Add local language handling

pipeline/LLAMA_CPP_GEMMA4_GUIDE.md

@@ -1,565 +0,0 @@
-# A Beginner's Guide to llama.cpp and Google Gemma 4
-
-## Table of Contents
-
-1. [Introduction](#introduction)
-2. [What is llama.cpp?](#what-is-llamacpp)
-3. [What is Google Gemma 4?](#what-is-google-gemma-4)
-4. [Why Use llama.cpp with Gemma 4?](#why-use-llamacpp-with-gemma-4)
-5. [Getting Started with llama.cpp](#getting-started-with-llamacpp)
-6. [Understanding Chat Templates](#understanding-chat-templates)
-7. [Gemma 4's Reasoning Engine](#gemma-4s-reasoning-engine)
-8. [Performance Optimization](#performance-optimization)
-9. [Common Pitfalls](#common-pitfalls)
-10. [References and Further Reading](#references-and-further-reading)
-
----
-
-## Introduction
-
-This guide is designed for developers and AI enthusiasts who want to run large language models locally and efficiently. Whether you're building a chatbot, conducting research, or simply exploring AI capabilities, understanding llama.cpp and Gemma 4 will help you make informed decisions about your setup.
-
-**Target Audience:** Developers with basic C/C++ knowledge, DevOps engineers, and AI practitioners.
-
----
-
-## What is llama.cpp?
-
-### Overview
-
-llama.cpp is a plain C/C++ implementation for Large Language Model (LLM) inference designed to enable efficient LLM inference with minimal setup and state-of-the-art performance across diverse hardware configurations—both locally and in the cloud.[^1]
-
-According to the official project description: *"The main goal of `llama.cpp` is to enable LLM inference with minimal setup and state-of-the-art performance on a wide range of hardware - locally and in the cloud."*[^1]
-
-### Key Features
-
-llama.cpp provides comprehensive support for inference acceleration:
-
-- **Plain C/C++ Implementation:** No complex dependencies, making it portable and lightweight[^1]
-- **Multi-Platform Support:** 
-  - Apple Silicon optimization via ARM NEON, Accelerate, and Metal frameworks[^1]
-  - x86 architectures: AVX, AVX2, AVX512, and AMX support[^1]
-  - RISC-V architectures: RVV, ZVFH, ZFH, ZICBOP, and ZIHINTPAUSE support[^1]
-
-- **Quantization Support:** 1.5-bit, 2-bit, 3-bit, 4-bit, 5-bit, 6-bit, and 8-bit integer quantization for faster inference and reduced memory usage[^1]
-
-- **GPU Acceleration:** 
-  - Custom CUDA kernels for NVIDIA GPUs[^1]
-  - AMD GPU support via HIP[^1]
-  - Vulkan and SYCL backend support[^1]
-
-- **Hybrid Inference:** CPU+GPU hybrid mode for models larger than total VRAM capacity[^1]
-
-### Installation
-
-llama.cpp can be installed through multiple methods:[^1]
-
-```bash
-# Package managers
-brew install llama.cpp          # macOS
-nix flake show github:ggml-org/llama.cpp  # NixOS
-winget install LlamaCpp         # Windows
-
-# Docker
-docker pull ghcr.io/ggml-org/llama.cpp:server-latest
-
-# From source
-git clone https://github.com/ggml-org/llama.cpp
-cd llama.cpp
-make
-```
-
-### Basic Usage
-
-Once installed, running llama.cpp is straightforward:[^1]
-
-```bash
-# Run locally with a GGUF model file
-llama-cli -m my_model.gguf
-
-# Download and run directly from Hugging Face
-llama-cli -hf ggml-org/gemma-3-1b-it-GGUF
-
-# Launch OpenAI-compatible API server
-llama-server -hf ggml-org/gemma-3-1b-it-GGUF
-```
-
----
-
-## What is Google Gemma 4?
-
-### Overview
-
-Google's Gemma is a family of open-source lightweight Large Language Models that represent the latest breakthroughs in AI research. Gemma models are built with the same research and technology used to create Gemini, Google's advanced AI model.[^2]
-
-The Gemma family includes various sizes optimized for different use cases:
-
-- **Gemma 2:** Available in 9B and 27B parameter variants[^3]
-- **Gemma 4:** The latest generation with advanced reasoning and instruction-tuning capabilities
-
-### Model Variants
-
-Gemma models are available in multiple configurations, with "-it" suffix indicating instruction-tuned versions optimized for chat and dialogue:
-
-- **Base Models:** Designed for text completion and continuation
-- **Instruction-Tuned Models (-it):** Fine-tuned for conversational interactions and following instructions[^3]
-
-### Architecture and Training
-
-Gemma models are built on proven transformer architecture with modern training techniques including:
-
-- Flash Attention for efficient attention computation[^4]
-- Robust quantization-friendly training
-- Extensive safety and alignment training
-
-*Reference:* "Gemma models are trained for safety and helpfulness, incorporating feedback from our safety team across all stages of development."[^2]
-
----
-
-## Why Use llama.cpp with Gemma 4?
-
-### Performance and Efficiency
-
-llama.cpp is specifically optimized for inference workloads, making it ideal for running Gemma 4 models:
-
-1. **Speed:** Highly optimized C/C++ implementation delivers faster token generation compared to Python frameworks[^1]
-2. **Memory Efficiency:** Support for aggressive quantization (4-bit, 3-bit) reduces model size significantly[^1]
-3. **Portability:** Run the same model on laptops, desktops, cloud instances, and edge devices[^1]
-4. **Resource Flexibility:** CPU-only inference is viable; GPU acceleration available when hardware permits[^1]
-
-### Use Cases
-
-**Development and Experimentation**
-- Rapid prototyping without GPU requirements
-- Local testing and debugging of prompts
-- Quantization experimentation
-
-**Production Deployment**
-- Low-latency API servers via `llama-server`[^1]
-- OpenAI-compatible REST API endpoints
-- Edge deployment on resource-constrained devices
-
-**Research**
-- Analyzing model behavior at scale
-- Benchmark studies with consistent inference runtime
-- Fine-tuning and adapter experiments
-
----
-
-## Getting Started with llama.cpp
-
-### Step 1: Build from Source
-
-```bash
-# Clone the repository
-git clone https://github.com/ggml-org/llama.cpp.git
-cd llama.cpp
-
-# Build with optimizations (CPU + optional GPU)
-make
-
-# Optional: Build with CUDA support
-make LLAMA_CUDA=1
-
-# Optional: Build with Metal (Apple Silicon)
-make LLAMA_METAL=1
-```
-
-### Step 2: Obtain a Model
-
-Gemma 4 models are available on Hugging Face in GGUF format (optimized for llama.cpp):[^5]
-
-```bash
-# Download Gemma 4 model (automatic via llama.cpp)
-llama-cli -hf google/gemma-4-9b-it-GGUF
-
-# Or manually download from:
-# https://huggingface.co/google/gemma-4-9b-it-GGUF
-```
-
-**GGUF Format:** GGUF (GUFF) is a quantized model format designed for efficient inference in llama.cpp. It stores model weights in a compressed binary format with metadata.[^6]
-
-### Step 3: Run Inference
-
-```bash
-# Interactive chat mode
-llama-cli -m gemma-4-9b-it.gguf -p "Hello, how are you?" -n 256
-
-# With explicit chat template (if needed)
-llama-cli -m gemma-4-9b-it.gguf --chat-template gemma -p "You are a helpful assistant."
-
-# Start API server
-llama-server -m gemma-4-9b-it.gguf -c 2048
-```
-
----
-
-## Understanding Chat Templates
-
-### What are Chat Templates?
-
-Chat templates are Jinja2-based formatting specifications that define how multi-turn conversations are structured for model input.[^7] They ensure consistent formatting of user messages, system prompts, and assistant responses.
-
-According to the llama.cpp documentation: *"Chat templates are Jinja templates that transform a list of messages into a formatted prompt suitable for the model's training format."*[^7]
-
-### Built-in Templates
-
-llama.cpp includes templates for popular models. The "gemma" template is a built-in alias:[^7]
-
-```bash
-# Use built-in Gemma template
-llama-server --chat-template gemma
-
-# List available templates
-llama-cli --list-templates
-```
-
-### Gemma Chat Format
-
-The Gemma chat template uses `<start_of_turn>` and `<end_of_turn>` markers:[^7]
-
-```
-<start_of_turn>user
-What is quantum computing?<end_of_turn>
-<start_of_turn>model
-Quantum computing uses quantum bits (qubits)...<end_of_turn>
-<start_of_turn>user
-Tell me more.<end_of_turn>
-<start_of_turn>model
-```
-
-### Custom Templates
-
-You can provide custom chat templates via file:
-
-```bash
-llama-server -m model.gguf --chat-template-file my_template.jinja
-```
-
-A custom template file example:
-
-```jinja
-{%- for message in messages %}
-[{{ message['role'].upper() }}]
-{{ message['content'] }}
-{% endfor -%}
-```
-
----
-
-## Gemma 4's Reasoning Engine
-
-### Introduction to Reasoning Capabilities
-
-Google Gemma 4 includes advanced reasoning capabilities that enable the model to think through problems step-by-step before generating responses.[^8]
-
-### Activating the Reasoning Engine
-
-To enable Gemma 4's thinking/reasoning mode, prepend the `<|think|>` token to your system prompt:[^8]
-
-```markdown
-<|think|>
-You are a helpful assistant that solves problems step-by-step.
-Please reason through the user's request carefully.
-```
-
-### How It Works
-
-When the reasoning token is detected, the model:
-
-1. **Allocates computational resources** for intermediate reasoning
-2. **Generates internal thoughts** before the final response
-3. **Produces more accurate answers** by working through logic explicitly
-
-### Example Usage
