return usage_stream.str();
+    })();
+    if (logger) {
+      logger->Log(LogDTO{.level = LogLevel::Info,
+                   .phase = PipelinePhase::Startup,
+                   .message = title});
+      logger->Log(LogDTO{.level = LogLevel::Info,
+                   .phase = PipelinePhase::Startup,
+                   .message = usage});
+    }
     return std::nullopt;
   }
 
@@ -73,7 +89,11 @@ std::optional<ApplicationOptions> ParseArguments(const int argc, char** argv) {
     if (var_map.contains("help")) {
       std::stringstream help_stream;
       help_stream << "\n" << desc;
-      spdlog::info(help_stream.str());
+      if (logger) {
+        logger->Log(LogDTO{.level = LogLevel::Info,
+                     .phase = PipelinePhase::Startup,
+                     .message = help_stream.str()});
+      }
       return std::nullopt;
     }
 
@@ -82,34 +102,57 @@ std::optional<ApplicationOptions> ParseArguments(const int argc, char** argv) {
     options.pipeline.output_path = var_map["output"].as<std::string>();
     options.pipeline.log_path = var_map["log-path"].as<std::string>();
     options.pipeline.prompt_dir = var_map["prompt-dir"].as<std::string>();
+    options.pipeline.location_count = var_map["location-count"].as<uint32_t>();
 
     const bool use_mocked = var_map["mocked"].as<bool>();
     const std::string model_path = var_map["model"].as<std::string>();
+    const int n_gpu_layers = var_map["n-gpu-layers"].as<int>();
 
     // Enforce mutual exclusivity before any further configuration is applied.
     if (use_mocked && !model_path.empty()) {
-      spdlog::error(
+      const std::string msg =
-          "Invalid arguments: --mocked and --model are mutually exclusive");
+          "Invalid arguments: --mocked and --model are mutually exclusive";
+      if (logger) {
+        logger->Log(LogDTO{.level = LogLevel::Error,
+                     .phase = PipelinePhase::Startup,
+                     .message = msg});
+      } else {
+        std::cerr << msg << std::endl;
+      }
       return std::nullopt;
     }
 
     if (!use_mocked && model_path.empty()) {
-      spdlog::error(
+      const std::string msg =
-          "Invalid arguments: either --mocked or --model must be specified");
+          "Invalid arguments: either --mocked or --model must be specified";
+      if (logger) {
+        logger->Log(LogDTO{.level = LogLevel::Error,
+                     .phase = PipelinePhase::Startup,
+                     .message = msg});
+      } else {
+        std::cerr << msg << std::endl;
+      }
       return std::nullopt;
     }
 
     // Prompt directory is only meaningful for live inference — the mock
     // generator has no use for it and should not require it to be present.
     if (!use_mocked && options.pipeline.prompt_dir.empty()) {
-      spdlog::error(
+      const std::string msg =
-          "Invalid arguments: --prompt-dir is required when not using "
+          "Invalid arguments: --prompt-dir is required when not using --mocked";
-          "--mocked");
+      if (logger) {
+        logger->Log({.level = LogLevel::Error,
+                     .phase = PipelinePhase::Startup,
+                     .message = msg});
+      } else {
+        std::cerr << msg << std::endl;
+      }
       return std::nullopt;
     }
 
     options.generator.use_mocked = use_mocked;
     options.generator.model_path = model_path;
+    // options.generator.n_gpu_layers = n_gpu_layers;
 
     // Only populate sampling config when the user explicitly overrides at
     // least one value. Leaving it as std::nullopt lets LlamaGenerator fall
@@ -118,13 +161,21 @@ std::optional<ApplicationOptions> ParseArguments(const int argc, char** argv) {
     const bool user_provided_sampling =
         !var_map["temperature"].defaulted() || !var_map["top-p"].defaulted() ||
         !var_map["top-k"].defaulted() || !var_map["n-ctx"].defaulted() ||
-        !var_map["seed"].defaulted();
+        !var_map["seed"].defaulted() || !var_map["n_gpu_layers"].defaulted();
 
     if (user_provided_sampling) {
       // Warn but do not fail — the run is still valid, the flags are just
       // silently irrelevant when no model is loaded.
       if (use_mocked) {
-        spdlog::warn("Sampling parameters are ignored when using --mocked");
+        const std::string msg =
+            "Sampling parameters are ignored when using --mocked";
+        if (logger) {
+          logger->Log(LogDTO{.level = LogLevel::Warn,
+                       .phase = PipelinePhase::Startup,
+                       .message = msg});
+        } else {
+          std::cerr << msg << std::endl;
+        }
       } else {
         SamplingOptions sampling;
         sampling.temperature = var_map["temperature"].as<float>();
@@ -132,6 +183,7 @@ std::optional<ApplicationOptions> ParseArguments(const int argc, char** argv) {
         sampling.top_k = var_map["top-k"].as<uint32_t>();
         sampling.n_ctx = var_map["n-ctx"].as<uint32_t>();
         sampling.seed = var_map["seed"].as<int>();
+        sampling.n_gpu_layers = var_map["n-gpu-layers"].as<int>();
 
         options.generator.sampling = sampling;
       }
@@ -140,11 +192,23 @@ std::optional<ApplicationOptions> ParseArguments(const int argc, char** argv) {
     return options;
 
   } catch (const std::exception& exception) {
-    spdlog::error("Failed to parse command-line arguments: {}",
+    const std::string msg =
-                  exception.what());
+        std::string("Failed to parse command-line arguments: ") +
+        exception.what();
+    if (logger) {
+      logger->Log(LogDTO{.level = LogLevel::Error,
+                   .phase = PipelinePhase::Startup,
+                   .message = msg});
+    }
     return std::nullopt;
   } catch (...) {
-    spdlog::error("Failed to parse command-line arguments: unknown error");
+    const std::string msg =
+        "Failed to parse command-line arguments: unknown error";
+    if (logger) {
+      logger->Log(LogDTO{.level = LogLevel::Error,
+                   .phase = PipelinePhase::Startup,
+                   .message = msg});
+    }
     return std::nullopt;
   }
 }

tooling/pipeline/src/biergarten_data_generator/biergarten_data_generator.cc

@@ -1,16 +0,0 @@
-/**
- * @file biergarten_data_generator/biergarten_data_generator.cc
- * @brief BiergartenDataGenerator constructor implementation.
- */
-
-#include "biergarten_data_generator.h"
-
-#include <utility>
-
-BiergartenDataGenerator::BiergartenDataGenerator(
-    std::unique_ptr<IEnrichmentService> context_service,
-    std::unique_ptr<DataGenerator> generator,
-    std::unique_ptr<IExportService> exporter)
-    : context_service_(std::move(context_service)),
-      generator_(std::move(generator)),
-      exporter_(std::move(exporter)) {}

tooling/pipeline/src/biergarten_data_generator/generate_breweries.cc

@@ -1,58 +0,0 @@
-/**
- * @file biergarten_data_generator/generate_breweries.cc
- * @brief BiergartenDataGenerator::GenerateBreweries() implementation.
- */
-
-#include <spdlog/spdlog.h>
-
-#include "biergarten_data_generator.h"
-
-void BiergartenDataGenerator::GenerateBreweries(
-    std::span<const EnrichedCity> cities) {
-  spdlog::info("\n=== SAMPLE BREWERY GENERATION ===");
-
-  generated_breweries_.clear();
-  size_t skipped_count = 0;
-  size_t export_failed_count = 0;
-
-  for (const auto& [location, region_context] : cities) {
-    try {
-      const BreweryResult brewery =
-          generator_->GenerateBrewery(location, region_context);
-
-      const GeneratedBrewery gen{.location = location, .brewery = brewery};
-
-      generated_breweries_.push_back(gen);
-
-      try {
-        exporter_->ProcessRecord(gen);
-      } catch (const std::exception& export_exception) {
-        ++export_failed_count;
-
-        spdlog::warn(
-            "[Pipeline] Generated brewery for '{}' ({}) but SQLite export "
-            "failed: {}",
-            location.city, location.country, export_exception.what());
-      }
-    } catch (const std::exception& e) {
-      ++skipped_count;
-
-      spdlog::warn(
-          "[Pipeline] Skipping city '{}' ({}): brewery generation failed: "
-          "{}",
-          location.city, location.country, e.what());
-    }
-  }
-
-  if (skipped_count > 0) {
-    spdlog::warn("[Pipeline] Skipped {} city/cities due to generation errors",
-                 skipped_count);
-  }
-
-  if (export_failed_count > 0) {
-    spdlog::warn(
-        "[Pipeline] Failed to export {} generated brewery/breweries to "
-        "SQLite",
-        export_failed_count);
-  }
-}

