CQLAI - Modern AI-Powered CQL Shell

CQLAI is a fast, portable interactive terminal for Apache Cassandra, built in Go by AxonOps. It provides a modern, user-friendly alternative to cqlsh with an advanced terminal UI, client-side command parsing, and optional AI-powered query generation.

Why CQLAI?

Feature	cqlsh	CQLAI
Language	Python (requires Python)	Go (single binary)
Dependencies	Python 3.x required	None
AI Query Generation	No	Yes (optional)
Tab Completion	Basic	Context-aware
Output Formats	Limited	Table, JSON, CSV, Parquet
Terminal UI	Basic readline	Rich TUI with mouse support
Parquet Support	No	Full import/export
Memory Management	Can OOM on large results	Virtualized, memory-bounded

Key Features

Zero Dependencies

# Download and run - that is it!
./cqlai --host cassandra.example.com

CQLAI is distributed as a single static binary for:

• Linux x86-64 and aarch64
• macOS x86-64 and arm64 (Apple Silicon)
• Windows x86-64

Rich Terminal UI

• Full-screen alternate buffer (preserves terminal history)
• Virtualized, scrollable tables for large result sets
• Vim-style keyboard navigation
• Full mouse support including wheel scrolling
• Sticky status bar with connection details and query latency
• Modal overlays for history, help, and completion

Optional AI-Powered Query Generation

Convert natural language to CQL queries:

.ai show all active users created in the last 7 days
.ai create a table for storing product inventory
.ai find orders with total greater than 1000

Supports multiple AI providers:

• OpenAI (GPT-4, GPT-3.5)
• Anthropic (Claude 3)
• Google Gemini
• Ollama (local models)
• OpenRouter
• Synthetic

AI is completely optional - CQLAI works as a full-featured CQL shell without any AI configuration.

Apache Parquet Support

Export and import data in Parquet format for analytics:

-- Export to Parquet
COPY users TO 'users.parquet';

-- Export with compression
COPY events TO 'events.parquet' WITH COMPRESSION='ZSTD';

-- Import from Parquet
COPY analytics_data FROM 'dataset.parquet';

Smart Tab Completion

Context-aware completion for:

• CQL keywords and commands
• Keyspace and table names
• Column names
• Consistency levels
• File paths

SELECT * FROM us<Tab>    -- Completes to 'users'
CONSISTENCY LOC<Tab>     -- Shows LOCAL_ONE, LOCAL_QUORUM, LOCAL_SERIAL

---

Quick Start

Installation

# Using Go
go install github.com/axonops/cqlai/cmd/cqlai@latest

# Or download from releases
# https://github.com/axonops/cqlai/releases

See Installation Guide for detailed instructions including package managers.

Connect to Cassandra

# Basic connection
cqlai --host 127.0.0.1

# With authentication
cqlai --host cassandra.example.com -u myuser
# Password: [hidden prompt]

# With keyspace
cqlai --host 127.0.0.1 -k my_keyspace

Run Queries

-- Standard CQL
SELECT * FROM users;
DESCRIBE TABLES;
CONSISTENCY QUORUM;

-- AI-powered (requires configuration)
.ai list all tables in this keyspace

---

---

Comparison with cqlsh

What CQLAI Does Better

1. No Python Required

   • cqlsh requires Python 3.x with cassandra-driver
   • CQLAI is a single binary with zero dependencies

2. Better Performance

   • Faster startup time
   • Memory-bounded result display (will not OOM on large queries)
   • Streaming export for large tables

3. Modern Terminal Experience

   • Rich UI with mouse support
   • Vim-style navigation
   • Better auto-completion

4. AI Integration

   • Generate queries from natural language
   • Schema-aware suggestions

5. Parquet Support

   • Direct export to Parquet format
   • Integration with analytics workflows

Full Compatibility

CQLAI supports:

• All CQL commands your cluster supports
• Meta-commands: DESCRIBE, SHOW, CONSISTENCY, TRACING, etc.
• COPY TO/FROM for data import/export
• SSL/TLS and authentication
• User-Defined Types (UDTs) and complex types
• Batch mode for scripting

---

Example Session

$ cqlai --host 127.0.0.1 -k my_app

Connected to: 127.0.0.1:9042 | Keyspace: my_app | CL: LOCAL_ONE

cqlai> DESCRIBE TABLES;

users
orders
products
events

cqlai> SELECT * FROM users LIMIT 5;

┌──────────────────────────────────────┬────────────┬─────────────────────┐
│ user_id                              │ username   │ email               │
├──────────────────────────────────────┼────────────┼─────────────────────┤
│ 550e8400-e29b-41d4-a716-446655440000 │ john_doe   │ john@example.com    │
│ 660e8400-e29b-41d4-a716-446655440001 │ jane_smith │ jane@example.com    │
│ 770e8400-e29b-41d4-a716-446655440002 │ bob_wilson │ bob@example.com     │
└──────────────────────────────────────┴────────────┴─────────────────────┘
(3 rows)  |  Latency: 2.3ms

cqlai> .ai find users who registered this month

Generated CQL:
  SELECT * FROM users
  WHERE created_at >= '2024-01-01'
  AND created_at < '2024-02-01'
  ALLOW FILTERING;

Execute? [Y/n]: y

┌──────────────────────────────────────┬────────────┬─────────────────────┐
│ user_id                              │ username   │ created_at          │
├──────────────────────────────────────┼────────────┼─────────────────────┤
│ 880e8400-e29b-41d4-a716-446655440003 │ new_user   │ 2024-01-15 10:30:00 │
└──────────────────────────────────────┴────────────┴─────────────────────┘
(1 row)  |  Latency: 4.1ms

cqlai> COPY users TO 'users_backup.parquet' WITH COMPRESSION='SNAPPY';

Exported 3 rows to users_backup.parquet (Parquet, Snappy)

cqlai>

---

Resources

• GitHub: github.com/axonops/cqlai
• Releases: github.com/axonops/cqlai/releases
• Issues: github.com/axonops/cqlai/issues
• Discussions: github.com/axonops/cqlai/discussions

---

Next Steps

• GitHub Repository - Download and documentation