Skip to content

CLI.md

Latest commit

 

History

History
447 lines (321 loc) · 9.35 KB

CLI.md

File metadata and controls

447 lines (321 loc) · 9.35 KB

CLI Command Reference

Complete reference for the exa-ai command-line interface.

Table of Contents

Global Options

All commands support these options:

  • --api-key KEY - Override API key from environment
  • --output-format FORMAT - json, pretty, or text (varies by command)
  • --help, -h - Show command help
  • exa-ai --version - Show version
  • exa-ai --help - Show available commands

search

Search the web using fast or deep search.

Basic Usage

exa-ai search "ruby programming"
exa-ai search "machine learning" --num-results 10
exa-ai search "AI" --type deep

Filtering Options

# Include specific domains
exa-ai search "tutorials" --include-domains "github.com,dev.to"

# Exclude specific domains
exa-ai search "programming" --exclude-domains "outdated-site.com"

# Combine filters
exa-ai search "Python" \
  --include-domains "github.com" \
  --exclude-domains "reddit.com"

Date Filtering

# Filter by published date
exa-ai search "AI research" \
  --start-published-date "2025-01-01T00:00:00.000Z" \
  --end-published-date "2025-12-31T23:59:59.999Z"

# Filter by crawl date
exa-ai search "news" \
  --start-crawl-date "2025-10-01T00:00:00.000Z" \
  --end-crawl-date "2025-10-31T23:59:59.999Z"

Text Filtering

# Must include phrase
exa-ai search "machine learning" --include-text "neural networks"

# Must exclude phrase
exa-ai search "programming" --exclude-text "paid-partnership"

# Combine inclusion and exclusion
exa-ai search "Python" \
  --include-text "open source" \
  --exclude-text "deprecated"

Content Extraction

# Extract full webpage text
exa-ai search "Ruby" --text

# With character limit
exa-ai search "AI" --text --text-max-characters 3000

# Include HTML tags in text
exa-ai search "web" --text --include-html-tags

# Generate AI summaries
exa-ai search "climate change" \
  --summary \
  --summary-query "What are the main points?"

# Format as LLM context
exa-ai search "kubernetes" \
  --context \
  --context-max-characters 5000

# Crawl subpages
exa-ai search "documentation" \
  --subpages 1 \
  --subpage-target about \
  --subpage-target docs

# Extract links
exa-ai search "web development" \
  --links 3 \
  --image-links 2

Output Options

# Pretty format (human-readable)
exa-ai search "ruby" --output-format pretty

# JSON format (default)
exa-ai search "ruby" --output-format json

Options Summary

  • QUERY - Search query (required)
  • --num-results N - Number of results (default: 10)
  • --type TYPE - Search type: fast, deep, keyword, or auto (default: fast)
  • --include-domains DOMAINS - Comma-separated domains
  • --exclude-domains DOMAINS - Comma-separated domains
  • --start-published-date DATE - ISO 8601 date string
  • --end-published-date DATE - ISO 8601 date string
  • --start-crawl-date DATE - ISO 8601 date string
  • --end-crawl-date DATE - ISO 8601 date string
  • --include-text TEXT - Phrase to include
  • --exclude-text TEXT - Phrase to exclude
  • --text - Include full webpage text
  • --text-max-characters N - Limit text length
  • --include-html-tags - Keep HTML tags in text
  • --summary - Include AI summary
  • --summary-query QUERY - Custom summary query
  • --context - Include LLM-formatted context
  • --context-max-characters N - Limit context length
  • --subpages N - Number of subpages to crawl
  • --subpage-target TARGET - Subpage URL pattern (repeatable)
  • --links N - Number of links to extract
  • --image-links N - Number of image links to extract
  • --output-format FORMAT - json or pretty

answer

Generate comprehensive answers to questions with source citations.

Basic Usage

exa-ai answer "What is the capital of France?"
exa-ai answer "Latest developments in quantum computing"

With Text Content

# Include full text from source pages
exa-ai answer "Ruby on Rails best practices" --text

Output Options

# Pretty format
exa-ai answer "How do I learn machine learning?" --output-format pretty

# JSON format (default)
exa-ai answer "What is AI?" --output-format json

Options Summary

  • QUERY - Question to answer (required)
  • --text - Include full text content from sources
  • --output-format FORMAT - json or pretty (default: json)

Response Fields

  • answer - The generated answer
  • citations - Array of source citations with URLs
  • cost_dollars - Cost of the request

context

Find relevant code snippets and documentation from repositories.

