Skip to content

CLI Commands Reference

Complete reference for all IdeaWeaver CLI commands and their options.

Overview

Usage: ideaweaver [OPTIONS] COMMAND [ARGS]...

  IdeaWeaver Model Training CLI - A comprehensive tool for AI model training,
  evaluation, and deployment.

  Features include LoRA/QLoRA fine-tuning, RAG systems, MCP integration, and
  enterprise-grade model management. For detailed documentation and examples,
  visit: https://github.com/ideaweaver-ai-code/ideaweaver

Options:
  --help  Show this message and exit.

Commands:
  agent       Intelligent agent workflows for creative and analytical tasks.
  download    Download a model from Hugging Face Hub
  evaluate    Evaluate a model using lm-evaluation-harness with...
  finetune    Supervised fine-tuning commands with LoRA, QLoRA, and full...
  list-tasks  List all available evaluation tasks from lm-evaluation-harness
  mcp         Model Context Protocol (MCP) integration commands
  rag         RAG (Retrieval-Augmented Generation) commands
  train       Train a model with AutoTrain Advanced.
  validate    Validate a configuration file

Model Management

download

Download models from Hugging Face Hub.

ideaweaver download [OPTIONS] MODEL_ID

Options:

  • --revision TEXT: Model revision to download (default: main)
  • --cache-dir PATH: Custom cache directory for models
  • --token TEXT: Hugging Face API token
  • --help: Show help and exit

Examples: 

# Download a model
ideaweaver download microsoft/DialoGPT-medium

# Download specific revision
ideaweaver download microsoft/DialoGPT-medium --revision v1.0

# Download with custom cache
ideaweaver download microsoft/DialoGPT-medium --cache-dir ./models

quantize

Optimize models through quantization for faster inference.

ideaweaver quantize [OPTIONS]

Options: - --model PATH: Path to model directory or Hugging Face model ID - --method TEXT: Quantization method (int8, int4, fp16) - --output PATH: Output directory for quantized model - --config PATH: Quantization configuration file - --help: Show help and exit

Examples: 

# Basic int8 quantization
ideaweaver quantize --model microsoft/DialoGPT-medium --method int8

# Custom output directory
ideaweaver quantize --model ./my-model --method int4 --output ./quantized

Training & Fine-tuning

train

Train models with flexible configuration options.

ideaweaver train [OPTIONS]

Options: - --config PATH: Training configuration YAML file - --model TEXT: Base model name or path - --dataset TEXT: Dataset name or path - --output-dir PATH: Output directory for trained model - --epochs INTEGER: Number of training epochs - --batch-size INTEGER: Training batch size - --learning-rate FLOAT: Learning rate - --help: Show help and exit

Examples: 

# Train with configuration file
ideaweaver train --config config/training.yaml

# Quick training with CLI options
ideaweaver train --model bert-base-uncased --dataset ./data.csv --epochs 3

finetune

Supervised fine-tuning with LoRA, QLoRA, and full parameter methods.

finetune lora

LoRA (Low-Rank Adaptation) fine-tuning.

ideaweaver finetune lora [OPTIONS]

Options: - --model TEXT: Base model for fine-tuning - --dataset TEXT: Training dataset - --output-dir PATH: Output directory - --rank INTEGER: LoRA rank (default: 16) - --alpha INTEGER: LoRA alpha parameter (default: 32) - --dropout FLOAT: LoRA dropout rate (default: 0.1) - --epochs INTEGER: Number of epochs (default: 3) - --help: Show help and exit

Examples: 

# Basic LoRA fine-tuning
ideaweaver finetune lora --model microsoft/DialoGPT-medium --dataset ./data

# Custom LoRA parameters
ideaweaver finetune lora --model bert-base --dataset ./data --rank 32 --alpha 64

finetune qlora

QLoRA (Quantized LoRA) for memory-efficient fine-tuning.

