Add YAML config file support to llama-cli

[devin-ai-integration]

Add YAML Configuration File Support to llama-cli

Make sure to read the contributing guidelines before submitting a PR

Overview

This PR adds support for YAML configuration files to llama-cli, allowing users to specify commonly used settings in a config file while maintaining full backward compatibility with existing command-line usage.

Link to Devin run: https://app.devin.ai/sessions/b282a23041034d3297e4b998d7e37c6c
Requested by: @jaime-leo

Key Changes

Core Implementation

• New --config CLI option: Accepts path to YAML configuration file
• common_params_load_from_yaml() function: Comprehensive YAML parsing that maps config keys to common_params structure
• Precedence handling: CLI arguments override YAML config values, which override defaults
• Full backward compatibility: All existing CLI functionality remains unchanged

Dependencies

• Added yaml-cpp library: Vendored as vendor/yaml-cpp/ following existing patterns (similar to nlohmann/json)
• Build system updates: Updated CMakeLists.txt files to include and link yaml-cpp

Testing & Documentation

• Comprehensive test suite: tests/test-yaml-config.cpp with 6 test cases covering:
  • Basic YAML config loading
  • CLI argument precedence over config
  • Invalid YAML file handling
  • Nested structure parsing (sampling, speculative, etc.)
  • Backward compatibility verification
  • Empty config file handling
• Documentation: Updated README with usage examples and config file structure
• Example config: Added examples/config.yaml with common use cases

Configuration File Structure

# Basic model configuration
model:
  path: "models/model.gguf"

# Generation parameters  
n_predict: 128
n_ctx: 4096
n_gpu_layers: 32

# Sampling parameters
sampling:
  temp: 0.8
  top_k: 40
  top_p: 0.95
  seed: 42

# Interactive settings
interactive: true
use_color: true

Testing Results

• ✅ All new YAML config tests pass
• ✅ Full test suite passes (40/40 tests, 0 failures)
• ✅ No regressions in existing functionality
• ✅ Backward compatibility verified

Human Review Checklist

⚠️ Critical Areas to Review:

1. Parameter Mapping Logic (common/arg.cpp:69-235): Complex mapping from YAML keys to common_params fields - verify all parameters are correctly handled and no fields are missed

2. CLI Precedence Logic (common/arg.cpp:1297-1300): Ensure CLI arguments properly override YAML config values in all scenarios

3. Error Handling: Review YAML parsing error handling for malformed files, missing files, and type conversion errors

4. Vendored Dependency: Assess impact of adding yaml-cpp library (~100 files, 91KB) on build time and binary size

5. Test Coverage: Verify test cases cover all major parameter types and edge cases, especially nested structures (sampling, speculative, etc.)

6. Build System Changes: Review CMakeLists.txt modifications and ensure proper linking across platforms

Lower Priority:

• Documentation accuracy and completeness
• Example config file usefulness
• Code style consistency with existing codebase

Verification Steps

1. Build with new dependency: cmake --build build --target llama-cli
2. Test basic functionality: ./llama-cli --config examples/config.yaml --help
3. Test precedence: ./llama-cli --config examples/config.yaml --temp 0.9 (should override config temp)
4. Test backward compatibility: existing CLI commands should work unchanged
5. Run test suite: make test (should pass 40/40 tests)

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring