Advanced Configuration Options

Overview

RepoMaster supports extensive runtime configuration through command-line flags, allowing you to customize behavior without modifying environment files.

Command Line Interface

Basic Usage

python launcher.py [--mode MODE] [--backend-mode BACKEND_MODE] [OPTIONS]

Getting Help

# View all available options
python launcher.py --help

# List available modes
python launcher.py --modes
# or
python launcher.py --list-modes

Mode Selection

Primary Mode

--mode {frontend,backend}
-m {frontend,backend}

Default: backend Select the primary runtime mode:

frontend: Streamlit web interface
backend: Command-line agent interface

Examples:

python launcher.py --mode frontend
python launcher.py -m backend

Backend Mode

--backend-mode {unified,deepsearch,general_assistant,repository_agent}
-b {unified,deepsearch,general_assistant,repository_agent}

Default: unified Select the backend sub-mode (only applies when --mode backend):

unified: Unified multi-agent interface (recommended)
deepsearch: Deep search agent direct access
general_assistant: Programming assistant direct access
repository_agent: Repository exploration agent direct access

Examples:

python launcher.py --backend-mode unified
python launcher.py -b deepsearch

API Configuration

API Provider Type

--api-type {basic,azure_openai,openai,claude,deepseek,basic_claude4,basic_deepseek_r1}
-a {basic,azure_openai,openai,claude,deepseek,basic_claude4,basic_deepseek_r1}

Default: basic Select which API provider to use for LLM operations.

The basic type uses OpenAI-compatible configuration and is suitable for OpenAI and compatible endpoints.

Provider Options:

Option	Provider	Configuration Used
`basic`	OpenAI-compatible	`OPENAI_API_KEY`, `OPENAI_MODEL`
`openai`	OpenAI	`OPENAI_API_KEY`, `OPENAI_MODEL`
`claude`	Anthropic	`ANTHROPIC_API_KEY`, `ANTHROPIC_MODEL`
`deepseek`	DeepSeek	`DEEPSEEK_API_KEY`, `DEEPSEEK_MODEL`
`azure_openai`	Azure OpenAI	`AZURE_OPENAI_*` variables
`basic_claude4`	Claude 4	`CLAUDE_API_KEY`, model: `claude-sonnet-4-20250514`
`basic_deepseek_r1`	DeepSeek R1	`DEEPSEEK_API_KEY`, model: `deepseek-r1-0528`

Examples:

# Use OpenAI
python launcher.py --api-type openai

# Use Claude
python launcher.py -a claude

# Use DeepSeek R1 model specifically
python launcher.py --api-type basic_deepseek_r1

Model Temperature

--temperature FLOAT
-t FLOAT

Default: 0.1
Range: 0.0 - 2.0 Control the randomness of model outputs:

0.0 - 0.3: More deterministic and focused
0.4 - 0.7: Balanced creativity and consistency
0.8 - 1.0: More creative and varied
1.0+: Highly creative (GPT-5 requires exactly 1.0)

GPT-5 only supports temperature=1.0. RepoMaster automatically adjusts this setting if you use GPT-5.

Examples:

# Very deterministic for code generation
python launcher.py --temperature 0.1

# More creative for brainstorming
python launcher.py -t 0.8

Request Timeout

--timeout SECONDS

Default: 120 (2 minutes) Maximum time (in seconds) to wait for API responses. Examples:

# Longer timeout for complex tasks
python launcher.py --timeout 300

# Shorter timeout for quick operations
python launcher.py --timeout 60

Max Tokens

--max-tokens COUNT

Default: 4000 Maximum number of tokens in model responses.

Different models have different context limits. Ensure your max-tokens value is within the model’s supported range.

Examples:

# Longer responses
python launcher.py --max-tokens 8000

# Concise responses
python launcher.py --max-tokens 2000

Max Conversation Turns

--max-turns COUNT

Default: 30 Maximum number of conversation turns in a single session. Examples:

# Extended conversation
python launcher.py --max-turns 50

# Quick interactions
python launcher.py --max-turns 10

Working Directory

Custom Work Directory

--work-dir PATH
-w PATH

Default: coding (with random subdirectory) Specify where RepoMaster stores working files and execution results. Behavior (from mode_config.py:35):

Not specified or "coding": Uses {pwd}/coding/{random_string()}
Relative path: Converted to {pwd}/{path}
Absolute path: Used as-is

Examples:

# Use custom directory
python launcher.py --work-dir /tmp/repomaster

# Use relative path (converted to absolute)
python launcher.py -w my_workspace

# Default with random subdirectory
python launcher.py  # Uses: ./coding/abc123xyz/

Frontend-Specific Options

Streamlit Port

--streamlit-port PORT
-p PORT

Default: 8501 Port number for the Streamlit web interface. Examples:

python launcher.py --mode frontend --streamlit-port 8080
python launcher.py -m frontend -p 9000

