AI Assistant Configuration

The AI Assistant is Loggator’s most powerful feature, enabling natural language queries and automated log analysis.

Overview

The AI Assistant can:

Search logs using natural language
Analyze container health
Explain errors and suggest solutions
Filter logs by time, container, or stream
Aggregate and summarize log patterns

Setup

1. Get an OpenRouter API Key

OpenRouter provides access to multiple AI models through a single API:

Visit OpenRouter.ai
Sign up for a free account
Go to Keys
Create a new API key
Copy the key (starts with sk-or-v1-)

2. Configure Environment Variables

Add to your .env file:

OPENROUTER_API_KEY=sk-or-v1-your-key-here
AI_MODEL=xiaomi/mimo-v2-flash:free
SITE_URL=http://localhost:3000

3. Restart Loggator

docker compose restart loggator

The AI chat will now be available in the UI!

Model Selection

Free Models

Perfect for getting started and most log analysis tasks:

xiaomi/mimo-v2-flash:free (Recommended)

Speed: Very fast
Quality: Good for log analysis
Best for: Quick queries, error detection
Rate Limits: Generous free tier

AI_MODEL: xiaomi/mimo-v2-flash:free

google/gemini-2.0-flash-thinking-exp:free

Speed: Fast
Quality: Excellent reasoning
Best for: Complex analysis, troubleshooting
Rate Limits: Moderate

AI_MODEL: google/gemini-2.0-flash-thinking-exp:free

meta-llama/llama-3.2-3b-instruct:free

Speed: Very fast
Quality: Good for simple queries
Best for: Basic log search
Rate Limits: High

AI_MODEL: meta-llama/llama-3.2-3b-instruct:free

Paid Models

For production use with high volumes:

anthropic/claude-3.5-sonnet

Speed: Moderate
Quality: Excellent
Best for: Complex analysis, detailed explanations
Cost: ~$3 per 1M input tokens

AI_MODEL: anthropic/claude-3.5-sonnet

openai/gpt-4-turbo

Speed: Moderate
Quality: Very good
Best for: General purpose
Cost: ~$10 per 1M input tokens

AI_MODEL: openai/gpt-4-turbo

google/gemini-pro-1.5

Speed: Fast
Quality: Good
Best for: High volume, cost-effective
Cost: ~$3.50 per 1M input tokens

AI_MODEL: google/gemini-pro-1.5

Available Tools

The AI assistant has access to these tools:

search_logs

Search container logs with filters.

Parameters:

query (required): Search terms
container (optional): Filter by container name
stream (optional): stdout or stderr
limit (optional): Max results (default: 50, max: 100)

Example Query:

“Show me all errors from nginx in the last hour”

list_containers

List all monitored Docker containers.

Parameters:

all (optional): Include stopped containers

Example Query:

“What containers are you monitoring?“

get_container_info

Get detailed information about a specific container.

Parameters:

containerName (required): Container name

Example Query:

“Tell me about the api container”

analyze_container_health

Analyze container health based on recent logs.

Parameters:

containerName (required): Container name
minutes (optional): Time window (default: 60, max: 1440)

Example Query:

“Is my web container healthy?”

Example Queries

Basic Search

Show me recent logs
Find errors
What's in the logs from nginx?

Time-Based

Show errors from the last 2 hours
What happened in the last 30 minutes?
Any warnings today?

Container-Specific

Show me logs from the api container
What errors does the database container have?
Is the web server working correctly?

Stream Filtering

Show me stderr logs
Any errors in the error stream?
What's in stdout for the api container?

Health Analysis

Is the api container healthy?
Analyze the web container
Check for problems in nginx
What's wrong with my database?

Complex Queries

Show me all 500 errors from the api container in the last hour
Find connection timeout errors
What containers are logging errors?
Analyze patterns in the nginx access logs

Chat Features

Persistent History

Your chat history is saved automatically:

Conversations persist across browser sessions
Context is maintained for follow-up questions
Clear history button available

Markdown Formatting

Responses include:

Code blocks with syntax highlighting
Tables for structured data
Lists for multiple items
Bold/Italic for emphasis

Tool Calling Visibility

Watch the AI work:

See which tools are being called
View tool parameters
Inspect tool results

Usage Limits

Free Tier

OpenRouter free models typically include:

Rate limits: 10-20 requests/minute
Daily quotas: 100-200 requests/day
Response size limits

Paid Tier

Paid models offer:

Higher rate limits
No daily quotas
Priority access
Better performance

Cost Optimization

1. Use Free Models

Free models are sufficient for most log analysis:

AI_MODEL: xiaomi/mimo-v2-flash:free

2. Be Specific

More specific queries = less token usage:

❌ “Tell me everything about my logs” ✅ “Show errors from nginx in the last hour”

3. Limit Results

Request only what you need:

❌ “Show me all logs” (thousands of results) ✅ “Show me the last 20 error logs”

4. Use Direct Search

For simple searches, use the search page instead of AI:

Faster response
No API costs
Full Meilisearch power

Security

API Key Protection

Never commit keys to version control
Use .env files (add to .gitignore)
Rotate keys regularly
Use separate keys per environment

Data Privacy

Logs sent to OpenRouter:

Are processed by the AI model
May be used for model improvement (check provider policy)
Should not contain sensitive data

Label-Based Access Control

The AI can only access containers with the monitoring label:

DOCKER_LABEL_FILTER: loggator.enable=true

This prevents unauthorized access to sensitive containers.

Troubleshooting

”AI features disabled”

Check:

OPENROUTER_API_KEY is set
Key is valid and active
Loggator was restarted after adding the key

# Verify environment variable
docker exec loggator env | grep OPENROUTER_API_KEY

# Restart
docker compose restart loggator

”Rate limit exceeded”

Solutions:

Wait for the limit to reset (usually 1 minute)
Upgrade to a paid model
Use direct search for simple queries

Poor responses

Try:

Be more specific in your query
Try a different model
Check if logs are being indexed: http://localhost:3000/search

Slow responses

Causes:

Model is slow (try a “flash” variant)
Large result sets (limit your query)
High load on OpenRouter

Solutions:

# Switch to a faster model
AI_MODEL: xiaomi/mimo-v2-flash:free  # Very fast

# Or
AI_MODEL: google/gemini-pro-1.5  # Fast paid option

Advanced Configuration

Custom System Prompt

The system prompt is defined in the source code. To customize:

Fork the repository
Edit src/routes/api/chat/+server.ts
Modify the SYSTEM_MESSAGE constant
Rebuild and deploy

Multiple AI Models

Run multiple Loggator instances with different models:

services:
  loggator-free:
    image: ghcr.io/mbeggiato/loggator:latest
    ports:
      - "3000:3000"
    environment:
      AI_MODEL: xiaomi/mimo-v2-flash:free

  loggator-premium:
    image: ghcr.io/mbeggiato/loggator:latest
    ports:
      - "3001:3000"
    environment:
      AI_MODEL: anthropic/claude-3.5-sonnet

Monitoring API Usage

Check OpenRouter dashboard for:

Request counts
Token usage
Costs (for paid models)
Rate limit status

Next Steps

Log Search - Manual search without AI
API Reference - Integrate AI chat in your apps
Container Monitoring - Dashboard features