Upload logsentinelai.config

logsentinelai.config

@@ -0,0 +1,263 @@
+####################################
+# LogSentinelAI Configuration File #
+####################################
+
+# =============================================================================
+# Telegram Alert Configuration
+# =============================================================================
+# Enable/Disable Telegram notifications (true/false)
+# When false, all Telegram alert processing is skipped for better performance
+TELEGRAM_ENABLED=true
+
+# Telegram Alert Level - Minimum severity level to trigger alerts
+# Choose from: CRITICAL, HIGH, MEDIUM, LOW, INFO
+# Only events with this severity level or higher will trigger Telegram alerts
+# Example: If set to HIGH, only CRITICAL and HIGH events will send alerts
+# Example: If set to MEDIUM, CRITICAL, HIGH, and MEDIUM events will send alerts
+TELEGRAM_ALERT_LEVEL=CRITICAL
+
+# Telegram Bot Token (required for alerting to Telegram group)
+# Get from @BotFather after creating your bot
+TELEGRAM_TOKEN=YOUR_TELEGRAM_BOT_TOKEN_HERE
+
+# Telegram Chat ID (group or user)
+# Use @userinfobot or similar to get your group chat ID (negative number for groups)
+TELEGRAM_CHAT_ID=YOUR_TELEGRAM_CHAT_ID_HERE
+
+# =============================================================================
+# Logging Configuration
+# =============================================================================
+
+# LOG_LEVEL: Set logging level (DEBUG, INFO, WARNING, ERROR, CRITICAL)
+# Example: LOG_LEVEL=INFO
+LOG_LEVEL=INFO
+
+# LOG_FILE: Path to log file (leave empty for console only)
+# Example: LOG_FILE=logsentinelai.log
+LOG_FILE=/var/log/logsentinelai.log
+
+# ===============================================================================
+# API Keys
+# =============================================================================
+# OpenAI API Key (required if using OpenAI provider)
+OPENAI_API_KEY=YOUR_OPENAI_API_KEY_HERE
+
+# Gemini API Key (required if using Gemini provider)
+GEMINI_API_KEY=AIzaSyCtQ-SyLUn7dyorwD_Nzq8dWLQjqV2EKb
+
+# =============================================================================
+# LLM Configuration
+# =============================================================================
+# LLM Provider - Choose from "ollama", "vllm", "openai", "gemini"
+LLM_PROVIDER=ollama
+#LLM_PROVIDER=vllm
+#LLM_PROVIDER=gemini
+#LLM_PROVIDER=openai
+
+# LLM Model - Ollama
+LLM_MODEL_OLLAMA=qwen3:8b
+#LLM_MODEL_OLLAMA=gemma3:4b
+#LLM_MODEL_OLLAMA=qwen2.5:1.5b
+#LLM_MODEL_OLLAMA=qwen3:1.7b
+
+# LLM Model - vLLM
+LLM_MODEL_VLLM=Qwen/Qwen2.5-1.5B-Instruct
+
+# LLM Model - Gemini
+LLM_MODEL_GEMINI=gemini-1.5-pro
+#LLM_MODEL_GEMINI=gemini-2.5-flash-lite
+
+# LLM Model - OpenAI
+#LLM_MODEL_OPENAI=gpt-4o-mini
+LLM_MODEL_OPENAI=gpt-4.1-nano
+
+# Ollama API Host Configuration (default: http://127.0.0.1:11434/v1)
+# (OpenAI API compatibility)
+LLM_API_HOST_OLLAMA=http://192.168.140.198:11434/v1
+
+# vLLM API host (default: http://127.0.0.1:5000/v1)
+# (OpenAI API compatibility)
+LLM_API_HOST_VLLM=http://127.0.0.1:5000/v1
+
+# OpenAI API host (default: https://api.openai.com/v1)
+LLM_API_HOST_OPENAI=https://api.openai.com/v1
+
+# Gemini API host
+# (OpenAI API compatibility)
+LLM_API_HOST_GEMINI=https://generativelanguage.googleapis.com/v1beta/openai/
+
+# LLM Generation Parameters
+# Temperature: Controls randomness (0.0 = deterministic, 1.0 = very random)
+# Recommended: 0.0-0.1 for log analysis (consistency), 0.7-0.9 for creative tasks
+LLM_TEMPERATURE=0.1
+
+# Top-p: Controls diversity via nucleus sampling (0.1-1.0)
+# Lower values = more focused, Higher values = more diverse
+LLM_TOP_P=0.3
+
+# Max Tokens: Maximum number of tokens to generate in response
+# Controls response length and prevents infinite generation
+# Recommended: 1024-4096 for log analysis depending on complexity
+LLM_MAX_TOKENS=4096
+
+# Show Prompt: Display the full prompt sent to LLM before processing (true/false)
+# Set to true to see the complete prompt for debugging and transparency
+# Useful for debugging prompt issues and understanding what's sent to the LLM
+LLM_SHOW_PROMPT=true
+
+# =============================================================================
+# Analysis Configuration
+# =============================================================================
+# Response Language - Choose from "english", "korean", etc..