ideaweaver finetune qlora [OPTIONS]

Options: - Similar to LoRA with additional quantization options - --bits INTEGER: Quantization bits (4, 8) - --double-quant BOOLEAN: Use double quantization

finetune full

Full parameter fine-tuning.

ideaweaver finetune full [OPTIONS]

Options: - Standard training options without rank/alpha parameters

Evaluation & Benchmarking

evaluate

Evaluate models using lm-evaluation-harness and custom benchmarks.

ideaweaver evaluate [OPTIONS]

Options: - --model PATH: Model to evaluate - --tasks TEXT: Comma-separated list of evaluation tasks - --batch-size INTEGER: Evaluation batch size - --num-fewshot INTEGER: Number of few-shot examples - --output PATH: Output file for results - --device TEXT: Device to use (cuda, cpu, auto) - --help: Show help and exit

Examples: 

# Evaluate on multiple tasks
ideaweaver evaluate --model ./my-model --tasks hellaswag,arc_easy,winogrande

# Custom few-shot evaluation
ideaweaver evaluate --model ./my-model --tasks hellaswag --num-fewshot 5

list-tasks

List all available evaluation tasks.

ideaweaver list-tasks [OPTIONS]

Options: - --category TEXT: Filter tasks by category - --search TEXT: Search tasks by name or description - --help: Show help and exit

compare

Compare multiple models on the same benchmarks.

ideaweaver compare [OPTIONS]

Options: - --models TEXT: Comma-separated list of models to compare - --tasks TEXT: Evaluation tasks for comparison - --output PATH: Output file for comparison results - --plot: Generate comparison plots - --help: Show help and exit

Examples: 

# Compare three models
ideaweaver compare --models model1,model2,model3 --tasks hellaswag,arc_easy

# Generate comparison with plots
ideaweaver compare --models model1,model2 --tasks all --plot --output comparison.html

RAG (Retrieval-Augmented Generation)

rag init

Initialize a new RAG system.

ideaweaver rag init [OPTIONS]

Options: - --vector-store TEXT: Vector store type (chroma, qdrant, faiss) - --embedding-model TEXT: Embedding model name - --llm TEXT: Language model for generation - --config PATH: RAG configuration file - --help: Show help and exit

rag add-documents

Add documents to the RAG knowledge base.

ideaweaver rag add-documents [OPTIONS]

Options: - --path PATH: Path to documents or directory - --recursive: Process directories recursively - --chunk-size INTEGER: Document chunk size - --chunk-overlap INTEGER: Overlap between chunks - --help: Show help and exit

rag query

Query the RAG system.

ideaweaver rag query [OPTIONS] QUERY

Options: - --config PATH: RAG configuration file - --top-k INTEGER: Number of retrieved documents - --temperature FLOAT: Generation temperature - --save-context: Save query context for training - --help: Show help and exit

Examples: 

# Basic RAG query
ideaweaver rag query "What is machine learning?"

# Custom retrieval parameters
ideaweaver rag query "Explain neural networks" --top-k 10 --temperature 0.7

MCP (Model Context Protocol)

mcp list-servers

List all available MCP servers.

ideaweaver mcp list-servers [OPTIONS]

mcp configure

Configure MCP server connections.

ideaweaver mcp configure [OPTIONS] SERVER_NAME

Supported Servers: - github: GitHub integration - slack: Slack workspace integration
- aws: AWS services integration - filesystem: Local filesystem access

mcp query

Query data through MCP servers.

ideaweaver mcp query [OPTIONS] SERVER_NAME QUERY

Examples: 

# Query GitHub repositories
ideaweaver mcp query github "List recent issues in my repository"

# Query Slack conversations
ideaweaver mcp query slack "Summarize today's discussions"

Llama.cpp Integration

llama convert

Convert models to GGML format for llama.cpp.

ideaweaver llama convert [OPTIONS]

