diff --git a/tooling/pipeline/includes/data_model/models.h b/tooling/pipeline/includes/data_model/models.h
index 05d4bc45..2666c3c1 100644
--- a/tooling/pipeline/includes/data_model/models.h
+++ b/tooling/pipeline/includes/data_model/models.h
@@ -191,9 +191,9 @@ struct SamplingOptions {
  * @brief Selects which DataGenerator backend the pipeline uses.
  */
 enum class GeneratorMode {
-   Mock,
-   Llama,
-   OpenAI,
+   kMock,
+   kLlama,
+   kOpenAI,
 };
 
 /**
@@ -202,25 +202,25 @@ enum class GeneratorMode {
 struct GeneratorOptions {
    /**
     * @brief Path to the LLM model file (gguf format). Only meaningful for
-    * GeneratorMode::Llama.
+    * GeneratorMode::kLlama.
     */
    std::filesystem::path model_path;
 
    /**
     * @brief Which generator backend to use.
     */
-   GeneratorMode mode = GeneratorMode::Mock;
+   GeneratorMode mode = GeneratorMode::kMock;
 
    /**
     * @brief Specific sampling parameters for this generator. Only
-    * meaningful for GeneratorMode::Llama; if nullopt, LlamaGenerator uses
+    * meaningful for GeneratorMode::kLlama; if nullopt, LlamaGenerator uses
     * its own defaults.
     */
    std::optional<SamplingOptions> sampling;
 
    /**
     * @brief OpenAI API key, read from the OPENAI_API_KEY environment
-    * variable. Only populated for GeneratorMode::OpenAI.
+    * variable. Only populated for GeneratorMode::kOpenAI.
     */
    std::string openai_api_key;
 
diff --git a/tooling/pipeline/src/application_options/parse_arguments.cc b/tooling/pipeline/src/application_options/parse_arguments.cc
index b1fe5c2a..1822584b 100644
--- a/tooling/pipeline/src/application_options/parse_arguments.cc
+++ b/tooling/pipeline/src/application_options/parse_arguments.cc
@@ -182,16 +182,16 @@ std::optional<ApplicationOptions> ParseArguments(
                             .phase = PipelinePhase::Startup,
                             .message = msg});
             } else {
-               std::cerr << msg << std::endl;
+               std::cerr << msg << '\n';
             }
             return std::nullopt;
          }
          options.generator.openai_api_key = api_key_env;
       }
 
-      options.generator.mode = use_mocked   ? GeneratorMode::Mock
-                               : use_openai ? GeneratorMode::OpenAI
-                                            : GeneratorMode::Llama;
+      options.generator.mode = use_mocked   ? GeneratorMode::kMock
+                               : use_openai ? GeneratorMode::kOpenAI
+                                            : GeneratorMode::kLlama;
       options.generator.model_path = model_path;
       options.generator.openai_model = var_map["openai-model"].as<std::string>();
 
@@ -208,7 +208,7 @@ std::optional<ApplicationOptions> ParseArguments(
       if (user_provided_sampling) {
          // Warn but do not fail — the run is still valid, the flags are just
          // silently irrelevant when no local model is loaded.
-         if (options.generator.mode != GeneratorMode::Llama) {
+         if (options.generator.mode != GeneratorMode::kLlama) {
             const std::string msg =
                 "Sampling parameters are ignored unless using --model "
                 "(Llama)";
diff --git a/tooling/pipeline/src/biergarten_pipeline_orchestrator/generate_breweries.cc b/tooling/pipeline/src/biergarten_pipeline_orchestrator/generate_breweries.cc
index 3a5d24a1..2fa65463 100644
--- a/tooling/pipeline/src/biergarten_pipeline_orchestrator/generate_breweries.cc
+++ b/tooling/pipeline/src/biergarten_pipeline_orchestrator/generate_breweries.cc
@@ -31,8 +31,8 @@ double RadiansToDegrees(double radians) {
 }
 }  // namespace
 
-static Coords GetRandomCoordsWithinRange(Coords const& centre,
-                                         int const distance_km,
+static Coords GetRandomCoordsWithinRange(const Coords& centre,
+                                         const int distance_km,
                                          std::mt19937& rng) {
    std::uniform_real_distribution<double> angle_dist(0.0,
                                                      2.0 * std::numbers::pi);
diff --git a/tooling/pipeline/src/data_generation/generation_json_validation.cc b/tooling/pipeline/src/data_generation/generation_json_validation.cc
index 0fbb3929..42344f16 100644
--- a/tooling/pipeline/src/data_generation/generation_json_validation.cc
+++ b/tooling/pipeline/src/data_generation/generation_json_validation.cc
@@ -8,12 +8,14 @@
 
 #include <algorithm>
 #include <array>
-#include <boost/json.hpp>
 #include <cctype>
+#include <format>
 #include <span>
 #include <string>
 #include <string_view>
 