Streamlit Host

--streamlit-host HOST

Default: localhost Host address for the Streamlit server. Examples:

# Listen on all interfaces
python launcher.py --mode frontend --streamlit-host 0.0.0.0

# Specific IP
python launcher.py --mode frontend --streamlit-host 192.168.1.100

Max Upload Size

--max-upload-size MB

Default: 200 (MB)
Range: 1 - 2000 (MB) Maximum file upload size for the web interface. Examples:

# Allow larger files (500MB)
python launcher.py --mode frontend --max-upload-size 500

# Restrict uploads (50MB)
python launcher.py --mode frontend --max-upload-size 50

Logging and Debugging

Log Level

--log-level {DEBUG,INFO,WARNING,ERROR}
-l {DEBUG,INFO,WARNING,ERROR}

Default: INFO Control logging verbosity:

DEBUG: Detailed debugging information
INFO: General informational messages
WARNING: Warning messages only
ERROR: Error messages only

Examples:

# Debug mode for troubleshooting
python launcher.py --log-level DEBUG

# Quiet mode - errors only
python launcher.py -l ERROR

Logging Configuration (from launcher.py:38):

logging.basicConfig(
    level=getattr(logging, log_level.upper()),
    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
)

When log level is not DEBUG, third-party library warnings are suppressed for cleaner output.

Code Execution

Docker Execution

--use-docker

Default: False Execute code in Docker containers for isolation and security.

Requires Docker to be installed and running. See Docker documentation for setup.

Examples:

# Enable Docker execution
python launcher.py --use-docker

# Combined with other options
python launcher.py --backend-mode unified --use-docker

Configuration Management

Skip Configuration Check

--skip-config-check

Default: False Skip API configuration validation on startup.

Not recommended for production use. This bypasses critical validation that ensures at least one API provider is configured.

Use Cases:

Testing with mock API responses
Development environments without API access
Automated testing pipelines

Examples:

python launcher.py --skip-config-check

Complete Examples

Production Backend Setup

python launcher.py \
  --mode backend \
  --backend-mode unified \
  --api-type openai \
  --temperature 0.1 \
  --timeout 300 \
  --max-tokens 8000 \
  --max-turns 50 \
  --work-dir /opt/repomaster/workspace \
  --log-level INFO

Development Frontend Setup

python launcher.py \
  --mode frontend \
  --streamlit-port 8080 \
  --streamlit-host 0.0.0.0 \
  --max-upload-size 500 \
  --log-level DEBUG

Deep Search Agent (Direct Access)

python launcher.py \
  --mode backend \
  --backend-mode deepsearch \
  --api-type claude \
  --temperature 0.2 \
  --timeout 180

Repository Agent with Docker

python launcher.py \
  --mode backend \
  --backend-mode repository_agent \
  --api-type deepseek \
  --use-docker \
  --work-dir /tmp/repo-tasks \
  --log-level INFO

Configuration Dataclasses

RepoMaster uses dataclasses for type-safe configuration (from mode_config.py):

RunConfig (Base)

@dataclass
class RunConfig:
    mode: str
    work_dir: str = ""
    log_level: str = "INFO"
    use_docker: bool = False
    timeout: int = 120

FrontendConfig

@dataclass
class FrontendConfig(RunConfig):
    mode: str = "frontend"
    streamlit_port: int = 8501
    streamlit_host: str = "localhost"
    file_watcher_type: str = "none"
    enable_auth: bool = True
    enable_file_browser: bool = True
    max_upload_size: int = 200  # MB

BackendConfig

@dataclass
class BackendConfig(RunConfig):
    mode: str = "backend"
    backend_mode: str = "deepsearch"
    api_type: str = "basic"
    temperature: float = 0.1
    max_tokens: int = 4000
    max_turns: int = 30

Mode-Specific Configs

Each backend mode has additional configuration options:

DeepSearchConfig

@dataclass
class DeepSearchConfig(BackendConfig):
    backend_mode: str = "deepsearch"
    enable_web_search: bool = True
    max_search_results: int = 10
    search_timeout: int = 30
    enable_code_tool: bool = True
    max_tool_messages: int = 2

GeneralAssistantConfig

@dataclass
class GeneralAssistantConfig(BackendConfig):
    backend_mode: str = "general_assistant"
    enable_venv: bool = True
    cleanup_venv: bool = False
    max_execution_time: int = 600
    supported_languages: list = [
        'python', 'javascript', 'typescript', 
        'java', 'cpp', 'go'
    ]