tooling/pipeline/src/biergarten_data_generator/log_results.cc

@@ -1,26 +0,0 @@
-/**
- * @file biergarten_data_generator/log_results.cc
- * @brief BiergartenDataGenerator::LogResults() implementation.
- */
-
-#include <spdlog/spdlog.h>
-
-#include "biergarten_data_generator.h"
-
-void BiergartenDataGenerator::LogResults() const {
-  spdlog::info("\n=== GENERATED DATA DUMP ===");
-  size_t index = 1;
-  for (const auto& [location, brewery] : generated_breweries_) {
-    spdlog::info(
-        "{}. city=\"{}\" country=\"{}\" state=\"{}\" "
-        "iso3166_2={} lat={} lon={}",
-        index, location.city, location.country, location.state_province,
-        location.iso3166_2, location.latitude, location.longitude);
-    spdlog::info("   brewery_name_en=\"{}\"", brewery.name_en);
-    spdlog::info("   brewery_description_en=\"{}\"", brewery.description_en);
-    spdlog::info("   brewery_name_local=\"{}\"", brewery.name_local);
-    spdlog::info("   brewery_description_local=\"{}\"",
-                 brewery.description_local);
-    ++index;
-  }
-}

tooling/pipeline/src/biergarten_data_generator/query_cities_with_countries.cc

@@ -1,41 +0,0 @@
-/**
- * @file biergarten_data_generator/query_cities_with_countries.cc
- * @brief BiergartenDataGenerator::QueryCitiesWithCountries() implementation.
- */
-
-#include <spdlog/spdlog.h>
-
-#include <algorithm>
-#include <filesystem>
-#include <iterator>
-#include <random>
-
-#include "biergarten_data_generator.h"
-#include "json_handling/json_loader.h"
-
-static constexpr size_t kBreweryAmount = 50;
-
-std::vector<Location> BiergartenDataGenerator::QueryCitiesWithCountries() {
-  spdlog::info("\n=== GEOGRAPHIC DATA OVERVIEW ===");
-
-  const std::filesystem::path locations_path = "locations.json";
-
-  auto all_locations = JsonLoader::LoadLocations(locations_path);
-  spdlog::info("  Locations available: {}", all_locations.size());
-
-  const size_t sample_count = std::min(kBreweryAmount, all_locations.size());
-
-  const auto sample_count_signed =
-      static_cast<std::iter_difference_t<decltype(all_locations.cbegin())>>(
-          sample_count);
-
-  std::vector<Location> sampled_locations;
-  sampled_locations.reserve(sample_count);
-
-  std::random_device random_generator;
-  std::ranges::sample(all_locations, std::back_inserter(sampled_locations),
-                      sample_count_signed, random_generator);
-
-  spdlog::info("  Sampled locations: {}", sampled_locations.size());
-  return sampled_locations;
-}

tooling/pipeline/src/biergarten_data_generator/run.cc

@@ -1,52 +0,0 @@
-/**
- * @file biergarten_data_generator/run.cc
- * @brief BiergartenDataGenerator::Run() implementation.
- */
-
-#include <spdlog/spdlog.h>
-
-#include <utility>
-
-#include "biergarten_data_generator.h"
-
-bool BiergartenDataGenerator::Run() {
-  try {
-    exporter_->Initialize();
-
-    std::vector<Location> cities = QueryCitiesWithCountries();
-    std::vector<EnrichedCity> enriched;
-    enriched.reserve(cities.size());
-
-    size_t skipped_count = 0;
-    for (auto& city : cities) {
-      try {
-        std::string region_context = context_service_->GetLocationContext(city);
-        spdlog::debug("[Pipeline] Context for '{}' ({}) gathered:\n{}",
-                      city.city, city.country, region_context);
-
-        enriched.push_back(
-            EnrichedCity{.location = std::move(city),
-                         .region_context = std::move(region_context)});
-      } catch (const std::exception& exception) {
-        ++skipped_count;
-        spdlog::warn(
-            "[Pipeline] Skipping city '{}' ({}): context lookup failed: {}",
-            city.city, city.country, exception.what());
-      }
-    }
-
-    if (skipped_count > 0) {
-      spdlog::warn(
-          "[Pipeline] Skipped {} city/cities due to context lookup errors",
-          skipped_count);
-    }
-
-    this->GenerateBreweries(enriched);
-    exporter_->Finalize();
-    this->LogResults();
-    return true;
-  } catch (const std::exception& e) {
-    spdlog::error("Pipeline execution failed with error: {}", e.what());
-    return false;
-  }
-}

tooling/pipeline/src/biergarten_pipeline_orchestrator/biergarten_pipeline_orchestrator.cc

@@ -0,0 +1,20 @@
+/**
+ * @file biergarten_pipeline_orchestrator/biergarten_pipeline_orchestrator.cc
+ * @brief BiergartenDataGenerator constructor implementation.
+ */
+
+#include "biergarten_pipeline_orchestrator.h"
+
+#include <utility>
+
+BiergartenPipelineOrchestrator::BiergartenPipelineOrchestrator(
+    std::shared_ptr<ILogger> logger,
+    std::unique_ptr<IEnrichmentService> context_service,
+    std::unique_ptr<DataGenerator> generator,
+    std::unique_ptr<IExportService> exporter,
+    const ApplicationOptions &app_options)
+    : logger_(std::move(logger)),
+      context_service_(std::move(context_service)),
+      generator_(std::move(generator)),
+      exporter_(std::move(exporter)),
+      application_options_(app_options) {}

tooling/pipeline/src/biergarten_pipeline_orchestrator/generate_breweries.cc

@@ -0,0 +1,68 @@
+/**
+ * @file biergarten_pipeline_orchestrator/generate_breweries.cc
+ * @brief BiergartenDataGenerator::GenerateBreweries() implementation.
+ */
+
+#include <chrono>
+#include <format>
+
+#include "biergarten_pipeline_orchestrator.h"
+#include "services/logging/logger.h"
+
+void BiergartenPipelineOrchestrator::GenerateBreweries(
+    std::span<const EnrichedCity> cities) {
+  logger_->Log({.level = LogLevel::Info,
+                .phase = PipelinePhase::BreweryAndBeerGeneration,
+                .message = "=== SAMPLE BREWERY GENERATION ==="});
+
+  generated_breweries_.clear();
+  size_t skipped_count = 0;
+  size_t export_failed_count = 0;
+
+  for (const auto& [location, region_context] : cities) {
+    try {
+      const BreweryResult brewery =
+          generator_->GenerateBrewery(location, region_context);
+
+      const GeneratedBrewery gen{.location = location, .brewery = brewery};
+
+      generated_breweries_.push_back(gen);
+
+      try {
+        exporter_->ProcessRecord(gen);
+      } catch (const std::exception& export_exception) {
+        ++export_failed_count;
+
+        logger_->Log(
+            {.level = LogLevel::Warn,
+             .phase = PipelinePhase::BreweryAndBeerGeneration,
+             .message =
+                 std::format("[Pipeline] Generated brewery for '{}' ({}) but SQLite export failed: {}",
+                 location.city, location.country, export_exception.what())});
+      }
+    } catch (const std::exception& e) {
+      ++skipped_count;
+
+      logger_->Log({.level = LogLevel::Warn,
+                    .phase = PipelinePhase::BreweryAndBeerGeneration,
+                    .message = std::format("[Pipeline] Skipping city '{}' ({}): brewery generation failed: {}",
+                               location.city, location.country, e.what())});
+    }
+  }
+
+  if (skipped_count > 0) {
+    logger_->Log({.level = LogLevel::Warn,
+                  .phase = PipelinePhase::BreweryAndBeerGeneration,
+                  .message = std::format(
+                      "[Pipeline] Skipped {} city/cities due to generation errors",
+                      skipped_count)});
+  }
+
+  if (export_failed_count > 0) {
+    logger_->Log({.level = LogLevel::Warn,
+                  .phase = PipelinePhase::Teardown,
+                  .message = std::format(
+                      "[Pipeline] Failed to export {} generated brewery/breweries to SQLite",
+                      export_failed_count)});
+  }
+}

tooling/pipeline/src/biergarten_pipeline_orchestrator/log_results.cc

@@ -0,0 +1,37 @@
+/**
+ * @file biergarten_pipeline_orchestrator/log_results.cc
+ * @brief BiergartenDataGenerator::LogResults() implementation.
+ */
+
+#include <boost/json/array.hpp>
+#include <chrono>
+#include <format>
+
+#include "../../includes/json_handling/pretty_print.h"
+#include "biergarten_pipeline_orchestrator.h"
+#include "services/logging/logger.h"
+void BiergartenPipelineOrchestrator::LogResults() const {
+  boost::json::array output;
+
+  for (const auto& [location, brewery] : generated_breweries_) {
+    output.push_back(boost::json::object{
+        {"name_en", brewery.name_en},
+        {"description_en", brewery.description_en},
+        {"name_local", brewery.name_local},
+        {"description_local", brewery.description_local},
+        {"location", boost::json::object{
+                         {"city", location.city},
+                         {"country", location.country},
+                         {"state_province", location.state_province},
+                         {"iso3166_2", location.iso3166_2},
+                         {"latitude", location.latitude},
+                         {"longitude", location.longitude},
+                     }}});
+  }
+
+  std::ostringstream oss;
+  PrettyPrint(oss, output);
+  logger_->Log({.level = LogLevel::Info,
+                .phase = PipelinePhase::Teardown,
+                .message = oss.str()});
+}

tooling/pipeline/src/biergarten_pipeline_orchestrator/query_cities_with_countries.cc