-
-**Without reasoning:**
-```
-Q: What is 47 × 8?
-A: 376
-```
-
-**With reasoning enabled:**
-```
-<|think|>You have advanced reasoning capabilities.
-
-Q: A store sells widgets at $3 each. If they sell 150 per week, 
-   what's their revenue per month assuming 4.3 weeks per month?
-
-A: [Model reasons through calculation internally]
-
-47 × 8 = 376. But let me verify: 40 × 8 = 320, 7 × 8 = 56, 
-so 320 + 56 = 376. ✓
-```
-
-### Implementation in Application Code
-
-In C++, activate reasoning by including the token in your system prompt:
-
-```cpp
-std::string system_prompt = 
-    "<|think|>\n"
-    "You are an expert problem solver that reasons step-by-step.\n"
-    "Always explain your reasoning before providing the answer.";
-
-std::string user_prompt = "What is the square root of 144?";
-
-// Pass to llama_chat_apply_template as normal
-std::string formatted = ToChatPrompt(model, system_prompt, user_prompt);
-```
-
----
-
-## Performance Optimization
-
-### Quantization Strategy
-
-Model quantization reduces file size and memory requirements while maintaining quality. Gemma 4 works well with multiple quantization levels:[^1]
-
-| Quantization | Size Reduction | Quality Impact | Best For |
-|--------------|----------------|----------------|----------|
-| Q8_0 (8-bit) | ~1/8           | Minimal        | Highest quality, CPU inference |
-| Q6_K         | ~1/4           | Very small     | Balanced (recommended) |
-| Q5_K         | ~1/5           | Small          | Good balance |
-| Q4_K_M       | ~1/3           | Noticeable     | GPU inference, moderate quality |
-| Q3_K         | ~1/3           | Moderate       | Limited memory, acceptable quality |
-
-**Recommendation for Gemma 4:** Use Q6_K or Q5_K quantization for optimal quality-to-performance ratio.[^1]
-
-### Buffer Management
-
-When processing prompts, llama.cpp dynamically resizes buffers to accommodate model output:[^9]
-
-```cpp
-// Initial buffer allocation
-std::vector<char> buffer(
-    std::max(min_buffer_size,
-             (system_prompt.size() + user_prompt.size()) * 4));
-
-// If needed, resize on second pass
-if (result >= buffer_size) {
-    buffer.resize(result + 1);  // Resize to actual required size
-    result = llama_chat_apply_template(
-        template_str, messages, n_msg, true, 
-        buffer.data(), static_cast<int32_t>(buffer.size())  // Use NEW size
-    );
-}
-```
-
-**Critical Point:** Always update the size parameter on retry to reflect the resized buffer capacity.[^9]
-
-### Context Window Optimization
-
-Larger context windows enable longer conversations but use more memory:
-
-```bash
-# Default context (2048 tokens)
-llama-server -m model.gguf
-
-# Larger context for longer conversations
-llama-server -m model.gguf -c 4096
-
-# Maximum context (may require GPU)
-llama-server -m model.gguf -c 16384 -ngl 35  # GPU layers
-```
-
----
-
-## Common Pitfalls
-
-### 1. Template Metadata Missing from GGUF
-
-**Problem:** Model lacks chat template metadata, causing fallback to raw text.
-
-**Solution:** Use the built-in "gemma" alias when metadata is unavailable:
-
-```cpp
-const char* tmpl = llama_model_chat_template(model, nullptr);
-if (tmpl == nullptr) {
-    tmpl = "gemma";  // Fall back to built-in alias
-}
-```
-
-### 2. Buffer Overflow During Template Application
-
-**Problem:** Initial buffer too small, causing truncated output.
-
-**Solution:** Implement dynamic resizing with correct size update:
-
-```cpp
-int32_t result = llama_chat_apply_template(
-    template_str, messages, msg_count, true,
-    buffer.data(), static_cast<int32_t>(buffer.size()));
-
-if (result >= static_cast<int32_t>(buffer.size())) {
-    buffer.resize(result + 1);
-    // IMPORTANT: Pass new buffer size
-    result = llama_chat_apply_template(
-        template_str, messages, msg_count, true,
-        buffer.data(), static_cast<int32_t>(buffer.size())  // New size!
-    );
-}
-```
-
-### 3. Incorrect System Prompt Format
-
-**Problem:** System prompt not recognized by Gemma template.
-
-**Solution:** Use standard role-based format with `<start_of_turn>`:
-
-```
-✓ Correct:
-<start_of_turn>user
-Your question here<end_of_turn>
-
-✗ Incorrect:
-System: [prompt]
-User: [question]
-```
-
-### 4. Token Limit Exceeded
-
-**Problem:** "Token count exceeds context window" errors.
-
-**Solution:** Check and limit input size before inference:
-
-```cpp
-const size_t max_tokens = context_size - safety_buffer;
-if (tokens.size() > max_tokens) {
-    // Truncate or summarize input
-    tokens.resize(max_tokens);
-}
-```
-
-### 5. GPU Memory Exhaustion
-
-**Problem:** Out of VRAM during inference.
-
-**Solution:** Reduce GPU layers or use CPU+GPU hybrid:
-
-```bash
-# Reduce GPU-accelerated layers
-llama-server -m model.gguf -ngl 20
-
-# Use hybrid inference
-llama-server -m model.gguf -ngl 15  # Only load 15 layers on GPU
-```
-
----
-
-## References and Further Reading
-
-### Official Documentation
-
-[^1]: **llama.cpp GitHub Repository**
-    - URL: https://github.com/ggml-org/llama.cpp
-    - Content: Official README with installation, build, and usage instructions
-    - Accessed: April 16, 2026
-
-[^7]: **llama.cpp Chat Template Documentation**
-    - URL: https://github.com/ggml-org/llama.cpp/wiki/Templates-supported-by-llama_chat_apply_template
-    - Content: Comprehensive guide to chat templates and built-in aliases including "gemma"
-    - Accessed: April 16, 2026
-
-### Google Gemma Resources
-
-[^2]: **Google Gemma Official Page**
-    - URL: https://ai.google.dev/gemma
-    - Content: Overview of Gemma model family, architecture, and training details
-    - Accessed: April 16, 2026
-
-[^3]: **Gemma 2 on Hugging Face**
-    - URL: https://huggingface.co/google/gemma-2-9b-it
-    - Content: Model card with architecture details, downloads: 324,845
-    - Accessed: April 16, 2026
-
-[^4]: **Google AI Blog: Gemma Training Details**
-    - URL: https://ai.google.dev/gemma/docs
-    - Content: Technical details on Flash Attention, quantization training, and safety alignment
-    - Accessed: April 16, 2026
-
-[^8]: **Google Gemma Thinking/Reasoning Documentation**
-    - URL: https://ai.google.dev/gemma/docs/capabilities/thinking
-    - Content: Guide to enabling and using Gemma 4's advanced reasoning engine
-    - Accessed: April 16, 2026
-
-### Technical References
-
-[^5]: **Gemma 4 GGUF Models on Hugging Face**
-    - URL: https://huggingface.co/google/gemma-4-9b-it-GGUF
-    - Content: GGUF quantized models optimized for llama.cpp inference
-    - Accessed: April 16, 2026
-
-[^6]: **GGUF Format Specification**
-    - URL: https://github.com/ggml-org/ggml/blob/master/docs/gguf.md
-    - Content: Technical specification of the GGUF binary format for quantized models
-    - Accessed: April 16, 2026
-
-[^9]: **llama.cpp API Reference: Chat Template Application**
-    - URL: https://github.com/ggml-org/llama.cpp/blob/master/include/llama.h
-    - Content: `llama_chat_apply_template()` function signature and buffer management patterns
-    - Accessed: April 16, 2026
-
-### Additional Resources
-
-- **llama.cpp Build Guide:** https://github.com/ggml-org/llama.cpp/blob/master/docs/build.md
-- **Model Quantization Guide:** https://github.com/ggml-org/llama.cpp/blob/master/docs/quantization.md
-- **Docker Support:** https://github.com/ggml-org/llama.cpp/blob/master/docs/docker.md
-- **Hugging Face Model Hub:** https://huggingface.co/models?search=gemma
-
----
-
-## Quick Reference Card
-
-### Common Commands
-
-```bash
-# Interactive chat
-llama-cli -m model.gguf --chat-template gemma
-
-# Start API server
-llama-server -m model.gguf -c 2048
-
-# With GPU acceleration
-llama-server -m model.gguf -ngl 35 -c 4096
-
-# Download and run from Hugging Face
-llama-cli -hf google/gemma-4-9b-it-GGUF
-```
-
-### System Prompt Template for Gemma 4 with Reasoning
-
-```markdown
-<|think|>
-[Model will allocate reasoning resources here]
-
-You are an expert assistant trained to solve problems carefully.
-Your role is to:
-1. Understand the user's question completely
-2. Think through the solution step-by-step
-3. Provide accurate and helpful responses
-4. Explain your reasoning when helpful
-```
-
-### Recommended Settings
-
-- **Model:** Gemma-4-9B-IT (9B parameter instruction-tuned variant)
-- **Quantization:** Q6_K (best quality-performance balance)
-- **Context:** 4096 tokens (good balance for most use cases)
-- **Temperature:** 0.7 (balanced creativity and consistency)
-- **Top-P:** 0.95 (good diversity without nonsense)
-
----
-
-## Conclusion
-
-llama.cpp and Google Gemma 4 represent a powerful combination for running state-of-the-art language models efficiently on various hardware configurations. By understanding chat templates, reasoning capabilities, and performance optimization techniques, you can build robust AI applications that leverage these technologies effectively.
-
-For the latest updates and community support, join the llama.cpp community discussions at https://github.com/ggml-org/llama.cpp/discussions.
-
----
-
-**Last Updated:** April 16, 2026  
-**Guide Version:** 1.0  
-**Compatible With:** llama.cpp b8742+, Gemma 4 models
-

