Configuration

Configuration#

ML Research Tools can be configured through multiple methods with a cascading priority system.

Configuration Priority#

Configuration values are determined with the following priority (highest to lowest):

  1. Command-line arguments

  2. Configuration file

  3. Default values

Configuration File#

By default, the configuration is stored in ~/.config/ml_research_tools/config.yaml. If this file doesn’t exist, it will be created with default values when the tool is first run.

Example Configuration#

logging:
  level: INFO
redis:
  host: localhost
  port: 6379
  db: 0
  enabled: true
  ttl: 604800  # 7 days in seconds
llm:
  default: "openai"  # Default preset to use
  presets:
    openai:
      base_url: https://api.openai.com/v1
      model: gpt-3.5-turbo
      max_tokens: 8000
      temperature: 0.01
      top_p: 1.0
      retry_attempts: 3
      retry_delay: 5
      api_key: null
      tier: standard

    ollama:
      model: gemma3
      base_url: http://127.0.0.1:3333/v1/

    perplexity:
      base_url: https://api.perplexity.ai/
      model: sonar-pro
      max_tokens: 128000
      temperature: 0.01
      api_key: null
      tier: premium

Command-line Arguments#

Configuration can be overridden using command-line arguments. Global options apply to all tools:

ml_research_tools --log-level DEBUG --redis-host redis.example.com --llm-preset premium

Configuration Reference#

Usage: ml_research_tools [-h] [--config CONFIG]
                         [--log-level {DEBUG,INFO,WARNING,ERROR,CRITICAL}]
                         [--log-file LOG_FILE] [-v] [--redis-host REDIS_HOST]
                         [--redis-port REDIS_PORT] [--redis-db REDIS_DB]
                         [--redis-password REDIS_PASSWORD] [--redis-disable]
                         [--redis-recache] [--llm-preset LLM_PRESET]
                         [--llm-tier LLM_TIER] [--llm-api-key LLM_API_KEY]
                         [--llm-base-url LLM_BASE_URL] [--llm-model LLM_MODEL]
                         [--llm-max-tokens LLM_MAX_TOKENS]
                         [--llm-temperature LLM_TEMPERATURE]
                         [--llm-top-p LLM_TOP_P]
                         [--llm-retry-attempts LLM_RETRY_ATTEMPTS]
                         [--llm-retry-delay LLM_RETRY_DELAY] [--list-presets]
                         [--list-tools]
                         [--rich-capture-output RICH_CAPTURE_OUTPUT]
                         {help,ask-document,wandb-downloader,kube-pod-forward,latex-grammar}
                         ...

ML Research Tools - A collection of utilities for ML research

Options:
  -h, --help            show this help message and exit
  --list-presets        List available LLM presets and exit
  --list-tools          List available tools and exit
  --rich-capture-output RICH_CAPTURE_OUTPUT
                        Capture output in a rich format to this file

Configuration:
  --config CONFIG       Path to configuration file (default:
                        ~/.config/ml_research_tools/config.yaml)

Logging:
  --log-level {DEBUG,INFO,WARNING,ERROR,CRITICAL}
                        Logging level
  --log-file LOG_FILE   Path to log file
  -v, --verbose         Enable verbose logging

Redis:
  --redis-host REDIS_HOST
                        Redis host
  --redis-port REDIS_PORT
                        Redis port
  --redis-db REDIS_DB   Redis database number
  --redis-password REDIS_PASSWORD
                        Redis password
  --redis-disable       Disable Redis caching
  --redis-recache       Disable Redis caching retrieval, but allow saving

Llm:
  --llm-preset LLM_PRESET
                        LLM preset name to use (e.g., 'standard', 'premium')
  --llm-tier LLM_TIER   LLM tier to use (e.g., 'standard', 'premium')
  --llm-api-key LLM_API_KEY
                        API key for LLM service
  --llm-base-url LLM_BASE_URL
                        Base URL for the LLM API endpoint
  --llm-model LLM_MODEL
                        LLM model to use
  --llm-max-tokens LLM_MAX_TOKENS
                        Maximum tokens for LLM response
  --llm-temperature LLM_TEMPERATURE
                        Temperature for LLM sampling
  --llm-top-p LLM_TOP_P
                        Top-p value for LLM sampling
  --llm-retry-attempts LLM_RETRY_ATTEMPTS
                        Number of retry attempts for LLM API calls
  --llm-retry-delay LLM_RETRY_DELAY
                        Delay between retry attempts for LLM API calls
                        (seconds)

Tools:
  {help,ask-document,wandb-downloader,kube-pod-forward,latex-grammar}
                        Available tools
    help                Display help information about available tools
    ask-document        Interactive chat with document content using LLMs
    wandb-downloader    Download artifacts and runs from Weights & Biases
    kube-pod-forward    Forward port to a Kubernetes pod with a specific name
                        pattern.
    latex-grammar       Check and improve LaTeX grammar and style with LLM

Tool-Specific Configuration#

Some tools have additional configuration options that can be specified in the configuration file or as command-line arguments. See the documentation for each tool for details.

Contents