@@ -0,0 +1,51 @@
+/**
+ * @file biergarten_pipeline_orchestrator/query_cities_with_countries.cc
+ * @brief BiergartenDataGenerator::QueryCitiesWithCountries() implementation.
+ */
+
+#include <algorithm>
+#include <chrono>
+#include <filesystem>
+#include <format>
+#include <iterator>
+#include <random>
+
+#include "biergarten_pipeline_orchestrator.h"
+#include "json_handling/json_loader.h"
+#include "services/logging/logger.h"
+
+std::vector<Location>
+BiergartenPipelineOrchestrator::QueryCitiesWithCountries() {
+  logger_->Log({.level = LogLevel::Info,
+                .phase = PipelinePhase::Startup,
+                .message = "=== GEOGRAPHIC DATA OVERVIEW ==="});
+
+  const std::filesystem::path locations_path = "locations.json";
+
+  auto all_locations = JsonLoader::LoadLocations(locations_path, logger_);
+
+  const size_t sample_count = std::min(
+      static_cast<size_t>(application_options_.pipeline.location_count),
+      all_locations.size());
+
+  const auto sample_count_signed =
+      static_cast<std::iter_difference_t<decltype(all_locations.cbegin())>>(
+          sample_count);
+
+  std::vector<Location> sampled_locations;
+  sampled_locations.reserve(sample_count);
+
+  std::random_device random_generator;
+  std::ranges::sample(all_locations, std::back_inserter(sampled_locations),
+                      sample_count_signed, random_generator);
+
+  logger_->Log({.level = LogLevel::Info,
+                .phase = PipelinePhase::Startup,
+                .message = std::format("  Locations available: {}",
+                                       all_locations.size())});
+  logger_->Log({.level = LogLevel::Info,
+                .phase = PipelinePhase::Startup,
+                .message = std::format("  Sampled locations: {}",
+                                       sampled_locations.size())});
+  return sampled_locations;
+}

tooling/pipeline/src/biergarten_pipeline_orchestrator/run.cc

@@ -0,0 +1,63 @@
+/**
+ * @file biergarten_pipeline_orchestrator/run.cc
+ * @brief BiergartenDataGenerator::Run() implementation.
+ */
+
+#include <chrono>
+#include <format>
+#include <utility>
+
+#include "biergarten_pipeline_orchestrator.h"
+#include "services/logging/logger.h"
+
+bool BiergartenPipelineOrchestrator::Run() {
+  try {
+    exporter_->Initialize();
+
+    std::vector<Location> cities = QueryCitiesWithCountries();
+    std::vector<EnrichedCity> enriched;
+    enriched.reserve(cities.size());
+
+    size_t skipped_count = 0;
+    for (auto& city : cities) {
+      try {
+        std::string region_context = context_service_->GetLocationContext(city);
+        // logger_->Log(LogLevel::Debug, PipelinePhase::UserGeneration,
+        //              "[Pipeline] Context for '" + city.city + "' (" +
+        //              city.iso3166_2 + ") gathered:\n" + region_context);
+
+        enriched.push_back(
+            EnrichedCity{.location = std::move(city),
+                         .region_context = std::move(region_context)});
+      } catch (const std::exception& exception) {
+        ++skipped_count;
+        logger_->Log(
+            {.level = LogLevel::Warn,
+             .phase = PipelinePhase::UserGeneration,
+             .message = std::format(
+                 "[Pipeline] Skipping city '{}' ({}): context lookup failed: {}",
+                 city.city, city.country, exception.what())});
+      }
+    }
+
+    if (skipped_count > 0) {
+      logger_->Log({.level = LogLevel::Warn,
+                    .phase = PipelinePhase::UserGeneration,
+                    .message = std::format(
+                        "[Pipeline] Skipped {} city/cities due to context lookup errors",
+                        skipped_count)});
+    }
+
+    this->GenerateBreweries(enriched);
+    exporter_->Finalize();
+    this->LogResults();
+    return true;
+  } catch (const std::exception& e) {
+    logger_->Log(
+        {.level = LogLevel::Error,
+         .phase = PipelinePhase::Teardown,
+         .message =
+             std::format("Pipeline execution failed with error: {}", e.what())});
+    return false;
+  }
+}

tooling/pipeline/src/data_generation/llama/generate_brewery.cc

@@ -4,8 +4,7 @@
  * inference, and validates structured JSON output for brewery records.
  */
 
-#include <spdlog/spdlog.h>
+#include <chrono>
-
 #include <format>
 #include <optional>
 #include <stdexcept>
@@ -100,8 +99,13 @@ BreweryResult LlamaGenerator::GenerateBrewery(
     // Generate brewery data from LLM
     raw = this->Infer(system_prompt, user_prompt, max_tokens,
                       kBreweryJsonGrammar);
-    spdlog::debug("LlamaGenerator: raw output (attempt {}): {}", attempt + 1,
+    if (logger_) {
-                  raw);
+      logger_->Log(
+          {.level = LogLevel::Debug,
+           .phase = PipelinePhase::BreweryAndBeerGeneration,
+           .message = std::format("LlamaGenerator: raw output (attempt {}): {}",
+                      attempt + 1, raw)});
+    }
 
     // Validate output: parse JSON and check required fields
 
@@ -112,9 +116,13 @@ BreweryResult LlamaGenerator::GenerateBrewery(
     if (!validation_error.has_value()) {
       // Success: return parsed brewery data
 
-      spdlog::info(
+      if (logger_) {
-          "LlamaGenerator: successfully generated brewery data on attempt {}",
+        logger_->Log(
-          attempt + 1);
+            {.level = LogLevel::Info,
+             .phase = PipelinePhase::BreweryAndBeerGeneration,
+             .message = std::format("LlamaGenerator: successfully generated brewery data on attempt {}",
+                        attempt + 1)});
+      }
 
       return brewery;
     }
@@ -122,8 +130,14 @@ BreweryResult LlamaGenerator::GenerateBrewery(
     // Validation failed: log error and prepare corrective feedback
 
     last_error = *validation_error;
-    spdlog::warn("LlamaGenerator: malformed brewery JSON (attempt {}): {}",
+    if (logger_) {
-                 attempt + 1, *validation_error);
+      logger_->Log(
+          {.level = LogLevel::Warn,
+           .phase = PipelinePhase::BreweryAndBeerGeneration,
+           .message =
+               std::format("LlamaGenerator: malformed brewery JSON (attempt {}): {}",
+               attempt + 1, *validation_error)});
+    }
 
     // Update prompt with error details to guide LLM toward correct output.
     user_prompt = std::format(
@@ -140,9 +154,13 @@ BreweryResult LlamaGenerator::GenerateBrewery(
   }
 
   // All retry attempts exhausted: log failure and throw exception
-  spdlog::error(
+  if (logger_) {
-      "LlamaGenerator: malformed brewery response after {} attempts: "
+    logger_->Log(
-      "{}",
+        {.level = LogLevel::Error,
-      max_attempts, last_error.empty() ? raw : last_error);
+         .phase = PipelinePhase::BreweryAndBeerGeneration,
+       .message = std::format(
+         "LlamaGenerator: malformed brewery response after {} attempts: {}",
+         max_attempts, last_error.empty() ? raw : last_error)});
+  }
   throw std::runtime_error("LlamaGenerator: malformed brewery response");
 }

tooling/pipeline/src/data_generation/llama/generate_user.cc

@@ -4,9 +4,8 @@
  * retry handling, and output sanitization for downstream parsing.
  */
 
-#include <spdlog/spdlog.h>
 
-#include <stdexcept>
+#include <format>
 #include <string>
 
 #include "data_generation/llama_generator.h"
@@ -21,5 +20,5 @@
 // 4. Return locale-aware username and biography
 UserResult LlamaGenerator::GenerateUser(const std::string& locale) {
   return {.username = "test_user",
-          .bio = "This is a test user profile from " + locale + "."};
+          .bio = std::format("This is a test user profile from {}.", locale)};
 }

tooling/pipeline/src/data_generation/llama/helpers.cc

@@ -16,11 +16,11 @@
 
 #include "data_generation/llama_generator_helpers.h"
 #include "llama.h"
-
+namespace {
 /**
  * String trimming: removes leading and trailing whitespace
  */
-static std::string Trim(std::string_view value) {
+std::string Trim(std::string_view value) {
   constexpr std::string_view whitespace = " \t\n\r\f\v";
   const size_t first_index = value.find_first_not_of(whitespace);
   if (first_index == std::string_view::npos) {
@@ -35,7 +35,7 @@ static std::string Trim(std::string_view value) {
  * Normalize whitespace: collapses multiple spaces/tabs/newlines into single
  * spaces
  */
-static std::string CondenseWhitespace(std::string_view text) {
+std::string CondenseWhitespace(std::string_view text) {
   std::string out;
   out.reserve(text.size());
 