pipeline/README.md

@@ -4,15 +4,6 @@ Biergarten Pipeline is a C++20 command-line tool that reads a local city list, r
 
 ## Tested Hardware & OS
 
-### x86/64 Linux, NVIDIA RTX 2000
-
-- **Host**: ThinkPad P1 Gen 7 (Fedora 43)
-- **CPU**: Intel Core Ultra 7 155H
-- **GPU**: NVIDIA RTX 2000 Ada Generation
-- **Memory**: 32GB
-- **Model**: Gemma 4 E4B: efficient local reasoning; released Apr 2, 2026.
-- **Inference**: llama.cpp with CUDA 12.x support
-
 ### ARM MacOS, M1 Pro
 
 - **Host**: MacBook Pro 14" (2021)
@@ -22,6 +13,15 @@ Biergarten Pipeline is a C++20 command-line tool that reads a local city list, r
 - **Model**: Gemma 4 E4B: efficient local reasoning; released Apr 2, 2026.
 - **Inference**: llama.cpp with Metal (MPS) support
 
+### x86/64 Linux, NVIDIA RTX 2000
+
+- **Host**: ThinkPad P1 Gen 7 (Fedora 43)
+- **CPU**: Intel Core Ultra 7 155H
+- **GPU**: NVIDIA RTX 2000 Ada Generation
+- **Memory**: 32GB
+- **Model**: Gemma 4 E4B: efficient local reasoning; released Apr 2, 2026.
+- **Inference**: llama.cpp with CUDA 12.x support
+
 ## Pipeline
 
 | Stage    | What happens                                                            |

pipeline/biergarten_pipeline.puml

@@ -66,8 +66,10 @@ package "Data models" {
   }
 
   class BreweryResult <<struct>> {
-    +name: std::string
-    +description: std::string
+    +name_en: std::string
+    +description_en: std::string
+    +name_local: std::string
+    +description_local: std::string
   }
 
   class UserResult <<struct>> {

pipeline/includes/data_generation/llama_generator.h

@@ -36,7 +36,7 @@ class LlamaGenerator final : public DataGenerator {
    */
   LlamaGenerator(const ApplicationOptions& options,
                  const std::string& model_path,
-                 std::shared_ptr<IPromptFormatter> prompt_formatter);
+                 std::unique_ptr<IPromptFormatter> prompt_formatter);
 
   ~LlamaGenerator() override;
 

pipeline/includes/data_generation/llama_generator_helpers.h

@@ -12,6 +12,8 @@
 #include <string>
 #include <string_view>
 
+#include "data_model/brewery_result.h"
+
 struct llama_vocab;
 using llama_token = int32_t;
 
@@ -39,13 +41,10 @@ void AppendTokenPiece(const llama_vocab* vocab, llama_token token,
  * @brief Validates and parses brewery JSON output.
  *
  * @param raw Raw model output.
- * @param name_out Parsed brewery name.
- * @param description_out Parsed brewery description.
+ * @param brewery_out Parsed brewery payload.
  * @return Validation error message if invalid, or std::nullopt on success.
  */
 std::optional<std::string> ValidateBreweryJson(const std::string& raw,
-                                               std::string& name_out,
-                                               std::string& description_out,
-                                               std::string& reasoning_out);
+                                               BreweryResult& brewery_out);
 
 #endif  // BIERGARTEN_PIPELINE_INCLUDES_DATA_GENERATION_LLAMA_GENERATOR_HELPERS_H_

pipeline/includes/data_model/brewery_result.h

@@ -12,11 +12,17 @@
  * @brief Generated brewery payload.
  */
 struct BreweryResult {
-  /// @brief Brewery display name.
-  std::string name{};
+  /// @brief Brewery display name in English.
+  std::string name_en;
 
-  /// @brief Brewery description text.
-  std::string description{};
+  /// @brief Brewery description text in English.
+  std::string description_en;
+
+  /// @brief Brewery display name in the local language.
+  std::string name_local;
+
+  /// @brief Brewery description text in the local language.
+  std::string description_local;
 };
 
 #endif  // BIERGARTEN_PIPELINE_INCLUDES_DATA_MODEL_BREWERY_RESULT_H_

pipeline/includes/data_model/location.h

@@ -7,6 +7,7 @@
  */
 
 #include <string>
+#include <vector>
 
 /**
  * @brief Canonical location record for city-level generation.
@@ -27,6 +28,9 @@ struct Location {
   /// @brief ISO 3166-1 country code.
   std::string iso3166_1{};
 
+  /// @brief Local language codes in priority order.
+  std::vector<std::string> local_languages{};
+
   /// @brief Latitude in decimal degrees.
   double latitude{};
 