RepositoryAgentConfig

@dataclass
class RepositoryAgentConfig(BackendConfig):
    backend_mode: str = "repository_agent"
    enable_repository_search: bool = True
    max_repo_size_mb: int = 100
    clone_timeout: int = 300
    enable_parallel_execution: bool = True
    retry_times: int = 3

UnifiedConfig

@dataclass
class UnifiedConfig(BackendConfig):
    backend_mode: str = "unified"
    enable_web_search: bool = True
    enable_repository_search: bool = True
    enable_venv: bool = True
    cleanup_venv: bool = False
    max_search_results: int = 10
    search_timeout: int = 30
    max_execution_time: int = 600
    max_repo_size_mb: int = 100
    clone_timeout: int = 300
    retry_times: int = 3
    supported_languages: list = [
        'python', 'javascript', 'typescript',
        'java', 'cpp', 'go'
    ]

Environment vs CLI Priority

Priority Order

CLI Arguments (highest priority)
Environment Variables (.env file)
Default Values (lowest priority)

Example

# In .env file:
DEFAULT_API_PROVIDER=openai
OPENAI_MODEL=gpt-4o

# CLI overrides:
python launcher.py --api-type claude --temperature 0.5

# Result: Uses Claude provider with temperature 0.5
# (overrides env file settings)

Programmatic Configuration

For Python API usage:

from configs.mode_config import ModeConfigManager, UnifiedConfig

# Create custom configuration
config = UnifiedConfig(
    api_type='openai',
    temperature=0.2,
    work_dir='/custom/path',
    max_tokens=8000,
    timeout=300
)

# Use configuration
manager = ModeConfigManager()
manager.config = config

# Get LLM config
llm_config = manager.get_llm_config(api_type='openai')

Troubleshooting

Invalid Arguments

Error: error: unrecognized arguments: --invalid-flag Solution: Use --help to see valid arguments:

python launcher.py --help

Conflicting Options

Issue: Frontend-specific flags used with backend mode Example:

# This combination doesn't make sense
python launcher.py --mode backend --streamlit-port 8080

Solution: Frontend options only apply to frontend mode:

python launcher.py --mode frontend --streamlit-port 8080

Configuration Override Issues

Issue: Changes not taking effect Solutions:

# 1. Verify CLI arguments are correct
python launcher.py --log-level DEBUG  # See what's loaded

# 2. Check environment file isn't overriding
cat configs/.env

# 3. Use --skip-config-check to bypass env validation
python launcher.py --skip-config-check

Getting Started

Core Concepts

Agents

Usage Modes

Configuration

Use Cases

​Overview

​Command Line Interface

​Basic Usage

​Getting Help

​Mode Selection

​Primary Mode

​Backend Mode

​API Configuration

​API Provider Type

​Model Temperature

​Request Timeout

​Max Tokens

​Max Conversation Turns

​Working Directory

​Custom Work Directory

​Frontend-Specific Options

​Streamlit Port

​Streamlit Host

​Max Upload Size

​Logging and Debugging

​Log Level

​Code Execution

​Docker Execution

​Configuration Management

​Skip Configuration Check

​Complete Examples

​Production Backend Setup

​Development Frontend Setup

​Deep Search Agent (Direct Access)

​Repository Agent with Docker

​Configuration Dataclasses

​RunConfig (Base)

​FrontendConfig

​BackendConfig

​Mode-Specific Configs

​Environment vs CLI Priority

​Priority Order

​Example

​Programmatic Configuration

​Troubleshooting

​Invalid Arguments

​Conflicting Options

​Configuration Override Issues

​Next Steps

API Providers

Environment Setup

Build docs developers (and LLMs) love

Overview

Command Line Interface

Basic Usage

Getting Help

Mode Selection

Primary Mode

Backend Mode

API Configuration

API Provider Type

Model Temperature

Request Timeout

Max Tokens

Max Conversation Turns

Working Directory

Custom Work Directory

Frontend-Specific Options

Streamlit Port

Streamlit Host

Max Upload Size

Logging and Debugging

Log Level

Code Execution

Docker Execution

Configuration Management

Skip Configuration Check

Complete Examples

Production Backend Setup

Development Frontend Setup

Deep Search Agent (Direct Access)

Repository Agent with Docker

Configuration Dataclasses

RunConfig (Base)

FrontendConfig

BackendConfig

Mode-Specific Configs

Environment vs CLI Priority

Priority Order

Example

Programmatic Configuration

Troubleshooting

Invalid Arguments

Conflicting Options

Configuration Override Issues

Next Steps