@@ -61,7 +61,37 @@ static std::string CondenseWhitespace(std::string_view text) {
 // Guard against truncating in the first half of the string.
 // This preserves the critical opening content and avoids cutting critical
 // context words early in the region description.
-static constexpr size_t kTruncationGuardDivisor = 2;
+constexpr size_t kTruncationGuardDivisor = 2;
+
+bool ReadRequiredTrimmedStringField(const boost::json::object& obj,
+                                    std::string_view key, std::string& out,
+                                    std::string* error_out) {
+  const boost::json::value* field = obj.if_contains(key);
+  if (field == nullptr || !field->is_string()) {
+    return false;
+  }
+
+  const auto& string_value = field->as_string();
+  out = Trim(std::string_view(string_value.data(), string_value.size()));
+  return !out.empty();
+}
+
+bool HasSchemaPlaceholder(const std::array<std::string*, 4>& values) {
+  for (const std::string* value : values) {
+    std::string lowered = *value;
+    std::ranges::transform(lowered, lowered.begin(),
+                           [](const unsigned char character) {
+                             return static_cast<char>(std::tolower(character));
+                           });
+
+    if (lowered == "string") {
+      return true;
+    }
+  }
+
+  return false;
+}
+}  // namespace
 
 /**
  * Truncate region context to fit within max length while preserving word
@@ -121,47 +151,6 @@ void AppendTokenPiece(const llama_vocab* vocab, llama_token token,
       "LlamaGenerator: failed to decode sampled token piece");
 }
 
-static bool ReadRequiredTrimmedStringField(const boost::json::object& obj,
-                                           std::string_view key,
-                                           std::string& out,
-                                           std::string* error_out) {
-  const boost::json::value* field = obj.if_contains(key);
-  if (field == nullptr || !field->is_string()) {
-    if (error_out != nullptr) {
-      *error_out =
-          "JSON field '" + std::string(key) + "' is missing or not a string";
-    }
-    return false;
-  }
-
-  const auto& string_value = field->as_string();
-  out = Trim(std::string_view(string_value.data(), string_value.size()));
-  if (out.empty()) {
-    if (error_out != nullptr) {
-      *error_out = "JSON field '" + std::string(key) + "' must not be empty";
-    }
-    return false;
-  }
-
-  return true;
-}
-
-static bool HasSchemaPlaceholder(const std::array<std::string*, 4>& values) {
-  for (const std::string* value : values) {
-    std::string lowered = *value;
-    std::ranges::transform(lowered, lowered.begin(),
-                           [](unsigned char character) {
-                             return static_cast<char>(std::tolower(character));
-                           });
-
-    if (lowered == "string") {
-      return true;
-    }
-  }
-
-  return false;
-}
-
 std::optional<std::string> ValidateBreweryJson(const std::string& raw,
                                                BreweryResult& brewery_out) {
   boost::system::error_code error_code;
@@ -209,7 +198,7 @@ std::optional<std::string> ValidateBreweryJson(const std::string& raw,
     return validation_error;
   }
 
-  const std::array<std::string*, 4> schema_placeholders = {
+  const std::array schema_placeholders = {
       &brewery_out.name_en, &brewery_out.description_en,
       &brewery_out.name_local, &brewery_out.description_local};
   if (HasSchemaPlaceholder(schema_placeholders)) {

tooling/pipeline/src/data_generation/llama/infer.cc

@@ -5,9 +5,9 @@
  * output tokens back to text for system+user chat prompts.
  */
 
-#include <spdlog/spdlog.h>
-
 #include <algorithm>
+#include <chrono>
+#include <format>
 #include <memory>
 #include <stdexcept>
 #include <string>
@@ -107,7 +107,7 @@ std::string LlamaGenerator::InferFormatted(const std::string& formatted_prompt,
       .top_p = sampling_top_p_,
       .seed = static_cast<uint32_t>(rng_()),
   };
-  auto sampler = MakeSamplerChain(vocab, sampler_config, grammar);
+  const auto sampler = MakeSamplerChain(vocab, sampler_config, grammar);
 
   /**
    * Clear KV cache to ensure clean inference state (no residual context)
@@ -171,10 +171,14 @@ std::string LlamaGenerator::InferFormatted(const std::string& formatted_prompt,
    */
   prompt_tokens.resize(static_cast<size_t>(token_count));
   if (token_count > prompt_budget) {
-    spdlog::warn(
+    if (logger_) {
-        "LlamaGenerator: prompt too long ({} tokens), truncating to {} "
+      logger_->Log({.level = LogLevel::Warn,
-        "tokens to fit n_batch/n_ctx limits",
+                    .phase = PipelinePhase::BreweryAndBeerGeneration,
-        token_count, prompt_budget);
+                    .message = std::format(
+                        "LlamaGenerator: prompt too long ({} tokens), "
+                        "truncating to {} tokens to fit n_batch/n_ctx limits",
+                        token_count, prompt_budget)});
+    }
     prompt_tokens.resize(static_cast<size_t>(prompt_budget));
     token_count = prompt_budget;
   }

tooling/pipeline/src/data_generation/llama/llama_generator.cc

@@ -32,9 +32,11 @@ void LlamaGenerator::ContextDeleter::operator()(
 
 LlamaGenerator::LlamaGenerator(
     const ApplicationOptions& options, const std::string& model_path,
+    std::shared_ptr<ILogger> logger,
     std::unique_ptr<IPromptFormatter> prompt_formatter,
     std::unique_ptr<IPromptDirectory> prompt_directory)
     : rng_(std::random_device{}()),
+      logger_(std::move(logger)),
       prompt_formatter_(std::move(prompt_formatter)),
       prompt_directory_(std::move(prompt_directory)) {
   if (model_path.empty()) {
@@ -89,6 +91,7 @@ LlamaGenerator::LlamaGenerator(
   }
 
   n_ctx_ = sampling.n_ctx;
+  n_gpu_layers_ = sampling.n_gpu_layers;
 
   this->Load(model_path);
 }

tooling/pipeline/src/data_generation/llama/load.cc

@@ -4,14 +4,14 @@
  * context, and resets prior resources during model initialization.
  */
 
-#include <spdlog/spdlog.h>
-
 #include <algorithm>
+#include <chrono>
 #include <stdexcept>
 #include <string>
 #include <utility>
 
 #include "data_generation/llama_generator.h"
+#include "ggml-backend.h"
 #include "llama.h"
 
 // Maximum batch size for decode operations. Capping the batch prevents
@@ -22,9 +22,16 @@ void LlamaGenerator::Load(const std::string& model_path) {
   context_.reset();
   model_.reset();
 
-  const llama_model_params model_params = llama_model_default_params();
+  // Specifically load dynamic ggml backends (like CUDA) that are provided
-  LlamaGenerator::ModelHandle loaded_model(
+  // externally before attempting to load a model.
+  ggml_backend_load_all();
+
+  llama_model_params model_params = llama_model_default_params();
+  model_params.n_gpu_layers = n_gpu_layers_;
+
+  ModelHandle loaded_model(
       llama_model_load_from_file(model_path.c_str(), model_params));
+
   if (!loaded_model) {
     throw std::runtime_error(
         "LlamaGenerator: failed to load model from path: " + model_path);
@@ -34,8 +41,9 @@ void LlamaGenerator::Load(const std::string& model_path) {
   context_params.n_ctx = n_ctx_;
   context_params.n_batch = std::min(n_ctx_, kMaxBatchSize);
 
-  LlamaGenerator::ContextHandle loaded_context(
+  ContextHandle loaded_context(
       llama_init_from_model(loaded_model.get(), context_params));
+
   if (!loaded_context) {
     throw std::runtime_error("LlamaGenerator: failed to create context");
   }
@@ -43,5 +51,10 @@ void LlamaGenerator::Load(const std::string& model_path) {
   model_ = std::move(loaded_model);
   context_ = std::move(loaded_context);
 
-  spdlog::info("[LlamaGenerator] Loaded model: {}", model_path);
+  if (logger_) {
+    logger_->Log({.level = LogLevel::Info,
+                  .phase = PipelinePhase::Startup,
+                  .message = std::format("[LlamaGenerator] Loaded model: {} ",
+                                         model_path)});
+  }
 }

tooling/pipeline/src/json_handling/json_loader.cc

@@ -6,7 +6,9 @@
 
 #include "json_handling/json_loader.h"
 
-#include <spdlog/spdlog.h>
+#include <format>
+#include "services/logging/logger.h"
+#include <iostream>
 
