OMC — Observability Monitor CLI

A terminal observability tool that supports log search, distributed tracing (Trace) analysis, alarm event intelligent analysis, and Agent-driven deep investigation. Data sources include ClickHouse and Elasticsearch.

Installation

npm install -g observable-cli

After installation, the omc command is available.

Quick Start

# Start REPL interactive mode
omc

# Start with arguments (auto-execute then enter REPL)
omc --trace-id abc123
omc -a myapp -L ERROR
omc trace abc123 -v timeline

# Alarm event drill-down (non-interactive, outputs and exits)
omc alarm-drill <eventId>

# Alarm intelligent analysis (non-interactive)
omc alarm-analyze-event-online <eventId>
omc alarm-analyze-file <filePath>

# Agent deep analysis (with conversation loop)
omc alarm-analyze-event-online-agent <eventId>
omc alarm-analyze-file-agent <filePath>

# Metrics query
omc query-metrics 'up{app="myapp"}' --start 1745800000 --end 1745803600

In REPL mode, the prompt is omc>. Use / prefixed commands:

omc> /log                                # Query all logs
omc> /log -s "timeout" -L ERROR          # Search + level filter
omc> /log --trace-id abc123              # Query by traceId
omc> /trace abc123                       # Trace analysis
omc> /trace abc123 -v timeline           # Timeline view
omc> /alarm-drill <eventId>              # Alarm drill-down (logs + traces)
omc> /alarm-analyze-event-online <eventId>  # Alarm analysis (online)
omc> /alarm-analyze-file <filePath>      # Alarm analysis (from file)
omc> /alarm-analyze-event-online-agent <eventId>  # Agent deep analysis (online)
omc> /alarm-analyze-file-agent <filePath>  # Agent deep analysis (file)
omc> /query-metrics 'up{app="myapp"}'    # Prometheus metrics query
omc> /memory show                         # Show all memories
omc> /memory add preference "Check ERROR first" --content "Always filter ERROR level first" --tags "habits"
omc> /memory add alarm_case "myapp OOM" --content "Root cause: memory leak" --tags "OOM" --appname "myapp"
omc> /export                             # Export last query result to /tmp/omc/
omc> /export ~/my-exports                # Export to specified directory
omc> /help                               # Show help
omc> /quit                               # Exit

Command Reference

REPL Commands

After starting omc, you enter REPL interactive mode with the following commands:

Command	Description
`/alarm-drill <eventId>`	Alarm event drill-down (logs + traces combined view)
`/alarm-analyze-event-online <eventId>`	Alarm intelligent analysis (online data collection + LLM analysis)
`/alarm-analyze-file <filePath>`	Alarm intelligent analysis (from exported file)
`/alarm-analyze-event-online-agent <eventId>`	Agent deep analysis (online collection + Agent conversation)
`/alarm-analyze-file-agent <filePath>`	Agent deep analysis (file analysis + Agent conversation)
`/query-metrics <PromQL>`	Prometheus metrics query
`/memory add\|show\|clean\|backup`	Long-term memory management (requires `memory.enabled`)
`/export [dir]`	Export command parameters and results to JSON file
`/log [options]`	Query logs (also `/logs`)
`/trace <traceId> [options]`	Trace analysis
`/help`	Show help
`/quit` or `/exit`	Exit program

Log Query `/log`

Option	Short	Description	Default
`--limit <n>`	`-l`	Limit number of results	200
`--level <level>`	`-L`	Filter by log level (ERROR/WARN/INFO etc.)	-
`--search <text>`	`-s`	Search keyword	-
`--trace-id <id>`	-	Query by traceId	-
`--appname <name>`	`-a`	Filter by application name	-
`--start-time <time>`	-	Start time	-
`--end-time <time>`	-	End time	-
`--query <sql>`	`-q`	Custom SQL query (ClickHouse mode)	-

Trace Analysis `/trace`

Option	Short	Description	Default
`--view <type>`	`-v`	View type: `tree` / `timeline` / `simple`	tree
`--limit <n>`	`-l`	Associated log entry limit	200

Three view types:

tree — Hierarchical call relationships, showing parent-child service structure
timeline — Waterfall timeline, showing span duration distribution
simple — Flat list, all spans sorted by time

Metrics Query `/query-metrics`

Option	Description	Default
`<PromQL>`	Prometheus query expression	-
`--start <time>`	Start time (Unix timestamp)	-
`--end <time>`	End time (Unix timestamp)	-
`--step <duration>`	Query step	`30s`
`--lookback <duration>`	Lookback window	-

# Query in REPL
omc> /query-metrics 'up{app="myapp"}' --start 1745800000 --end 1745803600

# Query from command line
omc query-metrics 'up{app="myapp"}'

Requires pluginsConfig.metrics-prometheus configuration (remoteReadUrl + clusters + namespace).