pipeline/locations.json

@@ -6,7 +6,8 @@
     "country": "South Africa",
     "iso3166_1": "ZA",
     "latitude": -33.9249,
-    "longitude": 18.4241
+    "longitude": 18.4241,
+    "local_languages": ["af", "en", "xh"]
   },
   {
     "city": "Johannesburg",
@@ -15,7 +16,8 @@
     "country": "South Africa",
     "iso3166_1": "ZA",
     "latitude": -26.2041,
-    "longitude": 28.0473
+    "longitude": 28.0473,
+    "local_languages": ["en", "zu", "st", "af"]
   },
   {
     "city": "Durban",
@@ -24,7 +26,8 @@
     "country": "South Africa",
     "iso3166_1": "ZA",
     "latitude": -29.8587,
-    "longitude": 31.0218
+    "longitude": 31.0218,
+    "local_languages": ["zu", "en"]
   },
   {
     "city": "Franschhoek",
@@ -33,7 +36,8 @@
     "country": "South Africa",
     "iso3166_1": "ZA",
     "latitude": -33.9146,
-    "longitude": 19.1198
+    "longitude": 19.1198,
+    "local_languages": ["af", "en"]
   },
   {
     "city": "Nairobi",
@@ -42,7 +46,8 @@
     "country": "Kenya",
     "iso3166_1": "KE",
     "latitude": -1.2921,
-    "longitude": 36.8219
+    "longitude": 36.8219,
+    "local_languages": ["sw", "en"]
   },
   {
     "city": "Buenos Aires",
@@ -51,7 +56,8 @@
     "country": "Argentina",
     "iso3166_1": "AR",
     "latitude": -34.6037,
-    "longitude": -58.3816
+    "longitude": -58.3816,
+    "local_languages": ["es-AR"]
   },
   {
     "city": "Bariloche",
@@ -60,7 +66,8 @@
     "country": "Argentina",
     "iso3166_1": "AR",
     "latitude": -41.1335,
-    "longitude": -71.3103
+    "longitude": -71.3103,
+    "local_languages": ["es-AR"]
   },
   {
     "city": "Bogotá",
@@ -69,7 +76,8 @@
     "country": "Colombia",
     "iso3166_1": "CO",
     "latitude": 4.711,
-    "longitude": -74.0721
+    "longitude": -74.0721,
+    "local_languages": ["es-CO"]
   },
   {
     "city": "Medellín",
@@ -78,7 +86,8 @@
     "country": "Colombia",
     "iso3166_1": "CO",
     "latitude": 6.2442,
-    "longitude": -75.5812
+    "longitude": -75.5812,
+    "local_languages": ["es-CO"]
   },
   {
     "city": "São Paulo",
@@ -87,7 +96,8 @@
     "country": "Brazil",
     "iso3166_1": "BR",
     "latitude": -23.5505,
-    "longitude": -46.6333
+    "longitude": -46.6333,
+    "local_languages": ["pt-BR"]
   },
   {
     "city": "Curitiba",
@@ -96,7 +106,8 @@
     "country": "Brazil",
     "iso3166_1": "BR",
     "latitude": -25.4284,
-    "longitude": -49.2733
+    "longitude": -49.2733,
+    "local_languages": ["pt-BR"]
   },
   {
     "city": "Rio de Janeiro",
@@ -105,7 +116,8 @@
     "country": "Brazil",
     "iso3166_1": "BR",
     "latitude": -22.9068,
-    "longitude": -43.1729
+    "longitude": -43.1729,
+    "local_languages": ["pt-BR"]
   },
   {
     "city": "Santiago",
@@ -114,7 +126,8 @@
     "country": "Chile",
     "iso3166_1": "CL",
     "latitude": -33.4489,
-    "longitude": -70.6693
+    "longitude": -70.6693,
+    "local_languages": ["es-CL"]
   },
   {
     "city": "Valdivia",
@@ -123,7 +136,8 @@
     "country": "Chile",
     "iso3166_1": "CL",
     "latitude": -39.8142,
-    "longitude": -73.2459
+    "longitude": -73.2459,
+    "local_languages": ["es-CL"]
   },
   {
     "city": "Lima",
@@ -132,7 +146,8 @@
     "country": "Peru",
     "iso3166_1": "PE",
     "latitude": -12.0464,
-    "longitude": -77.0428
+    "longitude": -77.0428,
+    "local_languages": ["es-PE"]
   },
   {
     "city": "Tokyo",
@@ -141,7 +156,8 @@
     "country": "Japan",
     "iso3166_1": "JP",
     "latitude": 35.6762,
-    "longitude": 139.6503
+    "longitude": 139.6503,
+    "local_languages": ["ja"]
   },
   {
     "city": "Osaka",
@@ -150,7 +166,8 @@
     "country": "Japan",
     "iso3166_1": "JP",
     "latitude": 34.6937,
-    "longitude": 135.5023
+    "longitude": 135.5023,
+    "local_languages": ["ja"]
   },
   {
     "city": "Kyoto",
@@ -159,7 +176,8 @@
     "country": "Japan",
     "iso3166_1": "JP",
     "latitude": 35.0116,
-    "longitude": 135.7681
+    "longitude": 135.7681,
+    "local_languages": ["ja"]
   },
   {
     "city": "Sapporo",
@@ -168,7 +186,8 @@
     "country": "Japan",
     "iso3166_1": "JP",
     "latitude": 43.0618,
-    "longitude": 141.3545
+    "longitude": 141.3545,
+    "local_languages": ["ja"]
   },
   {
     "city": "Seoul",
@@ -177,7 +196,8 @@
     "country": "South Korea",
     "iso3166_1": "KR",
     "latitude": 37.5665,
-    "longitude": 126.978
+    "longitude": 126.978,
+    "local_languages": ["ko"]
   },
   {
     "city": "Busan",
@@ -186,7 +206,8 @@
     "country": "South Korea",
     "iso3166_1": "KR",
     "latitude": 35.1796,
-    "longitude": 129.0756
+    "longitude": 129.0756,
+    "local_languages": ["ko"]
   },
   {
     "city": "Ho Chi Minh City",
@@ -195,7 +216,8 @@
     "country": "Vietnam",
     "iso3166_1": "VN",
     "latitude": 10.8231,
-    "longitude": 106.6297
+    "longitude": 106.6297,
+    "local_languages": ["vi"]
   },
   {
     "city": "Hanoi",
@@ -204,7 +226,8 @@
     "country": "Vietnam",
     "iso3166_1": "VN",
     "latitude": 21.0285,
-    "longitude": 105.8542
+    "longitude": 105.8542,
+    "local_languages": ["vi"]
   },
   {
     "city": "Da Nang",
@@ -213,7 +236,8 @@
     "country": "Vietnam",
     "iso3166_1": "VN",
     "latitude": 16.0544,
-    "longitude": 108.2022
+    "longitude": 108.2022,
+    "local_languages": ["vi"]
   },
   {
     "city": "Bangkok",
@@ -222,7 +246,8 @@
     "country": "Thailand",
     "iso3166_1": "TH",
     "latitude": 13.7563,
-    "longitude": 100.5018
+    "longitude": 100.5018,
+    "local_languages": ["th"]
   },
   {
     "city": "Taipei",
@@ -231,7 +256,8 @@
     "country": "Taiwan",
     "iso3166_1": "TW",
     "latitude": 25.033,
-    "longitude": 121.5654
+    "longitude": 121.5654,
+    "local_languages": ["zh-TW"]
   },
   {
     "city": "Beijing",
@@ -240,7 +266,8 @@
     "country": "China",
     "iso3166_1": "CN",
     "latitude": 39.9042,
-    "longitude": 116.4074
+    "longitude": 116.4074,
+    "local_languages": ["zh-CN"]
   },
   {
     "city": "Shanghai",
@@ -249,7 +276,8 @@
     "country": "China",
     "iso3166_1": "CN",
     "latitude": 31.2304,
-    "longitude": 121.4737
+    "longitude": 121.4737,
+    "local_languages": ["zh-CN"]
   },
   {
     "city": "Bengaluru",
@@ -258,7 +286,8 @@
     "country": "India",
     "iso3166_1": "IN",
     "latitude": 12.9716,
-    "longitude": 77.5946
+    "longitude": 77.5946,
+    "local_languages": ["kn", "en"]
   },
   {
     "city": "Singapore",
@@ -267,7 +296,8 @@
     "country": "Singapore",
     "iso3166_1": "SG",
     "latitude": 1.3521,
-    "longitude": 103.8198
+    "longitude": 103.8198,
+    "local_languages": ["en", "zh", "ms", "ta"]
   },
   {
     "city": "Melbourne",
@@ -276,7 +306,8 @@
     "country": "Australia",
     "iso3166_1": "AU",
     "latitude": -37.8136,
-    "longitude": 144.9631
+    "longitude": 144.9631,
+    "local_languages": ["en-AU"]
   },
   {
     "city": "Sydney",
@@ -285,7 +316,8 @@
     "country": "Australia",
     "iso3166_1": "AU",
     "latitude": -33.8688,
-    "longitude": 151.2093
+    "longitude": 151.2093,
+    "local_languages": ["en-AU"]
   },
   {
     "city": "Brisbane",
@@ -294,7 +326,8 @@
     "country": "Australia",
     "iso3166_1": "AU",
     "latitude": -27.4705,
-    "longitude": 153.026
+    "longitude": 153.026,
+    "local_languages": ["en-AU"]
   },
   {
     "city": "Adelaide",
@@ -303,7 +336,8 @@
     "country": "Australia",
     "iso3166_1": "AU",
     "latitude": -34.9285,
-    "longitude": 138.6007
+    "longitude": 138.6007,
+    "local_languages": ["en-AU"]
   },
   {
     "city": "Perth",
@@ -312,7 +346,8 @@
     "country": "Australia",
     "iso3166_1": "AU",
     "latitude": -31.9505,
-    "longitude": 115.8605
+    "longitude": 115.8605,
+    "local_languages": ["en-AU"]
   },
   {
     "city": "Hobart",
@@ -321,7 +356,8 @@
     "country": "Australia",
     "iso3166_1": "AU",
     "latitude": -42.8821,
-    "longitude": 147.3272
+    "longitude": 147.3272,
+    "local_languages": ["en-AU"]
   },
   {
     "city": "Wellington",
@@ -330,7 +366,8 @@
     "country": "New Zealand",
     "iso3166_1": "NZ",
     "latitude": -41.2865,
-    "longitude": 174.7762
+    "longitude": 174.7762,
+    "local_languages": ["en", "mi"]
   },
   {
     "city": "Auckland",
@@ -339,7 +376,8 @@
     "country": "New Zealand",
     "iso3166_1": "NZ",
     "latitude": -36.8485,
-    "longitude": 174.7633
+    "longitude": 174.7633,
+    "local_languages": ["en", "mi"]
   },
   {
     "city": "Christchurch",
@@ -348,7 +386,8 @@
     "country": "New Zealand",
     "iso3166_1": "NZ",
     "latitude": -43.532,
-    "longitude": 172.6306
+    "longitude": 172.6306,
+    "local_languages": ["en", "mi"]
   },
   {
     "city": "Nelson",
@@ -357,7 +396,8 @@
     "country": "New Zealand",
     "iso3166_1": "NZ",
     "latitude": -41.2706,
-    "longitude": 173.284
+    "longitude": 173.284,
+    "local_languages": ["en", "mi"]
   },
   {
     "city": "Munich",
@@ -366,7 +406,8 @@
     "country": "Germany",
     "iso3166_1": "DE",
     "latitude": 48.1351,
-    "longitude": 11.582
+    "longitude": 11.582,
+    "local_languages": ["de"]
   },
   {
     "city": "Berlin",
@@ -375,7 +416,8 @@
     "country": "Germany",
     "iso3166_1": "DE",
     "latitude": 52.52,
-    "longitude": 13.405
+    "longitude": 13.405,
+    "local_languages": ["de"]
   },
   {
     "city": "Cologne",
@@ -384,7 +426,8 @@
     "country": "Germany",
     "iso3166_1": "DE",
     "latitude": 50.9375,
-    "longitude": 6.9603
+    "longitude": 6.9603,
+    "local_languages": ["de"]
   },
   {
     "city": "Bamberg",
@@ -393,7 +436,8 @@
     "country": "Germany",
     "iso3166_1": "DE",
     "latitude": 49.8916,
-    "longitude": 10.8916
+    "longitude": 10.8916,
+    "local_languages": ["de"]
   },
   {
     "city": "Brussels",
@@ -402,7 +446,8 @@
     "country": "Belgium",
     "iso3166_1": "BE",
     "latitude": 50.8503,
-    "longitude": 4.3517
+    "longitude": 4.3517,
+    "local_languages": ["fr", "nl"]
   },
   {
     "city": "Antwerp",
@@ -411,7 +456,8 @@
     "country": "Belgium",
     "iso3166_1": "BE",
     "latitude": 51.2194,
-    "longitude": 4.4025
+    "longitude": 4.4025,
+    "local_languages": ["nl"]
   },
   {
     "city": "Bruges",
@@ -420,7 +466,8 @@
     "country": "Belgium",
     "iso3166_1": "BE",
     "latitude": 51.2093,
-    "longitude": 3.2247
+    "longitude": 3.2247,
+    "local_languages": ["nl"]
   },
   {
     "city": "London",
@@ -429,7 +476,8 @@
     "country": "United Kingdom",
     "iso3166_1": "GB",
     "latitude": 51.5074,
-    "longitude": -0.1278
+    "longitude": -0.1278,
+    "local_languages": ["en-GB"]
   },
   {
     "city": "Bristol",
@@ -438,7 +486,8 @@
     "country": "United Kingdom",
     "iso3166_1": "GB",
     "latitude": 51.4545,
-    "longitude": -2.5879
+    "longitude": -2.5879,
+    "local_languages": ["en-GB"]
   },
   {
     "city": "Edinburgh",
@@ -447,7 +496,8 @@
     "country": "United Kingdom",
     "iso3166_1": "GB",
     "latitude": 55.9533,
-    "longitude": -3.1883
+    "longitude": -3.1883,
+    "local_languages": ["en-GB", "gd"]
   },
   {
     "city": "Glasgow",