 #include <boost/json.hpp>
 #include <fstream>
@@ -19,8 +21,8 @@ static std::string ReadRequiredString(const boost::json::object& object,
                                       const char* key) {
   const boost::json::value* value = object.if_contains(key);
   if (value == nullptr || !value->is_string()) {
-    throw std::runtime_error(std::string("Missing or invalid string field: ") +
+    throw std::runtime_error(
-                             key);
+        std::format("Missing or invalid string field: {}", key));
   }
   const std::string_view text = value->as_string();
   return std::string(text);
@@ -30,8 +32,8 @@ static double ReadRequiredNumber(const boost::json::object& object,
                                  const char* key) {
   const boost::json::value* value = object.if_contains(key);
   if (value == nullptr || !value->is_number()) {
-    throw std::runtime_error(std::string("Missing or invalid numeric field: ") +
+    throw std::runtime_error(
-                             key);
+        std::format("Missing or invalid numeric field: {}", key));
   }
   return value->to_number<double>();
 }
@@ -41,7 +43,7 @@ static std::vector<std::string> ReadRequiredStringArray(
   const boost::json::value* value = object.if_contains(key);
   if (value == nullptr || !value->is_array()) {
     throw std::runtime_error(
-        std::string("Missing or invalid string array field: ") + key);
+      std::format("Missing or invalid string array field: {}", key));
   }
 
   const auto& array = value->as_array();
@@ -50,7 +52,7 @@ static std::vector<std::string> ReadRequiredStringArray(
   for (const auto& item : array) {
     if (!item.is_string()) {
       throw std::runtime_error(
-          std::string("Missing or invalid string array field: ") + key);
+          std::format("Missing or invalid string array field: {}", key));
     }
     items.emplace_back(item.as_string());
   }
@@ -58,7 +60,7 @@ static std::vector<std::string> ReadRequiredStringArray(
 }
 
 std::vector<Location> JsonLoader::LoadLocations(
-    const std::filesystem::path& filepath) {
+    const std::filesystem::path& filepath, std::shared_ptr<ILogger> logger) {
   std::ifstream input(filepath);
   if (!input.is_open()) {
     throw std::runtime_error("Failed to open locations file: " +
@@ -104,7 +106,5 @@ std::vector<Location> JsonLoader::LoadLocations(
     });
   }
 
-  spdlog::info("[JsonLoader] Loaded {} locations from {}", locations.size(),
-               filepath.string());
   return locations;
 }

tooling/pipeline/src/main.cc

@@ -4,48 +4,77 @@
  * initializes shared infrastructure, and executes the pipeline entry flow.
  */
 
+#include <spdlog/fmt/fmt.h>
 #include <spdlog/spdlog.h>
 
 #include <boost/di.hpp>
 #include <boost/program_options.hpp>
-
+#include <chrono>
 #include <exception>
+#include <format>
+#include <iostream>
 #include <memory>
 #include <optional>
-
 #include <string>
+#include <thread>
 
-#include "biergarten_data_generator.h"
+#include "biergarten_pipeline_orchestrator.h"
+#include "concurrency/bounded_channel.h"
 #include "data_generation/llama_generator.h"
 #include "data_generation/mock_generator.h"
 #include "data_generation/prompt_formatting/gemma4_jinja_prompt_formatter.h"
 #include "data_model/models.h"
 #include "llama_backend_state.h"
-#include "services/enrichment/enrichment_service.h"
 #include "services/database/export_service.h"
-#include "services/prompting/prompt_directory.h"
 #include "services/database/sqlite_export_service.h"
 #include "services/datetime/timer.h"
+#include "services/enrichment/enrichment_service.h"
+#include "services/enrichment/mock_enrichment.h"
 #include "services/enrichment/wikipedia_service.h"
+#include "services/logging/log_dispatcher.h"
+#include "services/logging/log_entry.h"
+#include "services/logging/log_producer.h"
+#include "services/logging/logger.h"
+#include "services/prompting/prompt_directory.h"
 #include "web_client/http_web_client.h"
 
 namespace di = boost::di;
 
+static constexpr size_t kLogMaxCount = 512;
+
 int main(const int argc, char** argv) {
+  spdlog::set_level(spdlog::level::debug);
+  spdlog::set_pattern("│ %Y-%m-%d %H:%M:%S.%e │ %^%-7l%$ │ %v");
+  BoundedChannel<LogEntry> log_channel(kLogMaxCount);
+
+  auto log_dispatcher =  //
+      std::make_unique<LogDispatcher>(log_channel);
+  std::shared_ptr<ILogger> log_producer =
+      std::make_shared<LogProducer>(log_channel);
+
+  std::thread log_thread([&log_dispatcher] { log_dispatcher->Run(); });
+  auto shutdown = [&](const int exit_code) {
+    log_channel.Close();
+    log_thread.join();
+    return exit_code;
+  };
+
   try {
     Timer timer;
-    spdlog::set_pattern("[%Y-%m-%d %H:%M:%S.%e] [%^%l%$] %v");
 
 #ifndef BIERGARTEN_MOCK_ONLY
     const LlamaBackendState llama_backend_state;
 #endif
-#ifdef DEBUG
-    spdlog::set_level(spdlog::level::debug);
-#endif
 
-    const auto parsed_options = ParseArguments(argc, argv);
+    log_producer->Log({.level   = LogLevel::Info,
+                       .phase   = PipelinePhase::Startup,
+                       .message = "STARTING PIPELINE"});
+
+    const std::optional<ApplicationOptions> parsed_options =
+        ParseArguments(argc, argv, log_producer);
+
     if (!parsed_options.has_value()) {
-      return 0;
+      return shutdown(EXIT_FAILURE);
     }
 
     const auto options = *parsed_options;
@@ -54,55 +83,136 @@ int main(const int argc, char** argv) {
         options.generator.sampling.value_or(SamplingOptions{});
 
     std::unique_ptr<IPromptDirectory> prompt_directory;
+
     if (!options.generator.use_mocked) {
       try {
-        prompt_directory =
+        prompt_directory = std::make_unique<PromptDirectory>(
-            std::make_unique<PromptDirectory>(options.pipeline.prompt_dir);
+            options.pipeline.prompt_dir, log_producer);
       } catch (const std::exception& dir_error) {
-        spdlog::error("[Startup] Invalid --prompt-dir: {}", dir_error.what());
+        log_producer->Log({.level   = LogLevel::Error,
-        return 1;
+                           .phase   = PipelinePhase::Startup,
+                           .message = std::format("Invalid --prompt-dir: {}",
+                                                  dir_error.what())});
+
+        return shutdown(EXIT_FAILURE);
       }
     }
 
     const auto injector = di::make_injector(
-        di::bind<WebClient>().to<HttpWebClient>(),
+        di::bind<ILogger>().to(log_producer),
         di::bind<ApplicationOptions>().to(options),
-        di::bind<IEnrichmentService>().to<WikipediaService>(),
-        di::bind<IExportService>().to<SqliteExportService>(),
-        di::bind<IPromptFormatter>().to<Gemma4JinjaPromptFormatter>(),
         di::bind<std::string>().to(model_path),
-        di::bind<DataGenerator>().to(
+        di::bind<IExportService>().to<SqliteExportService>(),
-            [options, model_path, sampling, &prompt_directory](
+        di::bind<IPromptFormatter>().to([options, log_producer] {
-            const auto& inj) -> std::unique_ptr<DataGenerator> {
+          if (options.generator.use_mocked) {
+            {
+              log_producer->Log(
+                  {.level = LogLevel::Info,
+                   .phase   = PipelinePhase::Startup,
+                   .message = "Prompt formatter: none (mock mode)"});
+            }
+            return std::unique_ptr<IPromptFormatter>(nullptr);
+          }
+          {
+            log_producer->Log(
+                {.level = LogLevel::Info,
+                 .phase   = PipelinePhase::Startup,
+                 .message = "Prompt formatter: Gemma4JinjaPromptFormatter"});
+          }
+          return std::unique_ptr<IPromptFormatter>(
+              std::make_unique<Gemma4JinjaPromptFormatter>());
+        }),
+        di::bind<WebClient>().to([options, log_producer] {
+          if (options.generator.use_mocked) {
+            {
+              log_producer->Log({.level   = LogLevel::Info,
+                                 .phase   = PipelinePhase::Startup,
+                                 .message = "Web client: none (mock mode)"});
+            }
+            return std::unique_ptr<WebClient>(nullptr);
+          }
+          {
+            log_producer->Log({.level   = LogLevel::Info,
+                               .phase   = PipelinePhase::Startup,
+                               .message = "Web client: HttpWebClient"});
+          }
+          return std::unique_ptr<WebClient>(
+              std::make_unique<HttpWebClient>(log_producer));
+        }),
+        di::bind<IEnrichmentService>().to(
+            [options, &log_producer](
+                const auto& inj) -> std::unique_ptr<IEnrichmentService> {
               if (options.generator.use_mocked) {
-                spdlog::info(
+                {
-                    "[Generator] Using MockGenerator (no model path provided)");
+                  log_producer->Log({.level   = LogLevel::Info,
+                                     .phase   = PipelinePhase::Startup,
+                                     .message = "Enrichment: mock"});
+                }
+                return std::make_unique<MockEnrichmentService>();
+              }
+              {
+                log_producer->Log({.level   = LogLevel::Info,
+                                   .phase   = PipelinePhase::Startup,
+                                   .message = "Enrichment: Wikipedia"});
+              }
+              return std::make_unique<WikipediaEnrichmentService>(
+                  inj.template create<std::unique_ptr<WebClient>>(),
+                  log_producer);
+            }),
+        di::bind<DataGenerator>().to(
+            [&options, &model_path, &sampling, &prompt_directory,
+             &log_producer](const auto& inj) -> std::unique_ptr<DataGenerator> {
+              if (options.generator.use_mocked) {
+                {
+                  log_producer->Log({.level   = LogLevel::Info,
+                                     .phase   = PipelinePhase::Startup,
+                                     .message = "Generator: mock"});
+                }
                 return std::make_unique<MockGenerator>();
               }
-
+              {
-              spdlog::info(
+                log_producer->Log(
-                  "[Generator] Using LlamaGenerator: {} (temperature={}, "
+                    {.level = LogLevel::Info,
-                  "top-p={}, top-k={}, n_ctx={}, seed={})",
+                     .phase   = PipelinePhase::Startup,
-                  model_path, sampling.temperature, sampling.top_p,
+                     .message = std::format(
-                  sampling.top_k, sampling.n_ctx, sampling.seed);
+                         "Generator: LlamaGenerator | model={} | temp={:.2f} "
+                         "top_p={:.2f} top_k={} n_ctx={} seed={}",
+                         model_path, sampling.temperature, sampling.top_p,
+                         sampling.top_k, sampling.n_ctx, sampling.seed)});
+              }
               return std::make_unique<LlamaGenerator>(
-                  options, model_path,
+                  options, model_path, log_producer,
                   inj.template create<std::unique_ptr<IPromptFormatter>>(),
                   std::move(prompt_directory));
             }));
 
