Skip to content

cli-reference.md

Latest commit

 

History

History
149 lines (106 loc) · 4.15 KB

cli-reference.md

File metadata and controls

149 lines (106 loc) · 4.15 KB

Agents Backend CLI Reference

The Agents Backend CLI (agents-backend) is a powerful command-line tool for managing and testing all Agents Backend functionality.

Installation

# Install dependencies
pip install -r requirements.txt

# The agents-backend command becomes available after installing the package
pip install -e .

# Verify installation
agents-backend --help

Complete Command Reference

Core Commands

Command Description Example
agents-backend run Start the FastAPI server agents-backend run --host 0.0.0.0 --port 8000
agents-backend run --no-reload Production mode (no auto-reload) agents-backend run --no-reload --workers 4
agents-backend init Initialize database (drop & recreate tables, run SQL scripts) agents-backend init
agents-backend agent Run the Deep Research Agent interactively agents-backend agent --debug
agents-backend --sql <path> Execute a SQL script in a transaction agents-backend --sql ./scripts/init.sql

Agent Command Options

Option Default Description
--debug False Enable debug logging
--max-plan-iterations 1 Maximum number of plan revision cycles
--max-step-num 3 Maximum steps in each research plan
--enable-background-investigation True Run web search before planning
--enable-clarification False Enable multi-turn Q&A clarification
--max-clarification-rounds None Limit clarification rounds

Server Command Options

Option Default Description
--host 127.0.0.1 Host IP address
--port 8000 Port number
--no-reload False Disable auto-reload
--workers 1 Number of worker processes (use with --no-reload)

Celery Commands

Command Description Example
agents-backend celery worker Start Celery background worker agents-backend celery worker -l info
agents-backend celery beat Start Celery scheduler agents-backend celery beat -l info
agents-backend celery flower Start Celery monitoring UI agents-backend celery flower --port 8555

Code Generation

Command Description
agents-backend codegen Generate CRUD code from database tables (interactive)
agents-backend codegen import Import table schema for code generation

Quick Examples

Test the Deep Research Agent

# Interactive mode - asks for language and question
fba agent

# With debug logging
fba agent --debug

# With custom parameters
fba agent --max-step-num 5 --enable-clarification

What happens:

  1. Select language (English / 中文)
  2. Choose a built-in question or ask your own
  3. Watch the agent research, plan, and generate a report

Start Development Server

# Development mode with auto-reload
fba run

# Listen on all interfaces
fba run --host 0.0.0.0

# Production mode with multiple workers
fba run --no-reload --workers 4

Initialize Fresh Database

# WARNING: This drops all tables!
fba init

This will:

  1. Drop existing database tables
  2. Clear Redis caches
  3. Create fresh tables
  4. Execute all SQL initialization scripts
  5. Install plugin SQL scripts

Running Without fba in PATH

If fba command is not in your PATH, use Python directly:

# Using Python to invoke CLI
python -c "import sys; sys.argv=['fba', 'agent']; from backend.cli import main; main()"

# Or run as module
python -m backend.cli agent

Environment Variables

The CLI respects these environment variables (set in backend/.env):

Variable Description
OPENAI_API_KEY Required for LLM operations
TAVILY_API_KEY Required for web search
DATABASE_URL PostgreSQL connection string
REDIS_URL Redis connection string

Related Documentation