Alarm Drill-Down `/alarm-drill`

Starting from an alarm event, automatically collects logs and traces:

Query alarm event details from the alarm platform
Call LLM (OpenAI-compatible API) to extract query conditions:
- Time range (startTime / endTime)
- Application name (appname, required; aborts if not found)
- Pod name (optional)
- API path (optional)
Automatically query logs, extract top 5 distinct traceIds
Concurrently query 5 traces, combining call trees + associated logs

# Use in REPL
/alarm-drill <eventId>

# Use from command line (non-interactive)
omc alarm-drill <eventId>

Alarm Intelligent Analysis `/alarm-analyze-event-online` / `/alarm-analyze-file`

Performs LLM deep intelligent analysis from an alarm event or exported file:

Online analysis /alarm-analyze-event-online <eventId>: Online data collection → LLM analysis
File analysis /alarm-analyze-file <filePath>: Analyze from an exported JSON file

Analysis flow:

Collect data (online or from JSON file)
Route to analyzer via analysis.mappings
LLM analysis (when analyzer is configured) or statistical analysis (fallback when not configured)
Supports field filtering (fieldFilter) and data summarization (summary)

# Online analysis
omc> /alarm-analyze-event-online <eventId>

# File analysis
omc> /alarm-analyze-file /tmp/omc/omc-xxx.json

# From command line
omc alarm-analyze-event-online <eventId>
omc alarm-analyze-file <filePath>

Agent Deep Analysis `/alarm-analyze-event-online-agent` / `/alarm-analyze-file-agent`

Starts an Agent conversation loop for incremental investigation on top of intelligent analysis results:

Online Agent /alarm-analyze-event-online-agent <eventId>: Phase1 online collection + LLM analysis → Phase2 Agent deep analysis
File Agent /alarm-analyze-file-agent <filePath>: Phase1 file analysis → Phase2 Agent deep analysis

The Agent has 3 tools it can invoke autonomously:

query_logs — Query application logs
query_trace — Query distributed traces
query_metrics — Query Prometheus metrics

# Online Agent deep analysis
omc> /alarm-analyze-event-online-agent <eventId>

# File Agent deep analysis
omc> /alarm-analyze-file-agent /tmp/omc/omc-xxx.json

# Skip Phase1, load initial prompt from file
omc> /alarm-analyze-file-agent <filePath> --llm-input-file prompt.txt

# From command line
omc alarm-analyze-event-online-agent <eventId>
omc alarm-analyze-file-agent <filePath>

After analysis, the Agent enters a conversation loop. Ask questions to continue investigating, type exit or quit to leave.

Requires agentModel configuration (see Configuration section).

Default Agent (Natural Language Interaction)

When defaultAgent.enabled = true is configured, entering natural language (without / prefix) in REPL is routed to the default Agent, serving as an intelligent observability assistant.

omc> Check recent error logs for myapp
omc> eventId abc123, analyze this alarm for me
omc> What's wrong with traceId xyz789

The default Agent provides 7 tools (6 without memory.enabled): query_logs, query_trace, query_metrics, query_alarm_event, alarm_drill, alarm_analyze, memory_save.

The Agent is created at REPL startup, persists throughout the session, and maintains conversation context
Relationship with dedicated Agents: when a dedicated Agent (/alarm-analyze-*-agent) is active, it takes priority; after it finishes, the default Agent resumes automatically
When memory.enabled is on, the Agent auto-loads existing memories into the system prompt
/ prefixed commands still use the original routing; type exit / quit to exit REPL

Requires defaultAgent configuration (see Configuration section).

Result Export `/export`

After executing a query command, use /export to save command parameters and query results as a JSON file:

# Export to default directory /tmp/omc/
omc> /log -a myapp -L ERROR
  ... (view results)
omc> /export
  Exported to: /tmp/omc/omc-1745800000000-a1b2c3d4.json

# Export to specified directory
omc> /export ~/my-exports
  Exported to: ~/my-exports/omc-1745800000000-a1b2c3d4.json

Export file format:

{
  "command": "/log -a myapp -L ERROR",
  "timestamp": "2026-04-28T12:00:00.000Z",
  "params": { "command": "/log -a myapp -L ERROR", "appname": "myapp", "level": "ERROR" },
  "results": { "queryInfo": "Source: ES", "logSource": "es", "logs": [...] }
}

Supports export for all query commands (/log, /trace, /alarm-drill, /alert). Press q to return to REPL after viewing results, then run /export.

