LLMProc - Miscellaneous Notes & Details

This document contains additional information supplementing the main README.md. For detailed documentation, see the /docs directory. For design rationales and API decisions, see FAQ.md.

Installation Details

Full Installation Options

# Install with uv (recommended)
# Basic installation
uv pip install llmproc

# Install with development dependencies
uv pip install "llmproc[dev]"

# Install with specific provider support
uv pip install "llmproc[openai]"     # For OpenAI models
uv pip install "llmproc[anthropic]"  # For Anthropic/Claude models
uv pip install "llmproc[vertex]"     # For Google Vertex AI models
uv pip install "llmproc[gemini]"     # For Google Gemini models

# Install with all provider support
uv pip install "llmproc[all]"

# Development installation
uv sync --all-extras --all-groups

# Or with pip
pip install llmproc               # Base package
pip install "llmproc[openai]"     # For OpenAI models
pip install "llmproc[anthropic]"  # For Anthropic/Claude models
pip install "llmproc[vertex]"     # For Google Vertex AI models
pip install "llmproc[gemini]"     # For Google Gemini models
pip install "llmproc[all]"        # All providers

Environment Variables

LLMProc requires provider-specific API keys set as environment variables:

# Set API keys as environment variables
export OPENAI_API_KEY="your-key"            # For OpenAI models
export ANTHROPIC_API_KEY="your-key"         # For Claude models
export GOOGLE_API_KEY="your-key"            # For Gemini models
export ANTHROPIC_VERTEX_PROJECT_ID="id"     # For Claude on Vertex AI
export CLOUD_ML_REGION="us-central1"        # For Vertex AI (defaults to us-central1)

You can set these in your environment or include them in a .env file at the root of your project.

Key Features Reference

File Descriptor System

Handles large inputs/outputs by creating file-like references with paging support.

See file-descriptor-system.md

Program Linking

Connects multiple LLMProcess instances for collaborative problem-solving.

See program-linking.md

MCP Tool Support

Connect to external tool servers via Model Context Protocol.

See mcp-feature.md

Tool Aliases

Provides shorter, more intuitive names for tools.

See tool-aliases.md

Token Efficient Tool Use

Optimizes token usage for tool calls with Claude 3.7+.

See token-efficient-tool-use.md

Performance Considerations

• Resource Usage: Each Process instance requires memory for its state
• API Costs: Using multiple processes results in multiple API calls
• Linked Programs: Program linking creates additional processes with separate API calls
• Selective MCP Usage: MCP tools now use selective initialization for better performance

Note on compile() method: The public compile() method is intended to be used primarily when implementing program serialization/export functionality. For typical usage, the start() method handles necessary validation internally. Consider direct use of program.start() in most cases.

For more API patterns, see api/patterns.md.