@@ -456,7 +506,8 @@
     "country": "United Kingdom",
     "iso3166_1": "GB",
     "latitude": 55.8642,
-    "longitude": -4.2518
+    "longitude": -4.2518,
+    "local_languages": ["en-GB", "gd"]
   },
   {
     "city": "Prague",
@@ -465,7 +516,8 @@
     "country": "Czechia",
     "iso3166_1": "CZ",
     "latitude": 50.0755,
-    "longitude": 14.4378
+    "longitude": 14.4378,
+    "local_languages": ["cs"]
   },
   {
     "city": "Pilsen",
@@ -474,7 +526,8 @@
     "country": "Czechia",
     "iso3166_1": "CZ",
     "latitude": 49.7384,
-    "longitude": 13.3736
+    "longitude": 13.3736,
+    "local_languages": ["cs"]
   },
   {
     "city": "Amsterdam",
@@ -483,7 +536,8 @@
     "country": "Netherlands",
     "iso3166_1": "NL",
     "latitude": 52.3676,
-    "longitude": 4.9041
+    "longitude": 4.9041,
+    "local_languages": ["nl"]
   },
   {
     "city": "Copenhagen",
@@ -492,7 +546,8 @@
     "country": "Denmark",
     "iso3166_1": "DK",
     "latitude": 55.6761,
-    "longitude": 12.5683
+    "longitude": 12.5683,
+    "local_languages": ["da"]
   },
   {
     "city": "Warsaw",
@@ -501,7 +556,8 @@
     "country": "Poland",
     "iso3166_1": "PL",
     "latitude": 52.2297,
-    "longitude": 21.0122
+    "longitude": 21.0122,
+    "local_languages": ["pl"]
   },
   {
     "city": "Krakow",
@@ -510,7 +566,8 @@
     "country": "Poland",
     "iso3166_1": "PL",
     "latitude": 50.0647,
-    "longitude": 19.945
+    "longitude": 19.945,
+    "local_languages": ["pl"]
   },
   {
     "city": "Rome",
@@ -519,7 +576,8 @@
     "country": "Italy",
     "iso3166_1": "IT",
     "latitude": 41.9028,
-    "longitude": 12.4964
+    "longitude": 12.4964,
+    "local_languages": ["it"]
   },
   {
     "city": "Milan",
@@ -528,7 +586,8 @@
     "country": "Italy",
     "iso3166_1": "IT",
     "latitude": 45.4642,
-    "longitude": 9.19
+    "longitude": 9.19,
+    "local_languages": ["it"]
   },
   {
     "city": "Barcelona",
@@ -537,7 +596,8 @@
     "country": "Spain",
     "iso3166_1": "ES",
     "latitude": 41.3851,
-    "longitude": 2.1734
+    "longitude": 2.1734,
+    "local_languages": ["ca", "es"]
   },
   {
     "city": "Madrid",
@@ -546,7 +606,8 @@
     "country": "Spain",
     "iso3166_1": "ES",
     "latitude": 40.4168,
-    "longitude": -3.7038
+    "longitude": -3.7038,
+    "local_languages": ["es"]
   },
   {
     "city": "Paris",
@@ -555,7 +616,8 @@
     "country": "France",
     "iso3166_1": "FR",
     "latitude": 48.8566,
-    "longitude": 2.3522
+    "longitude": 2.3522,
+    "local_languages": ["fr"]
   },
   {
     "city": "Lyon",
@@ -564,7 +626,8 @@
     "country": "France",
     "iso3166_1": "FR",
     "latitude": 45.764,
-    "longitude": 4.8357
+    "longitude": 4.8357,
+    "local_languages": ["fr"]
   },
   {
     "city": "Stockholm",
@@ -573,7 +636,8 @@
     "country": "Sweden",
     "iso3166_1": "SE",
     "latitude": 59.3293,
-    "longitude": 18.0686
+    "longitude": 18.0686,
+    "local_languages": ["sv"]
   },
   {
     "city": "Gothenburg",
@@ -582,7 +646,8 @@
     "country": "Sweden",
     "iso3166_1": "SE",
     "latitude": 57.7089,
-    "longitude": 11.9746
+    "longitude": 11.9746,
+    "local_languages": ["sv"]
   },
   {
     "city": "Oslo",
@@ -591,7 +656,8 @@
     "country": "Norway",
     "iso3166_1": "NO",
     "latitude": 59.9139,
-    "longitude": 10.7522
+    "longitude": 10.7522,
+    "local_languages": ["no"]
   },
   {
     "city": "Dublin",
@@ -600,7 +666,8 @@
     "country": "Ireland",
     "iso3166_1": "IE",
     "latitude": 53.3498,
-    "longitude": -6.2603
+    "longitude": -6.2603,
+    "local_languages": ["en", "ga"]
   },
   {
     "city": "Vienna",
@@ -609,7 +676,8 @@
     "country": "Austria",
     "iso3166_1": "AT",
     "latitude": 48.2082,
-    "longitude": 16.3738
+    "longitude": 16.3738,
+    "local_languages": ["de-AT"]
   },
   {
     "city": "Zurich",
@@ -618,7 +686,8 @@
     "country": "Switzerland",
     "iso3166_1": "CH",
     "latitude": 47.3769,
-    "longitude": 8.5417
+    "longitude": 8.5417,
+    "local_languages": ["de-CH"]
   },
   {
     "city": "Tallinn",
@@ -627,7 +696,8 @@
     "country": "Estonia",
     "iso3166_1": "EE",
     "latitude": 59.437,
-    "longitude": 24.7536
+    "longitude": 24.7536,
+    "local_languages": ["et"]
   },
   {
     "city": "Denver",
@@ -636,7 +706,8 @@
     "country": "United States",
     "iso3166_1": "US",
     "latitude": 39.7392,
-    "longitude": -104.9903
+    "longitude": -104.9903,
+    "local_languages": ["en-US"]
   },
   {
     "city": "Portland",
@@ -645,7 +716,8 @@
     "country": "United States",
     "iso3166_1": "US",
     "latitude": 45.5152,
-    "longitude": -122.6784
+    "longitude": -122.6784,
+    "local_languages": ["en-US"]
   },
   {
     "city": "San Diego",
@@ -654,7 +726,8 @@
     "country": "United States",
     "iso3166_1": "US",
     "latitude": 32.7157,
-    "longitude": -117.1611
+    "longitude": -117.1611,
+    "local_languages": ["en-US", "es-US"]
   },
   {
     "city": "Asheville",
@@ -663,7 +736,8 @@
     "country": "United States",
     "iso3166_1": "US",
     "latitude": 35.5951,
-    "longitude": -82.5515
+    "longitude": -82.5515,
+    "local_languages": ["en-US"]
   },
   {
     "city": "Grand Rapids",
@@ -672,7 +746,8 @@
     "country": "United States",
     "iso3166_1": "US",
     "latitude": 42.9634,
-    "longitude": -85.6681
+    "longitude": -85.6681,
+    "local_languages": ["en-US"]
   },
   {
     "city": "Chicago",
@@ -681,7 +756,8 @@
     "country": "United States",
     "iso3166_1": "US",
     "latitude": 41.8781,
-    "longitude": -87.6298
+    "longitude": -87.6298,
+    "local_languages": ["en-US", "es-US"]
   },
   {
     "city": "Seattle",
@@ -690,7 +766,8 @@
     "country": "United States",
     "iso3166_1": "US",
     "latitude": 47.6062,
-    "longitude": -122.3321
+    "longitude": -122.3321,
+    "local_languages": ["en-US"]
   },
   {
     "city": "Austin",
@@ -699,7 +776,8 @@
     "country": "United States",
     "iso3166_1": "US",
     "latitude": 30.2672,
-    "longitude": -97.7431
+    "longitude": -97.7431,
+    "local_languages": ["en-US", "es-US"]
   },
   {
     "city": "Boston",
@@ -708,7 +786,8 @@
     "country": "United States",
     "iso3166_1": "US",
     "latitude": 42.3601,
-    "longitude": -71.0589
+    "longitude": -71.0589,
+    "local_languages": ["en-US"]
   },
   {
     "city": "Philadelphia",
@@ -717,7 +796,8 @@
     "country": "United States",
     "iso3166_1": "US",
     "latitude": 39.9526,
-    "longitude": -75.1652
+    "longitude": -75.1652,
+    "local_languages": ["en-US"]
   },
   {
     "city": "Brooklyn",
@@ -726,7 +806,8 @@
     "country": "United States",
     "iso3166_1": "US",
     "latitude": 40.6782,
-    "longitude": -73.9442
+    "longitude": -73.9442,
+    "local_languages": ["en-US", "es-US"]
   },
   {
     "city": "Milwaukee",
@@ -735,7 +816,8 @@
     "country": "United States",
     "iso3166_1": "US",
     "latitude": 43.0389,
-    "longitude": -87.9065
+    "longitude": -87.9065,
+    "local_languages": ["en-US"]
   },
   {
     "city": "Richmond",
@@ -744,7 +826,8 @@
     "country": "United States",
     "iso3166_1": "US",
     "latitude": 37.5407,
-    "longitude": -77.436
+    "longitude": -77.436,
+    "local_languages": ["en-US"]
   },
   {
     "city": "Cincinnati",
@@ -753,7 +836,8 @@
     "country": "United States",
     "iso3166_1": "US",
     "latitude": 39.1031,
-    "longitude": -84.512
+    "longitude": -84.512,
+    "local_languages": ["en-US"]
   },
   {
     "city": "St. Louis",
@@ -762,7 +846,8 @@
     "country": "United States",
     "iso3166_1": "US",
     "latitude": 38.627,
-    "longitude": -90.1994
+    "longitude": -90.1994,
+    "local_languages": ["en-US"]
   },
   {
     "city": "Tampa",
@@ -771,7 +856,8 @@
     "country": "United States",
     "iso3166_1": "US",
     "latitude": 27.9506,
-    "longitude": -82.4572
+    "longitude": -82.4572,
+    "local_languages": ["en-US", "es-US"]
   },
   {
     "city": "Minneapolis",
@@ -780,7 +866,8 @@
     "country": "United States",
     "iso3166_1": "US",
     "latitude": 44.9778,
-    "longitude": -93.265
+    "longitude": -93.265,
+    "local_languages": ["en-US"]
   },
   {
     "city": "Burlington",
@@ -789,7 +876,8 @@
     "country": "United States",
     "iso3166_1": "US",
     "latitude": 44.4759,
-    "longitude": -73.2121
+    "longitude": -73.2121,
+    "local_languages": ["en-US"]
   },
   {
     "city": "Portland",
@@ -798,7 +886,8 @@
     "country": "United States",
     "iso3166_1": "US",
     "latitude": 43.6591,
-    "longitude": -70.2568
+    "longitude": -70.2568,
+    "local_languages": ["en-US"]
   },
   {
     "city": "Atlanta",
@@ -807,7 +896,8 @@
     "country": "United States",
     "iso3166_1": "US",
     "latitude": 33.749,
-    "longitude": -84.388
+    "longitude": -84.388,
+    "local_languages": ["en-US"]
   },
   {
     "city": "Toronto",
@@ -816,7 +906,8 @@
     "country": "Canada",
     "iso3166_1": "CA",
     "latitude": 43.651,
-    "longitude": -79.347
+    "longitude": -79.347,
+    "local_languages": ["en-CA"]
   },
   {
     "city": "Vancouver",
@@ -825,7 +916,8 @@
     "country": "Canada",
     "iso3166_1": "CA",
     "latitude": 49.2827,
-    "longitude": -123.1207
+    "longitude": -123.1207,
+    "local_languages": ["en-CA"]
   },
   {
     "city": "Montreal",
@@ -834,7 +926,8 @@
     "country": "Canada",
     "iso3166_1": "CA",
     "latitude": 45.5017,
-    "longitude": -73.5673
+    "longitude": -73.5673,
+    "local_languages": ["fr-CA", "en-CA"]
   },
   {
     "city": "Calgary",
@@ -843,7 +936,8 @@
     "country": "Canada",
     "iso3166_1": "CA",
     "latitude": 51.0447,
-    "longitude": -114.0719
+    "longitude": -114.0719,
+    "local_languages": ["en-CA"]
   },
   {
     "city": "Halifax",
@@ -852,7 +946,8 @@
     "country": "Canada",
     "iso3166_1": "CA",
     "latitude": 44.6488,
-    "longitude": -63.5752
+    "longitude": -63.5752,
+    "local_languages": ["en-CA"]
   },
   {
     "city": "Mexico City",
@@ -861,7 +956,8 @@
     "country": "Mexico",
     "iso3166_1": "MX",
     "latitude": 19.4326,
-    "longitude": -99.1332
+    "longitude": -99.1332,
+    "local_languages": ["es-MX"]
   },
   {
     "city": "Tijuana",
@@ -870,7 +966,8 @@
     "country": "Mexico",
     "iso3166_1": "MX",
     "latitude": 32.5149,
-    "longitude": -117.0382
+    "longitude": -117.0382,
+    "local_languages": ["es-MX"]
   },
   {
     "city": "Monterrey",
@@ -879,7 +976,8 @@
     "country": "Mexico",
     "iso3166_1": "MX",
     "latitude": 25.6866,
-    "longitude": -100.3161
+    "longitude": -100.3161,
+    "local_languages": ["es-MX"]
   },
   {
     "city": "Guadalajara",
@@ -888,7 +986,8 @@
     "country": "Mexico",
     "iso3166_1": "MX",
     "latitude": 20.6597,
-    "longitude": -103.3496
+    "longitude": -103.3496,
+    "local_languages": ["es-MX"]
   },
   {
     "city": "Ensenada",
@@ -897,6 +996,7 @@
     "country": "Mexico",
     "iso3166_1": "MX",
     "latitude": 31.8667,
-    "longitude": -116.5964
+    "longitude": -116.5964,
+    "local_languages": ["es-MX"]
   }
 ]