Long-Term Memory `/memory`

Records user preferences and alarm investigation cases for cross-session Agent reference. Requires memory.enabled = true.

Two memory types:

Type	Identifier	Purpose	Example
User Preference	`preference`	Investigation habits, preferences	"Always check ERROR logs first"
Alarm Case	`alarm_case`	Alarm events and investigation results	"myapp OOM alarm, root cause: memory leak"

# Add preference
omc> /memory add preference "Check ERROR first" --content "Always filter ERROR level first" --tags "habits"

# Add alarm case
omc> /memory add alarm_case "myapp OOM" --content "Root cause: memory leak, restarted" --tags "OOM" --appname "myapp"

# Show all memories
omc> /memory show

# Clear memories (auto-backup)
omc> /memory clean

# Manual backup
omc> /memory backup before-clean

Storage location: ~/.omc/memory/memory.md (Markdown format)

Agent integration: When enabled, the default Agent auto-loads existing memories and can proactively save memories via the memory_save tool.

Command Line Arguments

You can also start directly with command line arguments, which auto-execute then enter REPL:

# Log query mode
omc [options]
  -l, --limit <n>       Limit results (default 200)
  -L, --level <level>   Filter by log level
  -s, --search <text>   Search keyword
  --trace-id <id>       Query by traceId
  -a, --appname <name>  Filter by application name
  --start-time <time>   Start time
  --end-time <time>     End time
  -q, --query <sql>     Custom SQL query
  -c, --config <path>   Specify config file

# Trace analysis mode
omc trace <traceId> [options]
  -v, --view <type>     View type: tree | timeline | simple
  -l, --limit <n>       Associated log entry limit

# Alarm event drill-down (non-interactive)
omc alarm-drill <eventId>

# Alarm intelligent analysis (non-interactive)
omc alarm-analyze-event-online <eventId>    Online collection + LLM analysis
omc alarm-analyze-file <filePath>           Analyze from file

# Agent deep analysis (non-interactive, with conversation loop)
omc alarm-analyze-event-online-agent <eventId> [--llm-input-file <path>]
omc alarm-analyze-file-agent <filePath> [--llm-input-file <path>]

# Prometheus metrics query
omc query-metrics <PromQL> [--start <time>] [--end <time>] [--step <duration>] [--lookback <duration>]

Keyboard Shortcuts

REPL Mode

After starting, you enter the omc> interactive prompt:

Key	Action
`↑` / `↓`	Navigate command history
`Enter`	Execute command
`Backspace`	Delete character
`Ctrl+C`	Exit program

Log/Trace View

Key	Action
`/`	Enter search mode
`j` / `↓`	Move down
`k` / `↑`	Move up
`g`	Jump to first entry
`G`	Jump to last entry
`Ctrl+e`	Scroll down one line
`Ctrl+y`	Scroll up one line
`q` / `Esc`	Return to REPL
`v`	Cycle trace view (tree → timeline → simple)

Alarm Analysis View

Key	Action
`j` / `↓`	Scroll down
`k` / `↑`	Scroll up
`Ctrl+e`	Scroll down one line
`Ctrl+y`	Scroll up one line
`g`	Jump to top
`G`	Jump to bottom
`q` / `Esc`	Return to REPL

Search Mode

Press / to enter search mode:

Type keywords to filter in real-time
Enter to confirm search and exit search mode
Esc to cancel search and restore all results
Backspace to delete last character

Configuration

The configuration file is located at ~/.omc.json, or specify a path via -c. Environment variables override the config file. Priority: environment variables > config file > defaults.