-    auto generator =
+    const auto orchestrator =
-        injector.create<std::unique_ptr<BiergartenDataGenerator>>();
+        injector.create<std::unique_ptr<BiergartenPipelineOrchestrator>>();
 
-    if (!generator->Run()) {
+    if (!orchestrator->Run()) {
-      spdlog::error("Pipeline execution failed");
+      log_producer->Log({.level   = LogLevel::Error,
-      return 1;
+                         .phase   = PipelinePhase::Teardown,
+                         .message = "Pipeline execution failed"});
+      return shutdown(EXIT_FAILURE);
     }
 
-    spdlog::info("Pipeline executed successfully in {} ms", timer.Elapsed());
+    log_producer->Log({.level   = LogLevel::Info,
-    return 0;
+                       .phase   = PipelinePhase::Teardown,
+                       .message = std::format("Pipeline complete in {} ms",
+                                              timer.Elapsed())});
+
+    return shutdown(EXIT_SUCCESS);
+
   } catch (const std::exception& exception) {
-    spdlog::critical("Unhandled fatal error in main: {}", exception.what());
+    const LogDTO log_entry{.level = LogLevel::Error,
-    return 1;
+                               .phase   = PipelinePhase::Teardown,
+                               .message = exception.what()};
+    if (log_producer) {
+      log_producer->Log(log_entry);
+    } else {
+      std::cerr << log_entry.message << std::endl;
+    }
+
+    return shutdown(EXIT_FAILURE);
   }
 }

tooling/pipeline/src/services/enrichment/wikipedia/fetch_extract.cc

@@ -0,0 +1,160 @@
+/**
+ * @file wikipedia/fetch_extract.cc
+ */
+
+#include <boost/json.hpp>
+#include <chrono>
+#include <format>
+#include <string>
+#include <string_view>
+#include <thread>
+
+#include "services/enrichment/wikipedia_service.h"
+
+using namespace boost;
+
+std::string WikipediaEnrichmentService::FetchExtract(std::string_view query) {
+  const std::string cache_key(query);
+
+  // 1. Cache Lookup
+  if (const auto cache_it = this->extract_cache_.find(cache_key);
+      cache_it != this->extract_cache_.end()) {
+    if (logger_) {
+      logger_->Log({.level = LogLevel::Debug,
+                    .phase = PipelinePhase::UserGeneration,
+                    .message = std::format("Wikipedia: Cache hit for {}!", cache_key)});
+    }
+    return cache_it->second;
+  }
+
+  const std::string encoded = this->client_->EncodeURL(cache_key);
+  const std::string url = std::format(
+      "https://en.wikipedia.org/w/"
+      "api.php?action=query&titles={}&prop=extracts&explaintext=1&format=json",
+      encoded);
+
+  const std::string body = this->client_->Get(url);
+  {
+    using namespace std::literals::chrono_literals;
+    std::this_thread::sleep_for(1s);
+  }
+
+  // 2. Parse JSON
+  system::error_code ec;
+  json::value doc = json::parse(body, ec);
+
+  if (ec) {
+    if (logger_) {
+      logger_->Log(
+          {.level = LogLevel::Warn,
+           .phase = PipelinePhase::UserGeneration,
+           .message = std::format("WikipediaService: JSON parse error for '{}': {}",
+                      std::string(query), ec.message())});
+    }
+    return {};
+  }
+
+  // 3. Safe Extraction
+  const json::object* obj = doc.if_object();
+  if (obj == nullptr) {
+    if (logger_) {
+      logger_->Log(
+          {.level = LogLevel::Warn,
+           .phase = PipelinePhase::UserGeneration,
+           .message =
+               std::format("WikipediaService: Expected root object for '{}'",
+               std::string(query))});
+    }
+    return {};
+  }
+
+  const json::value* query_ptr = obj->if_contains("query");
+  const json::value* pages_ptr =
+      ((query_ptr != nullptr) && query_ptr->is_object())
+          ? query_ptr->get_object().if_contains("pages")
+          : nullptr;
+
+  if ((pages_ptr == nullptr) || !pages_ptr->is_object()) {
+    if (logger_) {
+      logger_->Log(
+          {.level = LogLevel::Warn,
+           .phase = PipelinePhase::UserGeneration,
+           .message =
+               std::format("WikipediaService: Missing query.pages for '{}'",
+               std::string(query))});
+    }
+    return {};
+  }
+
+  const json::object& pages = pages_ptr->get_object();
+
+  if (pages.empty()) {
+    if (logger_) {
+      logger_->Log(
+          {.level = LogLevel::Warn,
+           .phase = PipelinePhase::UserGeneration,
+           .message = std::format("WikipediaService: No pages returned for '{}'",
+                      std::string(query))});
+    }
+    this->extract_cache_.emplace(cache_key, "");
+    return {};
+  }
+
+  // Wikipedia returns the page under a dynamic ID key; we just want the first
+  // one
+  const json::value& page_val = pages.begin()->value();
+
+  if (!page_val.is_object()) {
+    if (logger_) {
+      logger_->Log(
+          {.level = LogLevel::Warn,
+           .phase = PipelinePhase::UserGeneration,
+           .message =
+               std::format("WikipediaService: Unexpected page format for '{}'",
+               std::string(query))});
+    }
+    return {};
+  }
+
+  const json::object& page = page_val.get_object();
+
+  // Handle 404/Missing status
+  if (page.contains("missing")) {
+    if (logger_) {
+      logger_->Log({.level = LogLevel::Warn,
+                    .phase = PipelinePhase::UserGeneration,
+                    .message = std::format("WikipediaService: Page '{}' does not exist",
+                               std::string(query))});
+    }
+    this->extract_cache_.emplace(cache_key, "");
+    return {};
+  }
+
+  const json::value* extract_ptr = page.if_contains("extract");
+
+  if ((extract_ptr == nullptr) || !extract_ptr->is_string()) {
+    if (logger_) {
+      logger_->Log(
+          {.level = LogLevel::Warn,
+           .phase = PipelinePhase::UserGeneration,
+           .message =
+               std::format("WikipediaService: No extract string found for '{}'",
+               std::string(query))});
+    }
+    this->extract_cache_.emplace(cache_key, "");
+    return {};
+  }
+
+  // 4. Success
+  std::string extract(extract_ptr->as_string());
+  if (logger_) {
+    logger_->Log({.level = LogLevel::Info,
+                  .phase = PipelinePhase::UserGeneration,
+                  .message = std::format("WikipediaService: Fetched {} chars for '{}'",
+                             extract.size(), std::string(query))});
+  }
+
+  this->extract_cache_.insert_or_assign(cache_key, extract);
+
+  return extract;
+}