pipeline/prompts/system.md

@@ -1,7 +1,3 @@
-Return only one raw JSON object as the final answer, with exactly three keys: "reasoning", "name", and "description".
-The "reasoning" key MUST be the first key in the object.
-No markdown, code fences, preamble, or extra keys.
-
 # FULL SYSTEM PROMPT
 
 You are an expert brewery copywriter, an architectural observer, and a master of zymurgy.
@@ -18,40 +14,48 @@ $$City Name$$
 
 $$Country Name$$
 
+## LOCAL LANGUAGE CODES:
+
+$$Local language codes in priority order$$
+
 ## CONTEXT:
 
-$$Information about local beer culture, history, or geography$$
+$$Information about local beer culture, history, geography, or language context$$
 
 ## CRITICAL OUTPUT FORMAT (READ CAREFULLY):
 
 ABSOLUTELY NO MARKDOWN FORMATTING. Do NOT wrap your response in json or ``` blocks.
 
-NO PREAMBLE OR POSTSCRIPT outside the JSON object. Do not say "Here is the JSON" or "Enjoy!".
+Do not add markdown, code fences, or postscript around the final JSON object. Do not say "Here is the JSON" or "Enjoy!".
 
-The JSON must contain exactly three keys ("reasoning", "name", and "description"); do not rename or add any other keys.
+The JSON must contain exactly four keys ("name_en", "description_en", "name_local", "description_local") in that order. Do not rename or add any other keys.
 
-The "reasoning" key MUST be first in the object.
+ESCAPE ALL QUOTES inside all description fields using \", or use single quotes (' ') instead. This applies equally to description_en and description_local. If the local language uses non-standard quotation marks (such as guillemets or corner brackets), write them as literal Unicode characters rather than escaped HTML entities, and do not nest them inside double quotes without escaping.
 
-ESCAPE ALL QUOTES inside the description using ", or use single quotes (' ') instead. Escaping quotes perfectly is super important to avoid errors later.
+DO NOT use actual line breaks (\n) inside any string. Keep all descriptions as one continuous string each.
 
-DO NOT use actual line breaks (\n) inside the string. Keep the description as one continuous string.
+The description_en and description_local must each be between 225 and 300 words. Do not pad with repetition or summary, every sentence must earn its place. Be concise and specific.
 
 Expected JSON format:
+
+```json
 {
-"reasoning": "Briefly plan the environmental hook, the technical brewing detail, the architectural detail, and the objective invitation.",
-"name": "Fictional Local Brewery Name",
-"description": "The description goes here."
+  "name_en": "Fictional Local Brewery Name in English",
+  "description_en": "The English description goes here.",
+  "name_local": "Translated brewery name in the local language",
+  "description_local": "The localised description goes here."
 }
+```
 
 ## CONTENT RULES AND CONSTRAINTS:
 
 ### THE HOOK:
 
-The first sentence must be an immersive, sensory environmental hook. It needs to clearly establish the weather, smells, or sounds typical of that city. Do not start by using the brewery's name or standard welcoming phrases.
+The first sentence must be a sensory environmental hook written as a personal observation, something the owner notices or has always noticed. It should establish the local weather, smell, or soundscape of the city. Do not open with the brewery's name or a generic welcome.
 
 ### GEOGRAPHIC & CULTURAL ANCHOR:
 
-The story must be deeply tied to the provided geographic and cultural info. It should mix historical brewing facts with the gritty reality of modern craft brewing, making sure it fits the local culture perfectly.
+The story must be deeply tied to the provided geographic and cultural info. Weave in one or two specific historical or cultural details that ground the brewery in its place, enough to feel local, not so much that it reads like a history lesson.
 
 ### TECHNICAL BREWING DETAIL (VARY THIS!):
 
@@ -59,15 +63,26 @@ You must include one highly specialized technical brewing detail. To avoid sound
 
 ### ARCHITECTURAL DETAIL (VARY THIS!):
 
-You must include one specific architectural or environmental detail, highlighting the building's physical wear, structure, or history. Examples include rusty steel beams, weird acoustics from an old factory, decaying brickwork, or worn-out local infrastructure. Avoid overused industry clichés like repurposed dairy equipment or glycol chillers.
+You must include one specific architectural or environmental detail, highlighting the building's physical wear, structure, or history. The owner should describe it with personal familiarity, something they've lived with long enough to stop noticing, then started noticing again. Avoid overused industry clichés like repurposed dairy equipment or glycol chillers.
 
 ### THE INVITATION:
 
-The last sentence must be an atmospheric invitation to hang out in the space, kept totally objective. Good examples include suggesting where to stand, like "Observation may commence near the foundational supports," or "Positioning adjacent to the exterior loading apparatus is suggested." Avoid regular sayings like telling people to grab a seat or ask the bartender.
+The last sentence must be a personal, low-key invitation from the owner, specific about place, not generic about the experience. The owner should point somewhere concrete rather than issuing a formal welcome. Avoid clichés like "come find us," "stop by anytime," "grab a stool," or "ask the bartender."
+
+### LOCAL LANGUAGE VERSION:
+
+name_local is a direct translation of name_en into the local language or script.
+Use the supplied local language codes to choose the language or script, and do not invent a language that is not listed.
+
+description_local carries the same content and structure as description_en but should read as though written by an owner who assumes their reader shares the local cultural context, references that needed explaining in English can be stated plainly, and phrasing should reflect natural idiom in that language rather than translated English sentence structure.
+
+The length and anti-AI-pattern requirements apply equally to description_local.
+
+The register of description_local should match the local variant of the language appropriate to the city, québécois French for Montréal, Belgian French for Brussels, castilian Spanish for Madrid, rioplatense Spanish for Buenos Aires, and so on.
 
 ### THE BLOCKLIST (FORBIDDEN CONCEPTS):
 
-You absolutely cannot use the following words and phrases because they are overused and too casual. Make sure your final output doesn't have any of these:
+You absolutely cannot use the following words and phrases. Make sure your final output doesn't have any of these:
 
 - "hidden gem"
 - "passion"
@@ -79,19 +94,41 @@ You absolutely cannot use the following words and phrases because they are overu
 - "mash temperature"
 - "grab a stool"
 - "ask the bartender"
+- "come find us"
+- "stop by anytime"
+
+#### FORBIDDEN WRITING PATTERNS
+
+The following patterns are common AI writing pitfalls and must not appear in either description:
+
+- Negative parallelism constructions: "It's not X, it's Y" or "We're not about X, we're about Y"
+- Inflated significance phrases: "stands as a testament," "plays a vital role," "leaves a lasting impact," "watershed moment," "deeply rooted," "rich cultural heritage," "rich cultural tapestry," "enduring legacy"
+- Superficial trailing analyses: sentences ending in -ing words that add opinion without content ("ensuring consistency," "reflecting the city's spirit," "highlighting our commitment")
+- Promotional travel-copy tone: "breathtaking," "must-visit," "stunning," "vibrant"
+- Overused conjunctive transitions used as sentence openers: "Moreover," "Furthermore," "In addition," "In contrast"
+- Rule of three: do not consistently organise ideas or examples in triplets
 
 ### VOICE & PERSPECTIVE:
 
-The description must be written strictly in the third-person objective. You need to act like a detached architectural observer looking at the space and the brewing process from the outside. Do not use first-person or second-person pronouns, keeping an atmosphere of academic distance and professionalism.
+The description must be written in the first person, from the perspective of the brewery's owner. Favour "we" and "our" over "I" and "my." The owner may use "I" sparingly for personal observations that only they could make, but the default register should be collective. The tone should feel lived-in and a little weathered. Do not use third-person or second-person pronouns.
 