Configuration Example

{
  "clickhouseUrl": "http://localhost:8123",
  "database": "default",
  "table": "logs",
  "username": "default",
  "password": "",
  "traceIdField": "traceid",
  "logSource": "es",
  "pluginsConfig": {
    "trace-elasticsearch": {
      "host": "http://<your-es-host>:9200",
      "indexPattern": "jaeger-span-*",
      "username": "<your-es-username>",
      "password": "<your-es-password>"
    },
    "log-elasticsearch": {
      "host": "http://<your-es-host>:9200",
      "indexPattern": "log-*",
      "traceIdField": "traceId",
      "timestampField": "@timestamp",
      "username": "<your-es-username>",
      "password": "<your-es-password>"
    },
    "metrics-prometheus": {
      "remoteReadUrl": "http://localhost:9090",
      "clusters": ["my-cluster"],
      "namespace": "my-namespace",
      "rateInterval": "2m",
      "step": "30s",
      "timeRangeMinutes": 30
    }
  },
  "extractorModel": {
    "baseUrl": "https://<your-llm-api>/v1",
    "token": "<your-api-key>",
    "model": "model-name"
  },
  "agentModel": {
    "baseUrl": "https://<your-llm-api>/v1",
    "token": "<your-api-key>",
    "model": "model-name",
    "api": "openai-completions",
    "maxTokens": 4096,
    "contextWindow": 128000
  },
  "defaultAgent": {
    "enabled": true,
    "maxToolCalls": 10,
    "baseUrl": "https://<your-llm-api>/v1",
    "token": "<your-api-key>",
    "model": "model-name",
    "api": "openai-completions",
    "maxTokens": 4096,
    "contextWindow": 128000
  },
  "memory": {
    "enabled": true,
    "maxEntriesPerType": 20,
    "maxCharsPerType": 10000
  },
  "analysis": {
    "models": {
      "default": {
        "baseUrl": "https://your-llm-api/v1",
        "token": "your-api-key",
        "model": "model-name"
      }
    },
    "analyzers": {
      "sre": {
        "model": "default",
        "prompt": "sre-deep"
      }
    },
    "mappings": [
      {
        "apps": ["myapp"],
        "analyzer": "sre",
        "context": "myapp-context"
      }
    ],
    "fieldFilter": {
      "logFields": ["timestamp", "level", "message", "traceId"],
      "spanFields": ["operationName", "serviceName", "durationMs"],
      "maxDataChars": 80000
    },
    "summary": {
      "enabled": true,
      "model": "default",
      "chunkSize": 20000,
      "targetSize": 4000
    }
  }
}

Configuration Reference

ClickHouse

Field	Description	Default
`clickhouseUrl`	ClickHouse URL	`http://localhost:8123`
`database`	Database name	`default`
`table`	Table name	`logs`
`username`	Username	-
`password`	Password	-
`traceIdField`	traceId field name	-

Elasticsearch

Field	Description
`pluginsConfig.trace-elasticsearch.host`	Trace ES host
`pluginsConfig.trace-elasticsearch.indexPattern`	Trace ES index pattern
`pluginsConfig.log-elasticsearch.host`	Log ES host
`pluginsConfig.log-elasticsearch.indexPattern`	Log ES index pattern
`pluginsConfig.log-elasticsearch.traceIdField`	traceId field name in log ES
`pluginsConfig.log-elasticsearch.timestampField`	Timestamp field name in log ES
`pluginsConfig.log-elasticsearch.fieldMapping`	ES field mapping (appname/level/podname/uri etc.)

Data Source Selection

Field	Description	Options	Default
`logSource`	Log data source	`clickhouse` / `es`	`clickhouse`
`trace.source`	Trace associated log data source	`clickhouse` / `es` / `none`	`clickhouse`

Prometheus Metrics

Field	Description	Default
`pluginsConfig.metrics-prometheus.remoteReadUrl`	Prometheus API URL	-
`pluginsConfig.metrics-prometheus.clusters`	Cluster list	-
`pluginsConfig.metrics-prometheus.namespace`	Namespace	-
`pluginsConfig.metrics-prometheus.rateInterval`	Rate window	`2m`
`pluginsConfig.metrics-prometheus.step`	Query step	`30s`
`pluginsConfig.metrics-prometheus.timeRangeMinutes`	Time window (minutes)	`30`

LLM Parameter Extraction Model

Field	Description	Default
`extractorModel.baseUrl`	LLM API URL (OpenAI-compatible)	-
`extractorModel.token`	API Key	-
`extractorModel.model`	Model name	-

Agent Model

Field	Description	Default
`agentModel.baseUrl`	LLM API URL (OpenAI-compatible)	-
`agentModel.token`	API Key	-
`agentModel.model`	Model name	-
`agentModel.api`	API type	`openai-completions`
`agentModel.maxTokens`	Max tokens	`4096`
`agentModel.contextWindow`	Context window size	`128000`

