LLxprt Privacy-First Telemetry

Overview

LLxprt takes a privacy-first approach to telemetry and logging. All data stays on your local machine by default, with no external transmission to Google or any other third-party services. This documentation explains our privacy-centered telemetry system and how to use it effectively for debugging and analysis.

Key Privacy Principles

Local Only: No data is sent to external services by default
Opt-in: All logging and telemetry features are disabled by default
Transparent: You can see exactly what is logged and where it's stored
User-Controlled: Simple commands to enable, disable, and configure all features
Data Redaction: Sensitive information is automatically redacted from logs
No Upstream Telemetry: Unlike the original gemini-cli, LLxprt never sends data to Google

Conversation Logging

When enabled, LLxprt logs conversations locally to help with debugging, analysis, and improving your AI interactions.

Storage Details

Location: <data>/conversations/ (see Application Directories)
Format: JSONL (one JSON object per line)
File Naming: conversation-YYYY-MM-DD.jsonl (daily rotation)
Permissions: Files are readable only by your user account
Retention: Configurable retention period (default: 30 days)

What Gets Logged

When conversation logging is enabled, the following data is stored locally:

Request Data

User prompts and messages
Provider selection (e.g., "openai", "anthropic", "gemini")
Model parameters and configuration
Tool call requests and parameters
Timestamp and session information

Response Data

AI assistant responses
Tool call results
Token usage statistics
Response metadata (model, provider, timing)

Metadata

Session identifiers
Performance metrics
Error information (when applicable)

What Gets Redacted

LLxprt automatically redacts sensitive information before writing to log files:

API Keys and Tokens: Automatically detected and replaced with [REDACTED_API_KEY]
Credentials: Passwords, auth tokens, and secret keys
File Paths: Local file system paths (configurable)
URLs: Web URLs with potentially sensitive parameters
Email Addresses: Personal email addresses
Personal Information: Phone numbers, SSNs, and other PII patterns

The `/logging` Command

The /logging command provides complete control over conversation logging features.

`/logging status`

Shows the current state of conversation logging.

/logging status

Example Output:

Conversation Logging: Disabled

`/logging enable`

Enables conversation logging with automatic local storage.

/logging enable

Example Output:

Conversation logging enabled. Data stored locally only.

After enabling, all new conversations will be logged to your local <data>/conversations/ directory (see Application Directories).

`/logging disable`

Disables conversation logging. No future conversations will be logged.

/logging disable

Example Output:

Conversation logging disabled. No conversation data will be collected.

Note: This does not delete existing log files. Use your system's file manager to remove old logs if desired.

`/logging show [N]`

Displays the last N log entries from your conversation history (default: 50).

/logging show 25

Example Output:

Conversation Logs (3 entries):
────────────────────────────────────────────────────────────
[1] 14:32:15 → openai: What is TypeScript and how does it differ from JavaScript?...
[2] 14:32:18 ← openai: TypeScript is a superset of JavaScript that adds static typing...
[3] 14:35:22 → gemini: Can you help me debug this React component?...
────────────────────────────────────────────────────────────

`/logging redaction`

Configure what types of data get automatically redacted from logs.

View Current Settings

/logging redaction

Example Output:

Current Redaction Settings:
  • API Keys: Enabled
  • Credentials: Enabled
  • File Paths: Disabled
  • URLs: Enabled
  • Email Addresses: Enabled
  • Personal Info: Enabled

To modify settings:
  /logging redaction --api-keys=false
  /logging redaction --file-paths=true

Modify Redaction Settings

/logging redaction --file-paths=true --emails=false

Available Options:

--api-keys=true/false: Redact API keys and authentication tokens
--credentials=true/false: Redact passwords and credentials
--file-paths=true/false: Redact local file system paths
--urls=true/false: Redact URLs with sensitive parameters
--emails=true/false: Redact email addresses
--personal-info=true/false: Redact phone numbers, SSNs, and PII

Example Output:

Redaction settings updated:
  • redactFilePaths: enabled
  • redactEmails: disabled

Configuration Options

LLxprt's telemetry can be configured through settings files, environment variables, and command-line flags.