 ## EXAMPLE:
 
 Input:
-CITY: Sapporo
-COUNTRY: Japan
-CONTEXT: Sapporo is the capital of Hokkaido, Japan's northernmost main island, with a subarctic climate: winters are severe and protracted, with the city averaging over 6 metres of cumulative snowfall per season...
+CITY: Montréal
+COUNTRY: Canada
+LOCAL LANGUAGE CODES: fr-CA, en-CA
+CONTEXT: Montréal has been brewing since 1646 when Jesuit Brother Ambroise first introduced brewing to New France. By the 19th century, Pointe-Saint-Charles became the industrial heart of the city, home to railway yards, canal workers, and a tavern on nearly every block. Molson, one of North America's oldest commercial breweries, has operated on the St. Lawrence since 1786. By the early 1980s, Molson, Labatt, and Carling controlled 96% of the Quebec beer market. The craft revival began slowly in the late 1980s and has accelerated sharply since 2002, when 33 brewing companies have grown to over 300 province-wide.
 
 $$Truncated for brevity, but assumes full context provided$$
 
 Output:
-{ "name": "Tokachi Grain & Ferment", "description": "By February, the powder snow blowing off the Teine range buries the bicycle racks on Susukino's side streets to the crossbar. Sapporo has been in the business of serious lager since 1876, but Tokachi Grain & Ferment isn't interested in replicating the macro-brew legacy. Instead, they source base malt exclusively from Obihiro-area farms and run the entire grain bill through a rigorous Burtonization protocol, driving up calcium sulfate levels to pull a sharp, mineral snap into the finish. The taproom is carved from a former Meiji-era goods shed, where a single run of oxidized copper piping bisects the ceiling and weeps green verdigris onto the communal timber table below. Observation may commence beneath the deteriorating copper, where the pale ale may be procured while the surrounding acoustics are analyzed." }
+
+```json
+{
+  "name_en": "Canal Street Grain & Ferment",
+  "description_en": "In February the wind off the Lachine Canal has a particular quality, wet and cold in a way that feels industrial, like it's been sitting in the lock chambers since the last barge went through. Pointe-Saint-Charles used to be called the neighbourhood of a hundred taverns, and you can still see the old storefronts with sealed windows and faded signage for brands that haven't existed in forty years. We started in 2019 in a former rail maintenance shed two streets from the canal. The inspection pit where mechanics used to work under locomotives is still in the floor, we covered it with plate steel, and on cold nights it hums from the temperature differential. Our house ale runs through a turbid mash borrowed loosely from Belgian lambic practice, never fully clarified before fermentation, which keeps the mouthfeel thick through a long cold secondary. It took two winters to dial in. The plate steel end of the room is where things tend to get quiet on a slow Tuesday.",
+  "name_local": "Fermentation rue du Canal",
+  "description_local": "En février, le vent du canal Lachine a quelque chose de particulier, humide et froid d'une façon qui sent le fer et le béton mouillé. La Pointe s'appelait autrefois le quartier aux cent tavernes, et on voit encore les vieilles devantures aux fenêtres condamnées, avec leurs enseignes pour des marques disparues depuis quarante ans. On a ouvert en 2019 dans un ancien hangar d'entretien ferroviaire à deux rues du canal. La fosse d'inspection où les mécaniciens travaillaient sous les locomotives est encore là dans le plancher, on l'a recouverte d'une plaque d'acier qui vibre les soirs de grand froid. Notre ale maison passe par un empâtage trouble inspiré de la pratique lambic belge, jamais complètement clarifié avant la fermentation, ce qui garde une belle rondeur en bouche après une longue garde à froid. Ça nous a pris deux hivers à stabiliser. Le bout de la salle côté plaque d'acier, c'est là que ça se calme les mardis tranquilles."
+}
+```

pipeline/src/biergarten_data_generator/log_results.cc

@@ -16,8 +16,12 @@ void BiergartenDataGenerator::LogResults() const {
         "iso3166_2={} lat={} lon={}",
         index, location.city, location.country, location.state_province,
         location.iso3166_2, location.latitude, location.longitude);
-    spdlog::info("   brewery_name=\"{}\"", brewery.name);
-    spdlog::info("   brewery_description=\"{}\"", brewery.description);
+    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;
   }
 }

pipeline/src/biergarten_data_generator/query_cities_with_countries.cc

@@ -13,7 +13,7 @@
 #include "biergarten_data_generator.h"
 #include "json_handling/json_loader.h"
 
-static constexpr size_t kBreweryAmount = 5;
+static constexpr size_t kBreweryAmount = 50;
 
 std::vector<Location> BiergartenDataGenerator::QueryCitiesWithCountries() {
   spdlog::info("\n=== GEOGRAPHIC DATA OVERVIEW ===");

pipeline/src/biergarten_data_generator/run.cc

@@ -19,7 +19,7 @@ bool BiergartenDataGenerator::Run() {
     for (auto& city : cities) {
       try {
         std::string region_context = context_service_->GetLocationContext(city);
-        spdlog::info("[Pipeline] Context for '{}' ({}) gathered:\n{}",
+        spdlog::debug("[Pipeline] Context for '{}' ({}) gathered:\n{}",
                      city.city, city.country, region_context);
 
         enriched.push_back(

pipeline/src/data_generation/llama/generate_brewery.cc

@@ -4,20 +4,38 @@
  * inference, and validates structured JSON output for brewery records.
  */
 
-#include "data_generation/llama_generator.h"
+#include <spdlog/spdlog.h>
 
 #include <format>
 #include <optional>
 #include <stdexcept>
 #include <string>
 #include <string_view>
+#include <vector>
 
-#include <spdlog/spdlog.h>
-
+#include "data_generation/llama_generator.h"
 #include "data_generation/llama_generator_helpers.h"
 
+static std::string FormatLocalLanguageCodes(
+    const std::vector<std::string>& codes) {
+  if (codes.empty()) {
+    return "Not provided";
+  }
+
+  std::string formatted;
+  for (const std::string& code : codes) {
+    if (!formatted.empty()) {
+      formatted += ", ";
+    }
+    formatted += code;
+  }
+
+  return formatted;
+}
+
 static constexpr std::string_view kBreweryJsonGrammar = R"json_brewery(
-root ::= ws "{" ws "\"reasoning\"" ws ":" ws string ws "," ws "\"name\"" ws ":" ws string ws "," ws "\"description\"" ws ":" ws string ws "}" ws
+root ::= thought-block "{" ws "\"name_en\"" ws ":" ws string ws "," ws "\"description_en\"" ws ":" ws string ws "," ws "\"name_local\"" ws ":" ws string ws "," ws "\"description_local\"" ws ":" ws string ws "}" ws
+thought-block ::= [^{]*
 ws ::= [ \t\n\r]*
 string ::= "\"" char+ "\""
 char ::= [^"\\\x7F\x00-\x1F] | [\\] escape
@@ -34,8 +52,10 @@ BreweryResult LlamaGenerator::GenerateBrewery(
   /**
    * Preprocess and truncate region context to manageable size
    */
-  const std::string safe_region_context =
-      PrepareRegionContext(region_context);
+  const std::string safe_region_context = PrepareRegionContext(region_context);
+
+  const std::string local_language_codes =
+      FormatLocalLanguageCodes(location.local_languages);
 
   const std::string country_suffix =
       location.country.empty() ? std::string{}
@@ -47,16 +67,18 @@ BreweryResult LlamaGenerator::GenerateBrewery(
   const std::string system_prompt =
       LoadBrewerySystemPrompt("prompts/system.md");
 
-
   std::string user_prompt = std::format(
-      "## CITY:\n{}\n\n## COUNTRY:\n{}\n\n## CONTEXT:\n{}",
-      location.city, location.country, safe_region_context);
+      "## CITY:\n{}\n\n## COUNTRY:\n{}\n\n## LOCAL LANGUAGE CODES:\n{}\n\n## "
+      "CONTEXT:\n{}",
+      location.city, location.country, local_language_codes,
+      safe_region_context);
 
   /**
    * Store location context for retry prompts (without repeating full context)
    */
   const std::string retry_location =
-      std::format("Location: {}{}", location.city, country_suffix);
+      std::format("Location: {}{}\nLocal language codes: {}", location.city,
+                  country_suffix, local_language_codes);
 
   /**
    * RETRY LOOP with validation and error correction
@@ -67,33 +89,32 @@ BreweryResult LlamaGenerator::GenerateBrewery(
   std::string raw;
   std::string last_error;
 
-    // Token budget: too small risks truncating valid JSON mid-string.
-    // Start conservatively but allow adaptive increases on truncation.
-    int max_tokens = kBreweryInitialMaxTokens;
+  // Token budget: too small risks truncating valid JSON mid-string.
+  // Start conservatively but allow adaptive increases on truncation.
+  int max_tokens = kBreweryInitialMaxTokens;
 
