Default Configuration Options for Litho: A Complete Guide to deepwiki-rs Settings

Litho operates with sensible defaults defined in src/config.rs, allowing immediate use without a litho.toml file by automatically configuring project paths, LLM parameters, caching behavior, and file filtering rules.

When working with the sopaco/deepwiki-rs repository, understanding the default configuration options for Litho is essential for effective documentation generation. The tool centers around a Config struct that implements the Default trait, supplying out-of-the-box values for every configurable parameter. These defaults activate automatically when users run Litho without providing a custom litho.toml configuration file.

Core Configuration Defaults

The primary Config struct in src/config.rs defines the following default values:

• project_name: None (Litho infers the name from the repository via get_project_name)
• project_path: "." (current working directory)
• output_path: "./litho.docs" (generated documentation directory)
• internal_path: "./.litho" (hidden working directory for intermediate files)
• target_language: English (TargetLanguage::English) as defined in src/i18n.rs
• analyze_dependencies: true
• identify_components: true
• max_depth: 10 (maximum recursion depth for codebase traversal)
• core_component_percentage: 20.0 (percentage of files considered core components)
• max_file_size: 64 KB (files larger than this are skipped)
• include_tests: false
• include_hidden: false

File Filtering Defaults

Litho excludes specific directories, files, and extensions by default:

Excluded directories (excluded_dirs):

[".litho", "litho.docs", "target", "node_modules", ".git", 
 "build", "dist", "venv", ".svelte-kit", "__pycache__", 
 "__tests__", "__mocks__", "__fixtures__"]

Excluded files (excluded_files):

["litho.toml", "*.litho", "*.log", "*.tmp", "*.cache", 
 "bun.lock", "package-lock.json", "yarn.lock", "pnpm-lock.yaml", 
 "Cargo.lock", ".gitignore", "*.tpl", "*.md", "*.txt", ".env"]

Excluded extensions (excluded_extensions):

["jpg", "jpeg", "png", "gif", "bmp", "ico", "mp3", "mp4", 
 "avi", "pdf", "zip", "tar", "exe", "dll", "so", "archive"]

LLM Configuration Defaults

The LLMConfig struct, also defined in src/config.rs, provides these defaults:

• provider: LLMProvider::default() (typically Ollama or ModelScope)
• api_key: Reads from LITHO_LLM_API_KEY environment variable, falls back to empty string
• api_base_url: "https://api-inference.modelscope.cn/v1"
• model_efficient: "ZhipuAI/GLM-5" (used for routine inference)
• model_powerful: "Qwen/Qwen3-Next-80B-A3B-Instruct" (used for complex tasks or fallback)
• max_tokens: 131072
• temperature: 0.1
• retry_attempts: 3
• retry_delay_ms: 5000
• timeout_seconds: 300
• disable_preset_tools: false
• max_parallels: 3 (parallel LLM calls)

Cache and Chunking Defaults

Cache Configuration

The CacheConfig struct controls intermediate storage:

• enabled: true
• cache_dir: ".litho/cache"
• expire_hours: 8760 (approximately one year)

Chunking Configuration

For processing large documentation files, ChunkingConfig provides:

• enabled: true
• max_chunk_size: 8000 characters (approximately 2000 tokens)
• chunk_overlap: 200 characters
• strategy: "semantic" (section-based chunking)
• min_size_for_chunking: 10000 characters

How to Use Default Configuration in Code

You can instantiate the default configuration programmatically using the Default trait implementation:

use deepwiki_rs::config::Config;

// Create a configuration with all defaults
let default_cfg = Config::default();

println!("Output directory: {}", default_cfg.output_path.display());
println!("LLM efficient model: {}", default_cfg.llm.model_efficient);
println!("Cache enabled: {}", default_cfg.cache.enabled);

This approach yields a fully populated configuration ready for immediate use without requiring a litho.toml file on disk.

Summary

• Litho requires no configuration file to run, using sensible defaults defined in src/config.rs for all parameters.
• File filtering automatically excludes common build directories, lock files, and binary formats to focus analysis on source code.
• LLM defaults target Chinese model providers (ModelScope, ZhipuAI, Qwen) with conservative retry and timeout settings.
• Cache persistence lasts one year by default, storing intermediate data in .litho/cache.
• Chunking processes files larger than 10,000 characters using semantic section-based splitting with 8,000 character chunks.

Frequently Asked Questions

What happens if I don't provide a litho.toml file?

Litho automatically uses the Default implementation of the Config struct defined in src/config.rs. Every setting—from output paths to LLM parameters—receives a sensible default value, allowing the tool to execute immediately without user configuration.

Which LLM models does Litho use by default?

According to the LLMConfig defaults in src/config.rs, Litho uses "ZhipuAI/GLM-5" for efficient routine inference and "Qwen/Qwen3-Next-80B-A3B-Instruct" for complex tasks or fallback scenarios. Both default to the ModelScope API endpoint at https://api-inference.modelscope.cn/v1.

How long does the cache persist by default?

The CacheConfig struct sets expire_hours to 8760 by default, which equals approximately one year. Cached data stores in the .litho/cache directory relative to the project root, and caching remains enabled (enabled: true) unless explicitly disabled.

Can I override specific defaults without creating a full config file?

While the source code in src/config.rs provides the Default trait for initialization, runtime overrides typically require a litho.toml file or environment variables such as LITHO_LLM_API_KEY. The codebase does not currently expose command-line flags for individual configuration parameters beyond the standard CLI interface defined in src/cli.rs.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}