alprina chat

Start an interactive AI-powered security assistant to answer questions and analyze code.

Synopsis


alprina chat [OPTIONS]

Description

The alprina chat command launches an interactive conversation with Alprina’s AI security assistant. Ask questions about security, get help understanding vulnerabilities, analyze code snippets, and receive expert guidance on secure development practices.

Think of it as having a security expert available 24/7 to help you write safer code.

Options

`--model, -m`

Specify which LLM model to use (default: claude-3-5-sonnet-20241022).


alprina chat --model gpt-4
alprina chat --model claude-3-5-sonnet-20241022

Available models:

claude-3-5-sonnet-20241022 - Anthropic Claude 3.5 Sonnet (recommended)
gpt-4 - OpenAI GPT-4
gpt-4-turbo - OpenAI GPT-4 Turbo
gpt-3.5-turbo - OpenAI GPT-3.5 Turbo (faster, less capable)

`--streaming / --no-streaming`

Enable or disable streaming responses (default: enabled).


# Streaming enabled (default)
alprina chat --streaming
 
# Disable streaming
alprina chat --no-streaming

Streaming shows responses as they’re generated. Disable for cleaner output or slower connections.

`--load, -l`

Load scan results into the chat context for analysis.


alprina chat --load ~/.alprina/out/latest-results.json

The AI will have access to your scan findings and can help you understand and fix them.

Examples

Basic Chat Session

Start a conversation:


alprina chat


💬 Alprina AI Security Assistant
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Ask me anything about security, vulnerabilities, or secure coding!

Type 'help' for commands, 'exit' to quit.

You: What is SQL injection?

Alprina: SQL injection is a code injection attack where an attacker
can execute arbitrary SQL commands on a database...
[detailed explanation]

You: How do I prevent it in Python?

Alprina: Here are the best practices for preventing SQL injection
in Python:

1. Use parameterized queries...

Analyze Scan Results

Load and discuss scan findings:


alprina chat --load ~/.alprina/out/scan-results.json


💬 Alprina AI (with scan context loaded)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

I've loaded your scan results. Found 3 vulnerabilities.

You: Explain the SQL injection finding

Alprina: Looking at your scan results, I found a SQL injection
vulnerability in database.py at line 42:

    query = f"SELECT * FROM users WHERE id = {user_id}"

This is vulnerable because...
[detailed analysis with code examples]

Use Different Model

Try GPT-4 instead of Claude:


alprina chat --model gpt-4

Disable Streaming

For cleaner output:


alprina chat --no-streaming

Chat Commands

Special commands available in the chat:

Command	Description
`help`	Show available commands
`exit`, `quit`	Exit the chat
`clear`	Clear the conversation history
`context`	Show current context (scan results, etc.)
`save`	Save conversation to file
`load <file>`	Load scan results into context
`model <name>`	Switch LLM model

Example:


You: /help
Alprina: Available commands:
  /exit - Exit the chat
  /clear - Clear conversation
  /context - Show loaded context
  ...

Use Cases

Learn Security Concepts


You: What's the difference between XSS and CSRF?

Alprina: Great question! Let me explain...

Understand Vulnerabilities


You: I got a "weak cryptography" finding. What does that mean?

Alprina: Weak cryptography means you're using outdated or
insecure encryption algorithms...

Get Code Review


You: Is this code secure?
```python
password = input("Enter password: ")
if password == "admin123":
    grant_access()

Alprina: No, this code has several security issues:

Hardcoded password…
Plain text comparison…



### Debug Security Issues

You: Why am I getting “CORS error” in my API?

Alprina: CORS (Cross-Origin Resource Sharing) errors occur when… Here’s how to fix it securely…



### Best Practices Guidance

You: What’s the best way to store API keys in Python?

Alprina: Here are the best practices for API key storage:

Use environment variables…
Never commit keys to git…



## Context Awareness

When you load scan results, the AI has access to:
- All vulnerability findings
- File paths and line numbers
- Severity levels
- CWE/CVE references
- Your code snippets

This allows for highly relevant and specific advice.

## Conversation History

The chat maintains conversation context, so you can:
- Ask follow-up questions
- Reference previous answers
- Build on earlier topics

You: Tell me about SQL injection

Alprina: [explains SQL injection]

You: How do I prevent it?

Alprina: To prevent what we just discussed… [uses context]



## Saving Conversations

Save your chat for future reference:

You: /save security-qa.txt

Alprina: ✓ Conversation saved to security-qa.txt



## Privacy & Data

- Conversations are processed by the selected LLM provider
- Scan results loaded into chat are sent to the LLM
- No conversation data is stored on Alprina servers
- Local conversation history is stored in `~/.alprina/chat/`

To clear local history:
```bash
rm -rf ~/.alprina/chat/

Example Workflows

Workflow 1: Understand Findings


# 1. Run scan
alprina scan ./src
 
# 2. Chat about findings
alprina chat --load ~/.alprina/out/latest-results.json
 
# 3. Ask questions
You: What's the most critical issue?
You: How do I fix the SQL injection?
You: Show me secure code examples

Workflow 2: Learn Security


# Start chat
alprina chat
 
# Ask questions
You: What are the OWASP Top 10?
You: Explain each one with examples
You: How do I test for these in my code?

Workflow 3: Code Review


# Start chat
alprina chat
 
# Paste code
You: Review this code for security issues:
[paste code]
 
# Get feedback
Alprina: [detailed security analysis]

Tips for Better Responses

Be Specific: “How do I prevent SQL injection in FastAPI?” vs “How do I secure my API?”
Provide Context: Include language, framework, and specific code when relevant
Ask Follow-ups: Don’t hesitate to ask for clarification or more details
Load Scan Results: Use --load for contextual help with your specific vulnerabilities
Try Different Models: Some models excel at different tasks

Model Comparison

Model	Speed	Security Knowledge	Code Generation	Cost
Claude 3.5 Sonnet	Fast	Excellent	Excellent	Medium
GPT-4	Medium	Excellent	Excellent	High
GPT-4 Turbo	Fast	Very Good	Very Good	Medium
GPT-3.5 Turbo	Very Fast	Good	Good	Low

Limitations

AI responses are not guaranteed to be 100% accurate
Always verify suggestions before implementing
Complex security topics may require human expert review
Rate limits apply based on your subscription tier

alprina scan - Scan code first, then chat about findings
alprina fix - Apply AI-generated fixes
alprina mitigate - Get fix suggestions

Troubleshooting

”API key not configured”

Set up authentication first:


alprina auth login

Slow responses

Try a faster model or disable streaming:


alprina chat --model gpt-3.5-turbo --no-streaming

Out of context

Clear and start fresh:


You: /clear