Default Agent

Field	Description	Default
`defaultAgent.enabled`	Enable default Agent	`false`
`defaultAgent.maxToolCalls`	Max tool calls per conversation turn	`10`
`defaultAgent.baseUrl`	LLM API URL (OpenAI-compatible)	-
`defaultAgent.token`	API Key	-
`defaultAgent.model`	Model name	-
`defaultAgent.api`	API type	`openai-completions`
`defaultAgent.maxTokens`	Max tokens	`4096`
`defaultAgent.contextWindow`	Context window size	`128000`

Long-Term Memory

Field	Description	Default
`memory.enabled`	Enable long-term memory	`false`
`memory.maxEntriesPerType`	Max entries per type (oldest trimmed when exceeded)	`20`
`memory.maxCharsPerType`	Max characters per type (oldest trimmed when exceeded)	`10000`

Alarm Intelligent Analysis

Field	Description
`analysis.models`	Analysis model definitions (key → {baseUrl, token, model})
`analysis.analyzers`	Analyzer definitions (key → {model, prompt, temperature})
`analysis.mappings`	appname → analyzer + context routing mappings
`analysis.fieldFilter.logFields`	Log field whitelist (optional, uses default filter if not set)
`analysis.fieldFilter.spanFields`	Span field whitelist (optional, uses default filter if not set)
`analysis.fieldFilter.maxDataChars`	Max total data characters
`analysis.summary.enabled`	Enable LLM summarization
`analysis.summary.model`	Summary model reference
`analysis.summary.chunkSize`	Max characters per chunk
`analysis.summary.targetSize`	Target summary character count

fieldFilter default fields (when logFields / spanFields are not configured):

Data Type	Default Fields	Notes
Log	`timestamp`, `level`, `appname`, `traceId`, `message`	message truncated to 500 chars
Span	`operationName`, `serviceName`, `durationMs`, `hasError`	No truncation

Configuration modes:

logFields / spanFields not configured → use default filter above
logFields: ["*"] / spanFields: ["*"] → keep all fields, no truncation
logFields: ["field1", "field2"] → extract specified fields, no truncation

Environment Variables

All configuration items can be set via environment variables:

# ClickHouse
MONITOR_CLICKHOUSE_URL="http://localhost:8123"
MONITOR_DATABASE="default"
MONITOR_TABLE="logs"
MONITOR_TRACE_ID_FIELD="traceid"

# Trace ES
TRACE_ES_HOST="http://es-host:9200"
TRACE_ES_INDEX_PATTERN="jaeger-span-*"
TRACE_ES_USERNAME="user"
TRACE_ES_PASSWORD="pass"

# Log ES
LOG_ES_HOST="http://es-host:9200"
LOG_ES_INDEX_PATTERN="log-*"
LOG_ES_TRACE_ID_FIELD="traceId"
LOG_ES_TIMESTAMP_FIELD="@timestamp"
LOG_ES_USERNAME="user"
LOG_ES_PASSWORD="pass"

# Data source selection
MONITOR_LOG_SOURCE="es"
MONITOR_TRACE_LOG_SOURCE="es"

# Prometheus metrics
PROMETHEUS_REMOTE_READ_URL="http://localhost:9090"
PROMETHEUS_CLUSTERS="my-cluster"
PROMETHEUS_NAMESPACE="my-namespace"
PROMETHEUS_RATE_INTERVAL="2m"
PROMETHEUS_STEP="30s"

# Parameter extraction model
EXTRACTOR_MODEL_BASE_URL="https://your-llm-api/v1"
EXTRACTOR_MODEL_TOKEN="your-api-key"
EXTRACTOR_MODEL_NAME="model-name"

# Agent model
AGENT_MODEL_BASE_URL="https://your-llm-api/v1"
AGENT_MODEL_TOKEN="your-api-key"
AGENT_MODEL_NAME="model-name"
AGENT_MODEL_API="openai-completions"

# Default Agent
DEFAULT_AGENT_BASE_URL="https://your-llm-api/v1"
DEFAULT_AGENT_TOKEN="your-api-key"
DEFAULT_AGENT_MODEL="model-name"

# Long-term memory
MONITOR_MEMORY_ENABLED="true"
MONITOR_MEMORY_MAX_ENTRIES="20"
MONITOR_MEMORY_MAX_CHARS="10000"

# Debug switch
OMC_DEBUG=1    # Enable LLM request debug logs + Agent payload logs + /alert command

Prompt Management

All LLM prompts are stored as .md files in the src/prompts/ directory, allowing independent iteration without code changes.