Settings File Configuration

Add telemetry configuration to your user settings.json (in LLxprt's config directory) or workspace .llxprt/settings.json:

{
  "telemetry": {
    "logConversations": false,
    "logResponses": false,
    "redactSensitiveData": true,
    "redactFilePaths": false,
    "redactUrls": true,
    "redactEmails": true,
    "redactPersonalInfo": true,
    "conversationLogPath": "",
    "maxLogFiles": 10,
    "maxLogSizeMB": 50,
    "retentionDays": 30,
    "maxConversationsStored": 1000
  }
}

Leave conversationLogPath empty (or omit it) to use the default <data>/conversations location; set it to an absolute path to override.

Configuration Options Reference

Core Logging Settings

logConversations (boolean): Enable conversation logging (default: false)
logResponses (boolean): Include full AI responses in logs (default: false)
conversationLogPath (string): Directory for log files (default: <data>/conversations, see Application Directories)

Data Retention Settings

retentionDays (number): Days to keep log files (default: 30)
maxLogFiles (number): Maximum number of log files to keep (default: 5)
maxLogSizeMB (number): Maximum size of each log file in MB (default: 10)
maxConversationsStored (number): Maximum conversations to store (default: 1000)

Privacy and Redaction Settings

redactSensitiveData (boolean): Enable API key/credential redaction (default: true)
redactFilePaths (boolean): Redact local file paths (default: false)
redactUrls (boolean): Redact URLs with parameters (default: true)
redactEmails (boolean): Redact email addresses (default: true)
redactPersonalInfo (boolean): Redact PII patterns (default: true)

Environment Variables

You can also control telemetry through environment variables:

LLXPRT_LOG_CONVERSATIONS: Set to "true" to enable conversation logging
LLXPRT_CONVERSATION_LOG_PATH: Override the log directory path

Configuration Precedence

Settings are applied in the following order (highest precedence first):

Command-line flags (when using the llxprt CLI)
Environment variables
Workspace settings (.llxprt/settings.json in current directory)
User settings (your user settings.json in LLxprt's config directory)
Default values

Testing and Development

Local Telemetry for Tests

During testing and development, LLxprt uses local-only telemetry configuration:

// Integration tests use local telemetry target
const telemetryConfig = {
  target: 'local',
  enabled: true,
  logConversations: true,
};

This ensures that:

No test data leaves your development machine
Test telemetry is isolated from production usage
Integration tests can verify logging functionality safely

Debugging with Conversation Logs

To debug issues with LLxprt:

Enable logging: /logging enable
Reproduce the issue: Run the problematic commands
View recent logs: /logging show 20
Examine log files: Check <data>/conversations/ (see Application Directories) for detailed JSONL data

The log files contain structured data that can be analyzed with standard JSON tools:

# View today's conversation log. This Linux example honors LLXPRT_DATA_HOME,
# then LLXPRT_CONFIG_HOME, then the Linux default; see Application Directories
# for the macOS and Windows defaults.
DATA_DIR="${LLXPRT_DATA_HOME:-${LLXPRT_CONFIG_HOME:-$HOME/.local/share/llxprt-code}}"
cat "${DATA_DIR}/conversations/conversation-$(date +%Y-%m-%d).jsonl" | jq '.'

# Filter for specific providers
cat "${DATA_DIR}/conversations/conversation-"*.jsonl | jq 'select(.provider == "openai")'

# Count conversations by provider
cat "${DATA_DIR}/conversations/conversation-"*.jsonl | jq -r '.provider' | sort | uniq -c

Privacy Guarantees

What LLxprt Does NOT Do

No External Transmission: LLxprt never sends your conversation data to Google, OpenAI, Anthropic, or any other external service
No Analytics: No usage analytics or statistics are collected or transmitted
No Cloud Storage: All data remains on your local machine
No Tracking: No user behavior tracking or profiling