Options: - --model PATH: Model to convert - --output PATH: Output GGML file - --type TEXT: Conversion type (f16, f32, q4_0, q4_1, q5_0, q5_1, q8_0) - --help: Show help and exit

llama inference

Run inference using llama.cpp.

ideaweaver llama inference [OPTIONS]

Options: - --model PATH: GGML model file - --prompt TEXT: Input prompt - --max-tokens INTEGER: Maximum generated tokens - --temperature FLOAT: Sampling temperature - --help: Show help and exit

Workflow Integration

crew

CrewAI operations for multi-agent workflows.

ideaweaver crew [OPTIONS] COMMAND [ARGS]...

Commands: - generate-storybook: Generate storybooks using multi-agent approach - setup-crew: Configure CrewAI agents - run-workflow: Execute custom CrewAI workflows

zenml

ZenML pipeline operations.

ideaweaver zenml [OPTIONS] COMMAND [ARGS]...

Commands: - init-pipeline: Initialize ZenML pipelines - run-pipeline: Execute ZenML pipelines - list-pipelines: Show available pipelines

Utilities

validate

Validate configuration files.

ideaweaver validate [OPTIONS] CONFIG_FILE

Options: - --type TEXT: Configuration type (training, rag, evaluation) - --strict: Enable strict validation - --help: Show help and exit

Examples: 

# Validate training configuration
ideaweaver validate config/training.yaml --type training

# Validate RAG setup
ideaweaver validate config/rag.yaml --type rag

Global Options

These options are available for most commands:

  • --verbose, -v: Enable verbose output
  • --quiet, -q: Suppress non-error output
  • --config PATH: Global configuration file
  • --log-level TEXT: Set logging level (DEBUG, INFO, WARNING, ERROR)
  • --no-cache: Disable caching
  • --device TEXT: Force specific device (cuda, cpu, mps)

Configuration Files

Training Configuration

# config/training.yaml
model:
  name: "microsoft/DialoGPT-medium"
  task: "text-generation"

dataset:
  name: "your-dataset"
  split: "train"

training:
  output_dir: "./results"
  num_train_epochs: 3
  per_device_train_batch_size: 4
  learning_rate: 5e-5
  save_steps: 500
  logging_steps: 100

lora:
  rank: 16
  alpha: 32
  dropout: 0.1

RAG Configuration

# config/rag.yaml
vector_store:
  type: "chroma"
  persist_directory: "./chroma_db"

embeddings:
  model: "sentence-transformers/all-MiniLM-L6-v2"

llm:
  provider: "openai"
  model: "gpt-3.5-turbo"
  temperature: 0.7

retrieval:
  top_k: 5
  similarity_threshold: 0.7

chunking:
  chunk_size: 1000
  chunk_overlap: 200

Evaluation Configuration

# config/evaluation.yaml
model: "path/to/model"
tasks:
  - "hellaswag"
  - "arc_easy"
  - "winogrande"
batch_size: 8
num_fewshot: 5
device: "auto"
output_file: "results.json"

Environment Variables

Set these environment variables for full functionality:

# API Keys
export OPENAI_API_KEY="your-openai-key"
export HUGGINGFACE_HUB_TOKEN="your-hf-token"
export ANTHROPIC_API_KEY="your-anthropic-key"

# Experiment Tracking
export WANDB_API_KEY="your-wandb-key"
export MLFLOW_TRACKING_URI="your-mlflow-uri"

# Cloud Services
export AWS_ACCESS_KEY_ID="your-aws-key"
export AWS_SECRET_ACCESS_KEY="your-aws-secret"

# MCP Integrations
export GITHUB_TOKEN="your-github-token"
export SLACK_BOT_TOKEN="your-slack-token"

Exit Codes

  • 0: Success
  • 1: General error
  • 2: Configuration error
  • 3: Model not found
  • 4: Dataset error
  • 5: Training/evaluation error
  • 6: Network/API error