Loading Priority

User custom ~/.omc/prompts/<name>.md — Highest priority, can override any built-in prompt
Built-in src/prompts/<name>.md — Default version shipped with the package

Prompt Files

File	Purpose
`extract-params.md`	Extract query conditions from alarm event JSON (appname, podname, apiPath etc.)
`summary-system.md`	System prompt for log/trace incremental summarization
`summary-user.md`	User template for log/trace incremental summarization
`summary-refine.md`	Second-round refinement prompt when summary exceeds limit
`metrics-summary-system.md`	System prompt for Prometheus metrics incremental summarization
`metrics-summary-user.md`	User template for Prometheus metrics incremental summarization
`metrics-summary-refine.md`	Second-round refinement prompt for metrics summary
`sre-deep.md`	SRE deep analysis prompt (default analysis template)
`agent-sre-deep.md`	Agent deep analysis system prompt (tool usage guidelines + conversation norms)
`default-agent.md`	Default Agent system prompt (observability assistant + scenario flows + tool descriptions)

Template Variables

Prompt files use {{variable}} template variables, replaced at runtime:

Variable	Description	Used In
`{{data}}`	Analysis data (JSON or summary text)	`sre-deep.md`
`{{context}}`	Additional context knowledge	`sre-deep.md`
`{{previousSummary}}`	Previous round summary result	`summary-user.md`, `metrics-summary-user.md`
`{{chunk}}`	Current chunk of new data	`summary-user.md`, `metrics-summary-user.md`
`{{currentSummary}}`	Summary content to be refined	`summary-refine.md`, `metrics-summary-refine.md`
`{{targetSize}}`	Target character count for refinement	`summary-refine.md`, `metrics-summary-refine.md`

Custom Prompts

Create a .md file with the same name under ~/.omc/prompts/ to override the built-in version:

mkdir -p ~/.omc/prompts
cp prompts/extract-params.md ~/.omc/prompts/extract-params.md
# Edit your custom version
vim ~/.omc/prompts/extract-params.md

Analysis prompts can also be specified via the analysis.analyzers.<name>.prompt field in configuration (without .md suffix).

Development

# Install dependencies
npm install

# Run in development mode
npm run dev -- [options]

# Build
npm run build

# Run in production mode
npm start -- [options]

# Run tests
npm test

# Publish
npm publish

Tech Stack

TypeScript
Ink — React for CLI
pi-agent — Agent runtime framework (tool invocation + conversation loop)
@sinclair/typebox — Agent tool parameter Schema
chalk — Terminal ANSI colors
marked — Markdown parsing (terminal rendering)
pino — Structured logging (enabled when OMC_DEBUG=1)
ClickHouse / Elasticsearch
OpenAI-compatible LLM API (parameter extraction + intelligent analysis + Agent conversation)

Name		Name	Last commit message	Last commit date
Latest commit History 3 Commits
src		src
.gitignore		.gitignore
CLAUDE.md		CLAUDE.md
LICENSE		LICENSE
package.json		package.json
readme.md		readme.md
readme_zh.md		readme_zh.md
tsconfig.json		tsconfig.json

Folders and files

Latest commit

History

Repository files navigation

OMC — Observability Monitor CLI

Installation

Quick Start

Command Reference

REPL Commands

Log Query /log

Trace Analysis /trace

Metrics Query /query-metrics

Alarm Drill-Down /alarm-drill

Alarm Intelligent Analysis /alarm-analyze-event-online / /alarm-analyze-file

Agent Deep Analysis /alarm-analyze-event-online-agent / /alarm-analyze-file-agent

Default Agent (Natural Language Interaction)

Result Export /export

Long-Term Memory /memory

Command Line Arguments

Keyboard Shortcuts

REPL Mode

Log/Trace View

Alarm Analysis View

Search Mode

Configuration

Configuration Example

Configuration Reference

ClickHouse

Elasticsearch

Data Source Selection

Prometheus Metrics

LLM Parameter Extraction Model

Agent Model

Default Agent

Long-Term Memory

Alarm Intelligent Analysis

Environment Variables

Prompt Management

Loading Priority

Prompt Files

Template Variables

Custom Prompts

Development

Tech Stack

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Log Query `/log`

Trace Analysis `/trace`

Metrics Query `/query-metrics`

Alarm Drill-Down `/alarm-drill`

Alarm Intelligent Analysis `/alarm-analyze-event-online` / `/alarm-analyze-file`

Agent Deep Analysis `/alarm-analyze-event-online-agent` / `/alarm-analyze-file-agent`

Result Export `/export`

Long-Term Memory `/memory`

Packages