AI Providers

Configure AI-powered bucket name generation with OpenAI, Ollama, Anthropic, or Gemini.

Overview

s3finder uses LLMs to generate intelligent bucket name variations. The AI analyzes CT log data to discover organizational naming patterns and generates bucket names following those conventions.

AI generation is optional. The permutation engine alone generates 780+ variations per seed.

Context-Aware Generation

When using CT log discovery (-d), the AI receives discovered infrastructure words as context. It analyzes these patterns to generate bucket names that match the organization's naming conventions.

Default Models

Each provider has a default model optimized for bucket name generation:

Provider	Default Model	Environment Variable
OpenAI	`gpt-4o-mini`	`OPENAI_API_KEY`
Anthropic	`claude-3-5-haiku-20241022`	`ANTHROPIC_API_KEY`
Gemini	`gemini-3-flash-preview`	`GEMINI_API_KEY`
Ollama	`llama3.2`	N/A (local)

Supported Providers

OpenAI

Cloud-based, best quality. Requires API key with per-request costs.

Recommended

Ollama

Local, free, and private. Requires local GPU/resources.

Free

Anthropic

Cloud-based Claude models. High quality alternative.

Alternative

Gemini

Google's Gemini models. Fast and cost-effective.

New

OpenAI Configuration

Setup

bash

# Set API key via environment variable
export OPENAI_API_KEY=sk-xxxxx

# Or pass directly (not recommended)
s3finder -s target --ai --ai-key sk-xxxxx

Usage

bash

# Use default model
s3finder -s acme-corp --ai

# Use a specific model
s3finder -s acme-corp --ai --ai-model <model-name>

# Generate more names
s3finder -s acme-corp --ai --ai-count 100

Model Selection

Choose a model based on your needs:

Type	Speed	Quality	Cost
Mini/Small models	Fast	Good	Low
Standard models	Medium	Excellent	Medium
Large/Turbo models	Slow	Excellent	High

Check OpenAI's documentation for the latest available models.

Ollama Configuration

Prerequisites

bash

# Install Ollama (macOS)
brew install ollama

# Start Ollama server
ollama serve

# Download a model
ollama pull <model-name>

Usage

bash

# Use local Ollama
s3finder -s acme-corp --ai --ai-provider ollama

# Custom model
s3finder -s acme-corp --ai --ai-provider ollama --ai-model <model-name>

# Custom Ollama URL
s3finder -s acme-corp --ai --ai-provider ollama --ai-url http://192.168.1.100:11434

Model Selection

Choose a model based on your hardware:

Size	Quality	RAM Required
7-8B parameters	Good	8GB
13-14B parameters	Better	16GB
70B+ parameters	Excellent	48GB+

Check Ollama's model library for available models.

Anthropic Configuration

Setup

bash

# Set API key via environment variable
export ANTHROPIC_API_KEY=sk-ant-xxxxx

# Use Anthropic
s3finder -s acme-corp --ai --ai-provider anthropic

# Use a specific model
s3finder -s acme-corp --ai --ai-provider anthropic --ai-model <model-name>

Model Selection

Choose a model based on your needs:

Type	Speed	Quality	Cost
Haiku (fast)	Fast	Good	Low
Sonnet (balanced)	Medium	Excellent	Medium
Opus (powerful)	Slow	Best	High

Check Anthropic's documentation for the latest available models.

Gemini Configuration

Setup

bash

# Set API key via environment variable
export GEMINI_API_KEY=xxxxx

# Use Gemini
s3finder -s acme-corp --ai --ai-provider gemini

# Use a specific model
s3finder -s acme-corp --ai --ai-provider gemini --ai-model <model-name>

Model Selection

Choose a model based on your needs:

Type	Speed	Quality	Cost
Flash (fast)	Fast	Good	Low
Pro (balanced)	Medium	Excellent	Medium

Check Google's documentation for the latest available models.

Custom Base URL (Proxy Support)

All providers support custom base URLs for proxies or self-hosted endpoints:

bash

# OpenAI with proxy
s3finder -s acme-corp --ai --ai-provider openai --ai-url https://your-proxy.com/v1

# Anthropic with custom endpoint
s3finder -s acme-corp --ai --ai-provider anthropic --ai-url https://your-proxy.com

# Gemini with proxy
s3finder -s acme-corp --ai --ai-provider gemini --ai-url https://your-proxy.com

# Ollama on remote server
s3finder -s acme-corp --ai --ai-provider ollama --ai-url http://192.168.1.100:11434

Use --ai-url when you need to route requests through a corporate proxy, use a self-hosted LLM endpoint, or connect to a remote Ollama instance.

How AI Generation Works

The AI generates bucket names like a "lazy sysadmin" would create them:

Predictable patterns (backup, logs, assets)
Year suffixes (2023, 2024, 2025)
Environment prefixes (dev, prod, staging)
Region suffixes (us-east-1, eu-west-1)
Abbreviations and shortcuts

Example AI Output

For seed "acme-corp", the AI might generate:

text

acme-corp-backup-2024
acme-corp-internal-assets
dev-acme-corp-logs
acme-corp-db-exports
acme-corp-config-backup
staging-acme-corp-uploads
acme-corp-customer-data
acme-corp-analytics-reports

Best Practices

Recommendations:

Start with the permutation engine alone
Add AI if you need more coverage
Use Ollama for privacy-sensitive targets
Small/mini models offer the best cost/quality ratio