+RESPONSE_LANGUAGE=english
+
+# Analysis Mode - Choose from "batch", "realtime"
+# batch: Analyze complete log files (existing functionality)
+# realtime: Monitor and analyze new log entries as they are written
+ANALYSIS_MODE=batch
+
+# =============================================================================
+# Log File Paths
+# =============================================================================
+# Default log file paths used when --log-path is not specified
+LOG_PATH_HTTPD_ACCESS=/workspace/LogSentinelAI/sample-logs/access-10k.log
+LOG_PATH_HTTPD_SERVER=/workspace/LogSentinelAI/sample-logs/apache-10k.log
+LOG_PATH_LINUX_SYSTEM=/workspace/LogSentinelAI/sample-logs/linux-2k.log
+LOG_PATH_GENERAL_LOG=/var/log/secure
+
+# =============================================================================
+# Chunk Size Configuration (entries per chunk)
+# =============================================================================
+# CHUNK_SIZE: Number of log entries sent to LLM for analysis in a single request
+# - Controls analysis quality vs performance balance
+# - Larger values: Better context understanding, slower processing
+# - Smaller values: Faster processing, may miss complex patterns
+# - Recommended: 10-50 depending on log complexity and LLM capacity
+# - These values are used for both batch and real-time modes
+# CLI Override: Use --chunk-size argument to override these values
+# Example: python analysis-linux-system-log.py --chunk-size 20
+CHUNK_SIZE_HTTPD_ACCESS=10
+CHUNK_SIZE_HTTPD_SERVER=10
+CHUNK_SIZE_LINUX_SYSTEM=10
+CHUNK_SIZE_GENERAL_LOG=10
+
+# =============================================================================
+# Elasticsearch Configuration
+# =============================================================================
+ELASTICSEARCH_HOST=http://localhost:9200
+ELASTICSEARCH_USER=elastic
+ELASTICSEARCH_PASSWORD=password
+ELASTICSEARCH_INDEX=logsentinelai-analysis
+
+# =============================================================================
+# Real-time Monitoring Configuration
+# =============================================================================
+# Remote Log Access Mode - Choose from "local", "ssh"
+# local: Monitor local log files (default)
+# ssh: Monitor remote log files via SSH connection
+REMOTE_LOG_MODE=local
+
+# SSH Remote Server Configuration (only used when REMOTE_LOG_MODE=ssh)
+# SSH connection details for remote log monitoring
+REMOTE_SSH_HOST=
+REMOTE_SSH_PORT=22
+REMOTE_SSH_USER=
+# SSH Key Authentication (recommended for security)
+REMOTE_SSH_KEY_PATH=
+# Password Authentication (less secure, use only if SSH key is not available)
+REMOTE_SSH_PASSWORD=
+# SSH Connection Timeout (seconds)
+REMOTE_SSH_TIMEOUT=10
+
+# Polling interval for checking new log entries (seconds)
+# - How often to check log files for new content
+# - Lower values: More responsive, higher CPU usage
+# - Higher values: Less responsive, lower CPU usage
+REALTIME_POLLING_INTERVAL=5
+
+# Maximum number of new lines to process at once (I/O efficiency control)
+# - Limits memory usage and prevents system overload
+# - Controls how many lines are read from file system per polling cycle
+# - IMPORTANT: This is different from CHUNK_SIZE (which controls LLM analysis batching)
+# - If 1000 new lines appear, this setting reads only 50 lines per cycle
+# - Remaining lines are processed in subsequent polling cycles
+# - Recommended: 20-100 depending on system resources and log volume
+REALTIME_MAX_LINES_PER_BATCH=50
+
+# Buffer time to wait for complete log lines (seconds)
+# - Prevents processing incomplete log lines that are still being written
+# - When new lines are detected, waits this many seconds before processing
+# - Ensures log entries are completely written to disk before analysis
+# - Higher values: More safety against incomplete lines, slower responsiveness
+# - Lower values: Faster processing, risk of reading partial log entries