+#include <boost/json.hpp>
+
 namespace {
 /**
  * String trimming: removes leading and trailing whitespace
@@ -66,11 +68,18 @@ bool ReadRequiredTrimmedStringField(const boost::json::object& obj,
                                     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 =
+             std::format("Missing or invalid string field: {}", key);
+      }
       return false;
    }
 
    const auto& string_value = field->as_string();
    out = Trim(std::string_view(string_value.data(), string_value.size()));
+   if (out.empty() && error_out != nullptr) {
+      *error_out = std::format("Field must not be empty: {}", key);
+   }
    return !out.empty();
 }
 
diff --git a/tooling/pipeline/src/data_generation/openai/call_chat_completions_api.cc b/tooling/pipeline/src/data_generation/openai/call_chat_completions_api.cc
index 05ce3aaf..f878616f 100644
--- a/tooling/pipeline/src/data_generation/openai/call_chat_completions_api.cc
+++ b/tooling/pipeline/src/data_generation/openai/call_chat_completions_api.cc
@@ -9,7 +9,6 @@
  * Structured Outputs: https://platform.openai.com/docs/guides/structured-outputs
  */
 
-#include <boost/json.hpp>
 #include <format>
 #include <stdexcept>
 #include <string>
@@ -17,6 +16,8 @@
 #include <utility>
 #include <vector>
 
+#include <boost/json.hpp>
+
 #include "data_generation/openai_generator.h"
 
 namespace {
diff --git a/tooling/pipeline/src/main.cc b/tooling/pipeline/src/main.cc
index 7a02d08f..32b01709 100644
--- a/tooling/pipeline/src/main.cc
+++ b/tooling/pipeline/src/main.cc
@@ -67,7 +67,7 @@ int main(const int argc, char** argv) {
 
       std::unique_ptr<IPromptDirectory> prompt_directory;
 
-      if (options.generator.mode != GeneratorMode::Mock) {
+      if (options.generator.mode != GeneratorMode::kMock) {
          try {
             prompt_directory = std::make_unique<PromptDirectory>(
                 options.pipeline.prompt_dir, log_producer);
@@ -90,7 +90,7 @@ int main(const int argc, char** argv) {
           di::bind<IPostalCodeService>().to(
               [options,
                &log_producer]() -> std::unique_ptr<IPostalCodeService> {
-                 if (options.generator.mode == GeneratorMode::Mock) {
+                 if (options.generator.mode == GeneratorMode::kMock) {
                     log_producer->Log(
                         {.level = LogLevel::Info,
                          .phase = PipelinePhase::Startup,
@@ -110,7 +110,7 @@ int main(const int argc, char** argv) {
           di::bind<ICuratedDataService>().to(
               [options,
                &log_producer]() -> std::unique_ptr<ICuratedDataService> {
-                 // if (options.generator.mode == GeneratorMode::Mock) {
+                 // if (options.generator.mode == GeneratorMode::kMock) {
                  //    log_producer->Log({.level = LogLevel::Info,
                  //                       .phase = PipelinePhase::Startup,
                  //                       .message = "Curated data: mock"});
@@ -131,7 +131,7 @@ int main(const int argc, char** argv) {
                      });
               }),
           di::bind<IPromptFormatter>().to([options, log_producer] {
-             if (options.generator.mode != GeneratorMode::Llama) {
+             if (options.generator.mode != GeneratorMode::kLlama) {
                 log_producer->Log(
                     {.level = LogLevel::Info,
                      .phase = PipelinePhase::Startup,
@@ -150,7 +150,7 @@ int main(const int argc, char** argv) {
                  std::make_unique<Gemma4JinjaPromptFormatter>());
           }),
           di::bind<WebClient>().to([options, log_producer] {
-             if (options.generator.mode == GeneratorMode::Mock) {
+             if (options.generator.mode == GeneratorMode::kMock) {
                 {
                    log_producer->Log(
                        {.level = LogLevel::Info,
@@ -170,7 +170,7 @@ int main(const int argc, char** argv) {
           di::bind<IEnrichmentService>().to(
               [options, &log_producer](
                   const auto& inj) -> std::unique_ptr<IEnrichmentService> {
-                 if (options.generator.mode == GeneratorMode::Mock) {
+                 if (options.generator.mode == GeneratorMode::kMock) {
                     log_producer->Log({.level = LogLevel::Info,
                                        .phase = PipelinePhase::Startup,
                                        .message = "Enrichment: mock"});
@@ -190,7 +190,7 @@ int main(const int argc, char** argv) {
               [&options, &model_path, &sampling, &prompt_directory,
                &log_producer](
                   const auto& inj) -> std::unique_ptr<DataGenerator> {
-                 if (options.generator.mode == GeneratorMode::Mock) {
+                 if (options.generator.mode == GeneratorMode::kMock) {
                     log_producer->Log({.level = LogLevel::Info,
                                        .phase = PipelinePhase::Startup,
                                        .message = "Generator: mock"});
@@ -198,7 +198,7 @@ int main(const int argc, char** argv) {
                     return std::make_unique<MockGenerator>();
                  }
 
-                 if (options.generator.mode == GeneratorMode::OpenAI) {
+                 if (options.generator.mode == GeneratorMode::kOpenAI) {
                     log_producer->Log(
                         {.level = LogLevel::Info,
                          .phase = PipelinePhase::Startup,