   // Limit output length to keep it concise and focused
   for (int attempt = 0; attempt < max_attempts; ++attempt) {
     // Generate brewery data from LLM
-    raw = this->Infer(system_prompt, user_prompt, max_tokens, kBreweryJsonGrammar);
-    spdlog::debug("LlamaGenerator: raw output (attempt {}): {}", attempt + 1,
-                  raw);
+    raw = this->Infer(system_prompt, user_prompt, max_tokens,
+                      kBreweryJsonGrammar);
+    spdlog::info("LlamaGenerator: raw output (attempt {}): {}", attempt + 1,
+                 raw);
 
     // Validate output: parse JSON and check required fields
 
-    std::string name;
-    std::string description;
-    std::string reasoning;
+    BreweryResult brewery;
     const std::optional<std::string> validation_error =
-        ValidateBreweryJson(raw, name, description, reasoning);
+        ValidateBreweryJson(raw, brewery);
+
     if (!validation_error.has_value()) {
       // Success: return parsed brewery data
 
       spdlog::info(
-          "LlamaGenerator: successfully generated brewery data on attempt {}:\n reasoning='{}',\n name='{}',\n description='{}'",
-          attempt + 1, reasoning, name, description);
+          "LlamaGenerator: successfully generated brewery data on attempt {}",
+          attempt + 1);
 
-      return BreweryResult{.name = std::move(name),
-                           .description = std::move(description)};
+      return brewery;
     }
 
     // Validation failed: log error and prepare corrective feedback
@@ -102,28 +123,29 @@ BreweryResult LlamaGenerator::GenerateBrewery(
     spdlog::warn("LlamaGenerator: malformed brewery JSON (attempt {}): {}",
                  attempt + 1, *validation_error);
 
-
     if (last_error == "JSON parse error: incomplete JSON") {
       const int previous_max_tokens = max_tokens;
-            max_tokens = std::min(max_tokens + kBreweryTruncationRetryTokenBump,
-                                                        kBreweryMaxTokensCeiling);
+      max_tokens = std::min(max_tokens + kBreweryTruncationRetryTokenBump,
+                            kBreweryMaxTokensCeiling);
       spdlog::info(
-          "LlamaGenerator: detected truncated JSON; increasing max_tokens from {} to {} and retrying",
+          "LlamaGenerator: detected truncated JSON; increasing max_tokens from "
+          "{} to {} and retrying",
           previous_max_tokens, max_tokens);
 
-
       continue;
     }
 
     // Update prompt with error details to guide LLM toward correct output.
     user_prompt = std::format(
-        R"(Your previous response was invalid. Error: {}
-Return ONLY valid JSON with exactly these keys, in this exact order: {{"reasoning": "<brief planning summary>", "name": "<brewery name>", "description": "<single-paragraph description>"}}.
-Do not include markdown, comments, extra keys, or literal placeholder values.
-
-Keep the JSON strings concise enough to fit within the token budget.
-
-{})",
+        "Your previous response was invalid. Error: {}\nReturn the thought "
+        "process before the JSON if needed, then return ONLY valid JSON with "
+        "exactly these keys, in this exact order: {{\"name_en\": \"<English "
+        "brewery name>\", \"description_en\": \"<English single-paragraph "
+        "description>\", \"name_local\": \"<local-language brewery name>\", "
+        "\"description_local\": \"<local-language single-paragraph "
+        "description>\"}}.\nDo not include markdown, comments, extra keys, or "
+        "literal placeholder values.\n\nKeep the JSON strings concise enough "
+        "to fit within the token budget.\n\n{}",
         *validation_error, retry_location);
   }
 

pipeline/src/data_generation/llama/helpers.cc

@@ -115,90 +115,101 @@ 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,
-                                               std::string& name_out,
-                                               std::string& description_out,
-                                               std::string& reasoning_out) {
-  auto validate_object = [&](const boost::json::value& json_value,
-                             std::string& error_out) -> bool {
-    if (!json_value.is_object()) {
-      error_out = "JSON root must be an object";
-      return false;
-    }
-
-
-    const auto& obj = json_value.get_object();
-
-    if (!obj.contains("reasoning") || !obj.at("reasoning").is_string()) {
-      error_out = "JSON field 'reasoning' is missing or not a string";
-      return false;
-    }
-
-    if (!obj.contains("name") || !obj.at("name").is_string()) {
-      error_out = "JSON field 'name' is missing or not a string";
-      return false;
-    }
-
-    if (!obj.contains("description") || !obj.at("description").is_string()) {
-      error_out = "JSON field 'description' is missing or not a string";
-      return false;
-    }
-    const auto& reasoning_value = obj.at("reasoning").as_string();
-    reasoning_out = Trim(std::string_view(reasoning_value.data(), reasoning_value.size()));
-    if (reasoning_out.empty()) {
-      error_out = "JSON field 'reasoning' must not be empty";
-      return false;
-    }
-
-    const auto& name_value = obj.at("name").as_string();
-    const auto& description_value = obj.at("description").as_string();
-    name_out = Trim(std::string_view(name_value.data(), name_value.size()));
-    description_out = Trim(
-        std::string_view(description_value.data(), description_value.size()));
-
-    if (name_out.empty()) {
-      error_out = "JSON field 'name' must not be empty";
-      return false;
-    }
-
-    if (description_out.empty()) {
-      error_out = "JSON field 'description' must not be empty";
-      return false;
-    }
-
-    std::string name_lower = name_out;
-    std::string description_lower = description_out;
-
-
-    auto string_to_lower = [](std::string& str_out) {
-       std::ranges::transform(str_out, str_out.begin(),
-                             [](unsigned char character) {
-                               return static_cast<char>(std::tolower(character));
-                             });
-    };
-
-    string_to_lower(name_lower);
-    string_to_lower(description_lower);
-
-    if (name_lower == "string" || description_lower == "string") {
-      error_out = "JSON appears to be a schema placeholder, not content";
-      return false;
-    }
-
-    error_out.clear();
-    return true;
-  };
-
+                                               BreweryResult& brewery_out) {
   boost::system::error_code error_code;
-  boost::json::value json_value = boost::json::parse(raw, error_code);
-  std::string validation_error;
+  const std::string_view raw_view(raw);
+  const size_t opening_brace = raw_view.find('{');
+  if (opening_brace == std::string_view::npos) {
+    return "JSON parse error: missing opening brace '{'";
+  }
+
+  const std::string_view json_payload = raw_view.substr(opening_brace);
+  boost::json::value json_value = boost::json::parse(json_payload, error_code);
   if (error_code) {
     return "JSON parse error: " + error_code.message();
   }
 
-  if (!validate_object(json_value, validation_error)) {
+  if (!json_value.is_object()) {
+    return "JSON root must be an object";
+  }
+
+  const auto& obj = json_value.get_object();
+  if (obj.size() != 4) {
+    return "JSON object must contain exactly four keys";
+  }
+
+  std::string validation_error;
+  if (!ReadRequiredTrimmedStringField(obj, "name_en", brewery_out.name_en,
+                                      &validation_error)) {
     return validation_error;
   }
 
+  if (!ReadRequiredTrimmedStringField(obj, "description_en",
+                                      brewery_out.description_en,
+                                      &validation_error)) {
+    return validation_error;
+  }
+
+  if (!ReadRequiredTrimmedStringField(obj, "name_local",
+                                      brewery_out.name_local,
+                                      &validation_error)) {
+    return validation_error;
+  }
+
+  if (!ReadRequiredTrimmedStringField(obj, "description_local",
+                                      brewery_out.description_local,
+                                      &validation_error)) {
+    return validation_error;
+  }
+
+  const std::array<std::string*, 4> schema_placeholders = {
+      &brewery_out.name_en, &brewery_out.description_en,
+      &brewery_out.name_local, &brewery_out.description_local};
+  if (HasSchemaPlaceholder(schema_placeholders)) {
+    return "JSON appears to be a schema placeholder, not content";
+  }
+
   return std::nullopt;
 }

pipeline/src/data_generation/llama/llama_generator.cc

@@ -32,7 +32,7 @@ void LlamaGenerator::ContextDeleter::operator()(
 
 LlamaGenerator::LlamaGenerator(const ApplicationOptions& options,
                                const std::string& model_path,
-                               std::shared_ptr<IPromptFormatter> prompt_formatter)
+                               std::unique_ptr<IPromptFormatter> prompt_formatter)
     : rng_(std::random_device{}()),
       prompt_formatter_(std::move(prompt_formatter)) {
   if (model_path.empty()) {

pipeline/src/data_generation/mock/generate_brewery.cc

@@ -36,7 +36,9 @@ BreweryResult MockGenerator::GenerateBrewery(
                   state_suffix, country_suffix);
 
   return {
-      .name = name,
-      .description = description,
+      .name_en = name,
+      .description_en = description,
+      .name_local = name,
+      .description_local = description,
   };
 }

pipeline/src/json_handling/json_loader.cc

@@ -35,6 +35,27 @@ static double ReadRequiredNumber(const boost::json::object& object,
   return value->to_number<double>();
 }
 
+static std::vector<std::string> ReadRequiredStringArray(
+    const boost::json::object& object, const char* key) {
+  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);
+  }
+
+  const auto& array = value->as_array();
+  std::vector<std::string> items;
+  items.reserve(array.size());
+  for (const auto& item : array) {
+    if (!item.is_string()) {
+      throw std::runtime_error(std::string("Missing or invalid string array field: ") +
+                               key);
+    }
+    items.emplace_back(item.as_string());
+  }
+  return items;
+}
+
 std::vector<Location> JsonLoader::LoadLocations(
     const std::filesystem::path& filepath) {
   std::ifstream input(filepath);
@@ -76,6 +97,8 @@ std::vector<Location> JsonLoader::LoadLocations(
         .iso3166_2 = ReadRequiredString(object, "iso3166_2"),
         .country = ReadRequiredString(object, "country"),
         .iso3166_1 = ReadRequiredString(object, "iso3166_1"),
+        .local_languages =
+            ReadRequiredStringArray(object, "local_languages"),
         .latitude = ReadRequiredNumber(object, "latitude"),
         .longitude = ReadRequiredNumber(object, "longitude"),
     });