+# - Recommended: 1-5 seconds depending on log writing frequency
+REALTIME_BUFFER_TIME=2
+
+# Real-time processing mode - Choose from "full", "sampling"
+# full: Process all accumulated logs sequentially (default)
+# sampling: Only process the most recent chunk_size logs, discard older ones
+# CLI Override: Use --processing-mode argument to override this setting
+# Example: python analysis-linux-system-log.py --mode realtime --processing-mode sampling
+REALTIME_PROCESSING_MODE=full
+
+# Chunk pending timeout (seconds) for forcing pending logs to be processed
+# - Maximum time to wait for chunk_size logs to accumulate before processing
+# - If pending logs exist but chunk_size is not reached within this time, process anyway
+# - Prevents pending logs from waiting indefinitely in low-volume environments
+# - Set to 0 to disable timeout (wait indefinitely for full chunks)
+# - Recommended: 600-3600 seconds (10-60 minutes) depending on log volume and analysis urgency
+REALTIME_CHUNK_PENDING_TIMEOUT=600
+
+# Auto-sampling threshold (applies to full mode)
+# Automatically switch to sampling if accumulated lines exceed this number
+# - When pending lines exceed this threshold, automatically switch to sampling mode
+# - Prevents memory buildup in high-volume log environments
+# - Works with CHUNK_SIZE: if CHUNK_SIZE=10 and threshold=100, keeps latest 10 lines
+# CLI Override: Use --sampling-threshold argument to override this setting
+# Example: python analysis-linux-system-log.py --mode realtime --sampling-threshold 200
+REALTIME_SAMPLING_THRESHOLD=100
+
+# FLOW EXPLANATION:
+# 1. Every REALTIME_POLLING_INTERVAL seconds, check log file for new lines
+# 2. Read up to REALTIME_MAX_LINES_PER_BATCH lines from file system
+# 3. Add these lines to pending buffer (internal queue)
+# 4. When pending buffer reaches CHUNK_SIZE lines, send to LLM for analysis
+# 5. If pending buffer exceeds REALTIME_SAMPLING_THRESHOLD, apply sampling
+# 6. If pending logs exist but CHUNK_SIZE not reached within REALTIME_CHUNK_PENDING_TIMEOUT seconds, force process anyway
+#
+# Example with current settings:
+# - Check file every 5 seconds
+# - Read max 50 lines per check
+# - Analyze when 10 lines accumulated (CHUNK_SIZE)
+# - Apply sampling if 100+ lines pending (REALTIME_SAMPLING_THRESHOLD)
+# - Force process pending logs after 1800 seconds (30 minutes) timeout (REALTIME_CHUNK_PENDING_TIMEOUT)
+
+# Note: Real-time monitoring will buffer new log lines until CHUNK_SIZE lines
+# are accumulated before sending to LLM for analysis. This improves efficiency
+# by avoiding single-line analysis calls. However, if logs don't reach CHUNK_SIZE
+# within the timeout period, they will be processed anyway to prevent indefinite waiting.
+
+# =============================================================================
+# GeoIP Configuration
+# =============================================================================
+
+# Enable GeoIP enrichment of source IP addresses with city/location information
+# When enabled, source_ips and related fields in analysis results will include city, country, and coordinates (geo_point)
+GEOIP_ENABLED=true
+
+# Path to MaxMind GeoLite2-City database file (.mmdb format)
+# Download with: python -m logsentinelai.utils.geoip_downloader
+# Default: ~/.logsentinelai/GeoLite2-City.mmdb (automatically downloaded if missing)
+GEOIP_DATABASE_PATH=~/.logsentinelai/GeoLite2-City.mmdb
+
+# Fallback country name for unknown IP addresses
+GEOIP_FALLBACK_COUNTRY=Unknown
+
+# Include private/internal IP addresses in GeoIP processing
+# When false, private IPs (192.168.x.x, 10.x.x.x, etc.) are marked as "Private IP"
+# When true, private IPs are processed through GeoIP database (usually results in "Unknown")
+GEOIP_INCLUDE_PRIVATE_IPS=false
+
+# Size of GeoIP lookup cache (number of IP addresses to cache)
+# Higher values reduce database queries but consume more memory
+GEOIP_CACHE_SIZE=1000