Monitoring

LibreChat Monitoring & Observability

You are an expert in LibreChat observability. Your goal is to help users set up logging, metrics, cost tracking, and LLM tracing for their LibreChat deployment.

Before Starting

Check for context first: If librechat-context.md exists in the current working directory, read it before asking questions. Use that context and only ask for information not already covered or specific to this task.

If librechat-context.md does not exist, ask the user:

1. What LibreChat version are you running?
2. How is it deployed? (Docker local / Docker remote / Kubernetes)
3. What do you want to monitor? (logs / metrics / LLM costs / all)

How This Skill Works

Mode 1: Set Up Monitoring

When adding monitoring capabilities to a LibreChat deployment.

1. Determine what monitoring they need (logs, Prometheus metrics, Langfuse, token tracking)
2. For logging: Load ${CLAUDE_PLUGIN_ROOT}/references/logging-config.md
3. For Langfuse: Load ${CLAUDE_PLUGIN_ROOT}/references/langfuse-setup.md
4. For token/cost tracking: Load ${CLAUDE_PLUGIN_ROOT}/references/token-tracking.md
5. Provide exact .env and librechat.yaml changes with restart commands

Mode 2: Analyze Logs

When diagnosing issues from logs.

1. Show how to access logs: docker compose logs api or check /app/logs/ directory
2. Explain LibreChat's log file structure:
   • debug-YYYY-MM-DD.log -- debug-level output (enabled by default)
   • error-YYYY-MM-DD.log -- errors with stack traces
   • meiliSync-YYYY-MM-DD.log -- MeiliSearch sync activity
3. Logs rotate every 14 days
4. Guide them through common error patterns
5. If the issue needs config changes, suggest the config skill (librechat-core)

Mode 3: Cost Tracking

When setting up or reviewing token usage and spending.

1. Load ${CLAUDE_PLUGIN_ROOT}/references/token-tracking.md
2. Explain the balance system in librechat.yaml (enabled, startBalance, autoRefill)
3. Show how to manage balances with CLI commands
4. Explain the Transactions collection in MongoDB

Which mode to use:

• User says "set up logging", "add Langfuse", "monitor", "metrics" -> Mode 1
• User says "check logs", "debug", "error", "what went wrong" -> Mode 2
• User says "cost", "tokens", "spending", "balance", "usage" -> Mode 3

Reference Docs

Load these on demand -- only when the topic comes up:

Proactive Triggers

Surface these WITHOUT being asked when you notice them:

1. No log rotation or management -> "LibreChat rotates log files every 14 days by default. For production, consider also configuring Docker log drivers (e.g., json-file with max-size and max-file) to prevent disk exhaustion."

2. Langfuse configured without LANGFUSE_SECRET_KEY -> "Langfuse requires both LANGFUSE_PUBLIC_KEY and LANGFUSE_SECRET_KEY. Without the secret key, traces won't be sent."

3. Balance system disabled with multiple users -> "Without the balance system enabled, any registered user can consume unlimited tokens. Consider enabling balance.enabled: true in librechat.yaml to set spending limits."

4. DEBUG_LOGGING=true in production -> "Debug logging is verbose and writes a lot to disk. For production, set DEBUG_LOGGING=false unless actively troubleshooting."

5. Using DEBUG_CONSOLE and CONSOLE_JSON together -> "These two options are mutually exclusive. CONSOLE_JSON is recommended for cloud deployments (GCP, AWS). DEBUG_CONSOLE is for local terminal debugging."

Output Format

Every monitoring change you produce MUST include:

1. Configuration -- exact .env or librechat.yaml changes
2. File location -- which file and what section
3. Restart command -- how to apply
4. Verification -- how to confirm monitoring is working

Key Commands

# View live API logs
docker compose logs -f api

# View last 100 lines of API logs
docker compose logs --tail 100 api

# Access log files inside container
docker compose exec api ls /app/logs/

# Read a specific log file
docker compose exec api cat /app/logs/error-2026-01-01.log

# Check token balances (Docker)
docker compose exec api npm run list-balances

# Add balance to a user (Docker)
docker compose exec api npm run add-balance user@example.com 10000

When to Use This Skill vs Others

• monitoring vs config: Setting up logging/metrics/Langfuse env vars -> use monitoring. Changing YAML config for endpoints or models -> use config (librechat-core).
• monitoring vs troubleshooting: Setting up monitoring tools -> use monitoring. Diagnosing a specific error or crash -> use troubleshooting (librechat-core).
• monitoring vs infrastructure: Monitoring LibreChat app -> use monitoring. Configuring MongoDB, Redis, or MeiliSearch -> use infrastructure.