tooling/pipeline/src/services/enrichment/wikipedia/get_summary.cc

@@ -0,0 +1,70 @@
+/**
+ * @file wikipedia/get_summary.cc
+ * @brief WikipediaService::GetLocationContext() implementation.
+ */
+
+#include <chrono>
+#include <format>
+#include <string>
+#include <thread>
+
+#include "services/enrichment/wikipedia_service.h"
+
+std::string WikipediaEnrichmentService::GetLocationContext(
+    const Location& loc) {
+  using namespace std::literals::chrono_literals;
+  if (!this->client_) {
+    if (logger_) {
+      logger_->Log({.level = LogLevel::Warn,
+                    .phase = PipelinePhase::UserGeneration,
+                    .message = "Wikipedia client is nullptr."});
+    }
+    return {};
+  }
+
+  std::string result;
+
+  // std::string region_query(loc.city);
+  // if (!loc.country.empty()) {
+  //   region_query += loc.state_province,
+  //   region_query += ", ";
+  //   region_query += loc.country;
+  // }
+
+  constexpr std::string_view brewing_query = "brewing";
+  const std::string location_query =
+      std::format("{}, {}", loc.city, loc.iso3166_2);
+  const std::string beer_query = std::format("beer in {}", loc.country);
+
+  auto append_extract = [&result](const std::string& extract) -> void {
+    if (extract.empty()) {
+      return;
+    }
+    if (!result.empty()) {
+      result += "\n\n";
+    }
+    result += extract;
+  };
+
+  try {
+    append_extract(FetchExtract(brewing_query));
+    append_extract(FetchExtract(beer_query));
+    if (logger_) {
+      logger_->Log({.level = LogLevel::Info,
+                    .phase = PipelinePhase::UserGeneration,
+                    .message = std::format("Done fetching for {}. Sleeping for 10 seconds.",
+                               location_query)});
+    }
+    std::this_thread::sleep_for(10s);
+
+  } catch (const std::runtime_error& e) {
+    if (logger_) {
+      logger_->Log(
+          {.level = LogLevel::Debug,
+           .phase = PipelinePhase::UserGeneration,
+           .message = std::format("WikipediaService lookup failed for '{}': {}",
+                      location_query, e.what())});
+    }
+  }
+  return result;
+}

tooling/pipeline/src/services/enrichment/wikipedia/wikipedia_service.cc

@@ -0,0 +1,12 @@
+/**
+ * @file services/wikipedia/wikipedia_service.cc
+ * @brief WikipediaService constructor implementation.
+ */
+
+#include "services/enrichment/wikipedia_service.h"
+
+#include <utility>
+
+WikipediaEnrichmentService::WikipediaEnrichmentService(
+    std::unique_ptr<WebClient> client, std::shared_ptr<ILogger> logger)
+    : client_(std::move(client)), logger_(std::move(logger)) {}

tooling/pipeline/src/services/logging/log_dispatcher.cc

@@ -0,0 +1,74 @@
+/**
+ * @brief LogDispatcher implementation for asynchronous pipeline logging.
+ *
+ * LogDispatcher drains LogEntry items from a BoundedChannel and forwards them
+ * to spdlog for final output.
+ */
+#include "services/logging/log_dispatcher.h"
+
+#include <spdlog/spdlog.h>
+
+#include <string>
+
+#include "concurrency/bounded_channel.h"
+#include "services/logging/log_entry.h"
+
+namespace {
+[[nodiscard]] constexpr std::string_view PipelinePhaseToString(
+    PipelinePhase phase) {
+  switch (phase) {
+    case PipelinePhase::Startup:
+      return "Startup";
+    case PipelinePhase::UserGeneration:
+      return "User Generation";
+    case PipelinePhase::BreweryAndBeerGeneration:
+      return "Brewery & Beer Gen";
+    case PipelinePhase::CheckinGeneration:
+      return "Checkin Gen";
+    case PipelinePhase::RatingGeneration:
+      return "Rating Gen";
+    case PipelinePhase::FollowGeneration:
+      return "Follow Gen";
+    case PipelinePhase::Teardown:
+      return "Teardown";
+  }
+  return "Unknown";
+}
+}  // namespace
+
+LogDispatcher::LogDispatcher(BoundedChannel<LogEntry>& channel)
+    : channel_(channel) {}
+
+void LogDispatcher::Run() {
+  auto logger = spdlog::default_logger();
+
+  while (true) {
+    auto entry = channel_.Receive();
+    if (!entry.has_value()) {
+      // Channel is closed and drained.
+      break;
+    }
+
+    const auto& log = entry.value();
+
+    logger->log(ToSpdlogLevel(log.level),
+                "{:<20} │ thread: {:016x} │ [{}:{}] │ {}",
+                PipelinePhaseToString(log.phase),
+                std::hash<std::thread::id>{}(log.thread_id),
+                log.origin.file_name(), log.origin.line(), log.message);
+  }
+}
+
+spdlog::level::level_enum LogDispatcher::ToSpdlogLevel(LogLevel level) {
+  switch (level) {
+    case LogLevel::Debug:
+      return spdlog::level::debug;
+    case LogLevel::Info:
+      return spdlog::level::info;
+    case LogLevel::Warn:
+      return spdlog::level::warn;
+    case LogLevel::Error:
+      return spdlog::level::err;
+  }
+  return spdlog::level::info;
+}

tooling/pipeline/src/services/logging/log_producer.cc

@@ -0,0 +1,19 @@
+/**
+ * @file src/services/logging/log_producer.cc
+ * @brief LogProducer implementation for asynchronous pipeline logging.
+ */
+
+#include "services/logging/log_producer.h"
+
+#include <chrono>
+#include <optional>
+#include <string>
+#include <string_view>
+
+#include "concurrency/bounded_channel.h"
+#include "services/logging/log_entry.h"
+
+LogProducer::LogProducer(BoundedChannel<LogEntry>& channel)
+    : channel_(channel) {}
+
+void LogProducer::DoLog(LogEntry entry) { channel_.Send(std::move(entry)); }

tooling/pipeline/src/services/prompt_directory.cc

@@ -6,20 +6,25 @@
 
 #include "services/prompting/prompt_directory.h"
 
-#include <spdlog/spdlog.h>
+#include <chrono>
-
 #include <filesystem>
+#include <format>
 #include <fstream>
 #include <stdexcept>
 #include <string>
 #include <string_view>
+#include <utility>
 
 // ---------------------------------------------------------------------------
 // PromptDirectory
 // ---------------------------------------------------------------------------
 
 PromptDirectory::PromptDirectory(const std::filesystem::path& prompt_dir)
-    : prompt_dir_(prompt_dir) {
+    : PromptDirectory(prompt_dir, nullptr) {}
+
+PromptDirectory::PromptDirectory(const std::filesystem::path& prompt_dir,
+                                 std::shared_ptr<ILogger> logger)
+    : prompt_dir_(prompt_dir), logger_(std::move(logger)) {
   std::error_code ec;
 
   // Scenario 4: directory must exist.
@@ -40,12 +45,18 @@ PromptDirectory::PromptDirectory(const std::filesystem::path& prompt_dir)
   std::filesystem::directory_iterator probe(prompt_dir_, ec);
   if (ec) {
     throw std::runtime_error(
-        "PromptDirectory: prompt directory is not readable: " +
+        std::format("PromptDirectory: prompt directory is not readable: {} ({})",
-        prompt_dir_.string() + " (" + ec.message() + ")");
+        prompt_dir_.string(), ec.message()));
   }
 
-  spdlog::info("[PromptDirectory] Resolved prompt directory: {}",
+  if (logger_) {
-               prompt_dir_.string());
+    logger_->Log(
+        {.level = LogLevel::Info,
+         .phase = PipelinePhase::Startup,
+         .message =
+             std::string("[PromptDirectory] Resolved prompt directory: ") +
+             prompt_dir_.string()});
+  }
 }
 
 std::string PromptDirectory::Load(std::string_view key) {
@@ -59,13 +70,13 @@ std::string PromptDirectory::Load(std::string_view key) {
 
   // Scenario 3: resolve <prompt_dir>/<key>.md and require it to exist.
   const std::filesystem::path file_path =
-      prompt_dir_ / std::filesystem::path(key_str + ".md");
+      prompt_dir_ / std::filesystem::path(std::format("{}.md", key_str));
 
   std::ifstream file(file_path);
   if (!file.is_open()) {
     throw std::runtime_error(
-        "PromptDirectory: prompt file not found for key '" + key_str +
+        std::format("PromptDirectory: prompt file not found for key '{}': {}",
-        "': " + file_path.string());
+        key_str, file_path.string()));
   }
 
   std::string content((std::istreambuf_iterator<char>(file)),
@@ -73,12 +84,16 @@ std::string PromptDirectory::Load(std::string_view key) {
   file.close();
 
   if (content.empty()) {
-    throw std::runtime_error("PromptDirectory: prompt file for key '" +
+    throw std::runtime_error(std::format("PromptDirectory: prompt file for key '{}' is empty: {}",
-                             key_str + "' is empty: " + file_path.string());
+                             key_str, file_path.string()));
   }
 