Basic Usage

exa-ai context "authentication with JWT"
exa-ai context "React hooks"
exa-ai context "async/await patterns"

Token Allocation

# Fixed token count
exa-ai context "React hooks" --tokens-num 5000

# Dynamic allocation (default)
exa-ai context "authentication" --tokens-num dynamic

Output Options

# Text format
exa-ai context "async/await" --output-format text

# JSON format (default)
exa-ai context "authentication" --output-format json

Options Summary

  • QUERY - Search query (required)
  • --tokens-num NUM - Token allocation: integer or "dynamic" (default: dynamic)
  • --output-format FORMAT - json or text (default: json)

get-contents

Retrieve full text content from web pages.

Single Page

exa-ai get-contents "https://example.com/article"

Multiple Pages

# Comma-separated URLs
exa-ai get-contents "https://site1.com,https://site2.com,https://site3.com"

With Options

# Include full text
exa-ai get-contents "https://example.com" --text

# Include highlighted sections
exa-ai get-contents "https://example.com" --highlights

# Include summary
exa-ai get-contents "https://example.com" --summary

# Combine options
exa-ai get-contents "url1,url2" \
  --text \
  --highlights \
  --summary

Output Options

# Pretty format
exa-ai get-contents "https://example.com" --output-format pretty

# JSON format (default)
exa-ai get-contents "https://example.com" --output-format json

Options Summary

  • IDS - Page IDs or URLs (required, comma-separated)
  • --text - Include full text content
  • --highlights - Include highlighted sections
  • --summary - Include summary
  • --output-format FORMAT - json or pretty (default: json)

research-start

Start a long-running research task.

Basic Usage

exa-ai research-start --instructions "Find Ruby performance tips"

With Model

exa-ai research-start \
  --instructions "Analyze AI safety papers" \
  --model gpt-4

With Output Schema

exa-ai research-start \
  --instructions "Extract key metrics" \
  --output-schema '{"format":"json","fields":["metric","value"]}'

Wait for Completion

# Block until task completes
exa-ai research-start \
  --instructions "Research AI trends" \
  --wait

# Show event log while waiting
exa-ai research-start \
  --instructions "Research AI trends" \
  --wait \
  --events

Output Options

# Pretty format
exa-ai research-start --instructions "..." --output-format pretty

# JSON format (default)
exa-ai research-start --instructions "..." --output-format json

Options Summary

  • --instructions TEXT - Research instructions (required)
  • --model MODEL - Model to use (e.g., gpt-4)
  • --output-schema SCHEMA - JSON schema for structured output
  • --wait - Block until completion
  • --events - Show event log during polling
  • --output-format FORMAT - json or pretty (default: json)

research-get

Check status of a research task.

Basic Usage

exa-ai research-get abc-123

With Event Log

exa-ai research-get abc-123 --events

Stream Results (Premium)

exa-ai research-get abc-123 --stream

Output Options

# Pretty format
exa-ai research-get abc-123 --output-format pretty

# JSON format (default)
exa-ai research-get abc-123 --output-format json

Options Summary

  • RESEARCH_ID - Task ID (required)
  • --events - Include event log
  • --stream - Stream results (premium feature)
  • --output-format FORMAT - json or pretty (default: json)

research-list

List research tasks with pagination.

Basic Usage

exa-ai research-list

Pagination

# Set page size
exa-ai research-list --limit 20

# Get next page
exa-ai research-list --limit 20 --cursor "next_cursor"

Output Options

# Pretty table format
exa-ai research-list --output-format pretty

# JSON format (default)
exa-ai research-list --output-format json

Options Summary

  • --cursor CURSOR - Pagination cursor
  • --limit N - Results per page (default: 10)
  • --output-format FORMAT - json or pretty (default: json)

Output Formats

JSON Format

Default output format returns structured JSON:

exa-ai search "ruby" --output-format json
# Returns: {"results": [...], "cost_dollars": ...}

Pretty Format

Human-readable format with formatted titles and URLs:

exa-ai search "ruby" --output-format pretty
# Returns nicely formatted output for terminal display

Text Format

Plain text output (available for context command):

exa-ai context "React" --output-format text
# Returns plain text output