What LLxprt DOES Do

Local Storage Only: All logs are written to your local file system with proper permissions
Automatic Redaction: Sensitive information is automatically removed before logging
User Control: You have complete control over what gets logged and for how long
Transparency: All logging behavior is documented and configurable
Data Ownership: You own and control all logged data

Difference from Upstream Gemini CLI

LLxprt fundamentally differs from the original Google Gemini CLI in its approach to telemetry:

Upstream Gemini CLI (What We Don't Do)

May send telemetry data to Google services
Has different privacy policies and data handling
May collect usage statistics for Google's analysis
Telemetry configuration optimized for Google's needs

LLxprt Privacy-First Approach

Zero external transmission: All data stays local by default
Disabled by default: No telemetry or logging without explicit user consent
Complete user control: Users manage all aspects of data collection
Enhanced redaction: Advanced privacy protection with configurable redaction
Open transparency: Full documentation of all privacy practices

Best Practices

For Daily Use

Start with logging disabled: Only enable when you need debugging information
Configure redaction: Enable redaction for all sensitive data types in your workflow
Regular cleanup: Periodically review and clean old log files
Check settings: Verify your privacy settings match your comfort level

For Development

Local testing only: Use local telemetry targets for all development work
Sensitive data awareness: Be mindful of API keys and credentials in test scenarios
Log analysis: Use standard JSON tools to analyze conversation patterns
Documentation: Document any privacy-related configuration for your team

For Team Environments

Consistent settings: Share redaction configurations across team members
No shared logs: Never share raw conversation log files (they may contain sensitive data)
Privacy policies: Establish team policies for conversation logging
Access control: Ensure log files have appropriate file system permissions

Support and Troubleshooting

Common Issues

Q: I enabled logging but don't see any log files A: Check that the log directory exists and you have write permissions. The default path is <data>/conversations/ (see Application Directories).

Q: My log files are very large A: Configure maxLogSizeMB and maxLogFiles in your settings to control file rotation and size limits.

Q: I see sensitive information in logs despite redaction being enabled A: Some patterns may not be caught by automatic redaction. Consider adding custom redaction patterns or disabling logging for sensitive workflows.

Q: How do I permanently delete all conversation logs? A: Remove the entire conversation log directory. This Linux example honors LLXPRT_DATA_HOME, then LLXPRT_CONFIG_HOME, then the Linux default; see Application Directories for the macOS and Windows defaults:

DATA_DIR="${LLXPRT_DATA_HOME:-${LLXPRT_CONFIG_HOME:-$HOME/.local/share/llxprt-code}}"
rm -rf "${DATA_DIR}/conversations/"

Getting Help

For privacy-related questions or concerns:

Check this documentation first
Review your current settings with /logging status and /logging redaction
Examine log files to understand what data is being stored
File an issue in the LLxprt repository for additional support

Remember: Your privacy and data control are fundamental to LLxprt's design. All telemetry features are designed to serve you, not external parties.

LLxprt Privacy-First Telemetry

Overview

Key Privacy Principles

Conversation Logging

Storage Details

What Gets Logged

Request Data

Response Data

Metadata

What Gets Redacted

The /logging Command

/logging status

/logging enable

/logging disable

/logging show [N]

/logging redaction

View Current Settings

Modify Redaction Settings

Configuration Options

Settings File Configuration

Configuration Options Reference

Core Logging Settings

Data Retention Settings

Privacy and Redaction Settings

Environment Variables

Configuration Precedence

Testing and Development

Local Telemetry for Tests

Debugging with Conversation Logs

Privacy Guarantees

What LLxprt Does NOT Do

What LLxprt DOES Do

Difference from Upstream Gemini CLI

Upstream Gemini CLI (What We Don't Do)

LLxprt Privacy-First Approach

Best Practices

For Daily Use

For Development

For Team Environments

Support and Troubleshooting

Common Issues

Getting Help

The `/logging` Command

`/logging status`

`/logging enable`

`/logging disable`

`/logging show [N]`

`/logging redaction`