-  spdlog::info("[PromptDirectory] Loaded prompt '{}' from '{}' ({} chars)",
+  if (logger_) {
-               key_str, file_path.string(), content.size());
+    logger_->Log({.level = LogLevel::Info,
+                  .phase = PipelinePhase::Startup,
+                  .message = std::format("[PromptDirectory] Loaded prompt '{}' from '{}' ({} chars)",
+                             key_str, file_path.string(), content.size())});
+  }
 
   cache_.emplace(key_str, content);
   return content;

tooling/pipeline/src/services/sqlite/helpers/sqlite_connection_helpers.cc

@@ -1,5 +1,6 @@
 #include "services/database/sqlite_connection_helpers.h"
 
+#include <format>
 #include <stdexcept>
 
 namespace sqlite_export_service_internal {
@@ -20,7 +21,7 @@ void SqliteStatementDeleter::operator()(
 void ThrowSqliteError(sqlite3* db_handle, std::string_view action) {
   const std::string message =
       db_handle != nullptr ? sqlite3_errmsg(db_handle) : "unknown SQLite error";
-  throw std::runtime_error(std::string(action) + ": " + message);
+  throw std::runtime_error(std::format("{}: {}", action, message));
 }
 
 SqliteDatabaseHandle OpenDatabase(const std::filesystem::path& path) {
@@ -50,7 +51,7 @@ void ExecSql(const SqliteDatabaseHandle& db_handle, std::string_view sql,
                                     ? error_message
                                     : sqlite3_errmsg(db_handle.get());
     sqlite3_free(error_message);
-    throw std::runtime_error(std::string(action) + ": " + message);
+    throw std::runtime_error(std::format("{}: {}", action, message));
   }
 }
 

tooling/pipeline/src/services/sqlite/initialize.cc

@@ -4,6 +4,7 @@
  */
 
 #include <filesystem>
+#include <format>
 #include <memory>
 #include <stdexcept>
 #include <string>
@@ -18,8 +19,8 @@ std::filesystem::path SqliteExportService::BuildDatabasePath() const {
 
   for (int suffix = 1; std::filesystem::exists(candidate); ++suffix) {
     candidate = output_path_ /
-                std::filesystem::path("biergarten_seed_" + run_timestamp_utc_ +
+                std::filesystem::path(std::format("biergarten_seed_{}-{}.sqlite",
-                                      "-" + std::to_string(suffix) + ".sqlite");
+                                      run_timestamp_utc_, suffix));
   }
 
   return candidate;

tooling/pipeline/src/services/wikipedia/fetch_extract.cc

@@ -1,61 +0,0 @@
-/**
- * @file wikipedia/fetch_extract.cc
- * @brief WikipediaService::FetchExtract() implementation.
- */
-
-#include <spdlog/spdlog.h>
-
-#include <boost/json.hpp>
-#include <string>
-#include <string_view>
-
-#include "services/enrichment/wikipedia_service.h"
-
-std::string WikipediaService::FetchExtract(std::string_view query) {
-  const std::string cache_key(query);
-  const auto cache_it = this->extract_cache_.find(cache_key);
-  if (cache_it != this->extract_cache_.end()) {
-    return cache_it->second;
-  }
-
-  const std::string encoded = this->client_->UrlEncode(cache_key);
-  const std::string url =
-      "https://en.wikipedia.org/w/api.php?action=query&titles=" + encoded +
-      "&prop=extracts&explaintext=1&format=json";
-
-  const std::string body = this->client_->Get(url);
-
-  boost::system::error_code parse_error;
-  boost::json::value doc = boost::json::parse(body, parse_error);
-
-  if (!parse_error && doc.is_object()) {
-    try {
-      auto& pages = doc.at("query").at("pages").get_object();
-      if (!pages.empty()) {
-        auto& page = pages.begin()->value().get_object();
-        if (page.contains("extract") && page.at("extract").is_string()) {
-          const std::string_view extract_view = page.at("extract").as_string();
-          std::string extract(extract_view);
-
-          spdlog::debug("WikipediaService fetched {} chars for '{}'",
-                        extract.size(), query);
-
-          this->extract_cache_.emplace(cache_key, extract);
-          return extract;
-        }
-      }
-      this->extract_cache_.emplace(cache_key, std::string{});
-    } catch (const std::exception& e) {
-      spdlog::warn(
-          "WikipediaService: failed to parse response structure for '{}': "
-          "{}",
-          query, e.what());
-      return {};
-    }
-  } else if (parse_error) {
-    spdlog::warn("WikipediaService: JSON parse error for '{}': {}", query,
-                 parse_error.message());
-  }
-
-  return {};
-}

tooling/pipeline/src/services/wikipedia/get_summary.cc

@@ -1,47 +0,0 @@
-/**
- * @file wikipedia/get_summary.cc
- * @brief WikipediaService::GetLocationContext() implementation.
- */
-
-#include <spdlog/spdlog.h>
-
-#include <string>
-
-#include "services/enrichment/wikipedia_service.h"
-
-std::string WikipediaService::GetLocationContext(const Location& loc) {
-  if (!client_) {
-    return {};
-  }
-
-  std::string result;
-
-  std::string region_query(loc.city);
-  if (!loc.country.empty()) {
-    region_query += ", ";
-    region_query += loc.country;
-  }
-
-  const std::string beer_query = "beer in " + loc.country;
-  const std::string city_beer_query = "beer in " + loc.city;
-
-  auto append_extract = [&result](const std::string& extract) -> void {
-    if (extract.empty()) {
-      return;
-    }
-    if (!result.empty()) {
-      result += "\n\n";
-    }
-    result += extract;
-  };
-
-  try {
-    append_extract(FetchExtract(region_query));
-    append_extract(FetchExtract(beer_query));
-    append_extract(FetchExtract(city_beer_query));
-  } catch (const std::runtime_error& e) {
-    spdlog::debug("WikipediaService lookup failed for '{}': {}", region_query,
-                  e.what());
-  }
-  return result;
-}

tooling/pipeline/src/services/wikipedia/wikipedia_service.cc

@@ -1,11 +0,0 @@
-/**
- * @file services/wikipedia/wikipedia_service.cc
- * @brief WikipediaService constructor implementation.
- */
-
-#include "services/enrichment/wikipedia_service.h"
-
-#include <utility>
-
-WikipediaService::WikipediaService(std::unique_ptr<WebClient> client)
-    : client_(std::move(client)) {}

tooling/pipeline/src/web_client/http_web_client.cc

@@ -1,5 +1,5 @@
 /**
-* @file web_client/http_web_client.cc
+ * @file web_client/http_web_client.cc
  * @brief cpp-httplib implementation of WebClient.
  */
 
@@ -7,11 +7,15 @@
 
 #include <httplib.h>
 
+#include <chrono>
+#include <format>
 #include <regex>
 #include <stdexcept>
 #include <string>
 #include <utility>
 
+#include "services/logging/logger.h"
+
 namespace {
 constexpr time_t kConnectionTimeoutSeconds = 5;
 constexpr time_t kReadTimeoutSeconds = 10;
@@ -29,7 +33,7 @@ std::pair<std::string, std::string> SplitUrl(const std::string& url) {
 
   return {match[1].str(), match[2].matched ? match[2].str() : "/"};
 }
-} // namespace
+}  // namespace
 
 std::string HttpWebClient::Get(const std::string& url) {
   const auto [origin, path] = SplitUrl(url);
@@ -38,24 +42,32 @@ std::string HttpWebClient::Get(const std::string& url) {
   client.set_follow_location(true);
   client.set_connection_timeout(kConnectionTimeoutSeconds);
   client.set_read_timeout(kReadTimeoutSeconds);
+  client.set_default_headers({{"Accept", "application/json"},
+                              {"User-Agent", "biergarten-pipeline/1.0"}});
 
-  const auto result = client.Get(path);
+  const httplib::Result result = client.Get(path);
 
   if (!result) {
-    throw std::runtime_error(
+    throw std::runtime_error(std::format(
-        "[HttpWebClient] Request failed for URL: " + url +
+        "[HttpWebClient] Request failed for URL: {} — {}", url,
-        " — " + httplib::to_string(result.error()));
+        httplib::to_string(result.error())));
   }
 
   if (result->status < kSuccessMin || result->status >= kSuccessMax) {
-    throw std::runtime_error(
+    if (logger_) {
-        "[HttpWebClient] HTTP " + std::to_string(result->status) +
+      logger_->Log(
-        " for URL: " + url);
+          {.level = LogLevel::Error,
+           .phase = PipelinePhase::UserGeneration,
+         .message =
+           std::format("[HttpWebClient] Request failed for URL: {}", url)});
+    }
+    throw std::runtime_error(std::format("[HttpWebClient] HTTP {} for URL: {}",
+                       result->status, url));
   }
 
   return result->body;
 }
 
-std::string HttpWebClient::UrlEncode(const std::string& value) {
+std::string HttpWebClient::EncodeURL(const std::string& value) {
   return httplib::encode_uri_component(value);
-}
+}