SQL queries on self-hosted environment doesn't work

[mi875]

When I tried to launch Laminar on my server with Docker Compose, the trace and some other basic features worked.
However, features that require SQL queries don't work, resulting in the error "ClickHouse client is not configured". The ClickHouse container actually works and is accessible. That prevents me from using the dashboards and SQL editor.

Here is the screenshot of the dashboards.

Here is the docker-compose.yml

# This compose file is a lightweight version of docker-compose-full.yml.
# It is intended to be used for local quickstarts.
# It does not include RabbitMQ.
# It only includes frontend, clickhouse, postgres, and app-server.

name: lmnr

services:
  postgres:
    image: postgres:16
    volumes:
      - postgres-data:/var/lib/postgresql/data
    environment:
      POSTGRES_USER: ${POSTGRES_USER}
      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD}
      POSTGRES_DB: ${POSTGRES_DB}
    healthcheck:
      test:
        ["CMD", "pg_isready", "-U", "${POSTGRES_USER}", "-d", "${POSTGRES_DB}"]
      interval: 2s
      timeout: 5s
      retries: 5

  clickhouse:
    image: clickhouse/clickhouse-server:latest
    container_name: clickhouse
    volumes:
      - type: volume
        source: clickhouse-data
        target: /var/lib/clickhouse/
      - type: volume
        source: clickhouse-logs
        target: /var/log/clickhouse-server/
      - type: bind
        source: ./clickhouse-profiles-config.xml
        target: /etc/clickhouse-server/users.d/lmnr.xml
    environment:
      CLICKHOUSE_USER: ${CLICKHOUSE_USER}
      CLICKHOUSE_PASSWORD: ${CLICKHOUSE_PASSWORD}
    ulimits:
      nofile:
        soft: 262144
        hard: 262144
    ports:
      - "0.0.0.0:8123:8123"

  frontend:
    image: ghcr.io/lmnr-ai/frontend
    pull_policy: always
    ports:
      - "${FRONTEND_HOST_PORT:-5667}:5667"
    depends_on:
      postgres:
        condition: service_healthy
    environment:
      DATABASE_URL: postgres://${POSTGRES_USER}:${POSTGRES_PASSWORD}@postgres:5432/${POSTGRES_DB}
      PORT: 5667
      BACKEND_URL: http://app-server:8000
      SHARED_SECRET_TOKEN: ${SHARED_SECRET_TOKEN}
      NEXTAUTH_URL: http://<my-IP>:5667
      NEXTAUTH_SECRET: some_secret
      NEXT_PUBLIC_URL: http://<my-IP>:5667
      ENVIRONMENT: LITE # this disables some runtime dependencies
      CLICKHOUSE_URL: http://clickhouse:8123
      CLICKHOUSE_USER: ${CLICKHOUSE_USER}
      CLICKHOUSE_PASSWORD: ${CLICKHOUSE_PASSWORD}
      OPENAI_API_KEY: ${OPENAI_API_KEY}
      AEAD_SECRET_KEY: ${AEAD_SECRET_KEY}

  app-server:
    image: ghcr.io/lmnr-ai/app-server
    pull_policy: always
    ports:
      - "${APP_SERVER_HOST_PORT:-8002}:8000"
      - "${APP_SERVER_GRPC_HOST_PORT:-8001}:8001"
    depends_on:
      postgres:
        condition: service_healthy
      clickhouse:
        condition: service_started
    environment:
      PORT: 8000
      GRPC_PORT: 8001
      DATABASE_URL: postgres://${POSTGRES_USER}:${POSTGRES_PASSWORD}@postgres:5432/${POSTGRES_DB}
      SHARED_SECRET_TOKEN: ${SHARED_SECRET_TOKEN}
      CLICKHOUSE_URL: http://clickhouse:8123
      CLICKHOUSE_USER: ${CLICKHOUSE_USER}
      CLICKHOUSE_PASSWORD: ${CLICKHOUSE_PASSWORD}
      ENVIRONMENT: LITE # this disables runtime dependency on rabbitmq
      AEAD_SECRET_KEY: ${AEAD_SECRET_KEY}

volumes:
  postgres-data:
  clickhouse-data:
  clickhouse-logs:

[dinmukhamedm]

Hey @mi875 thanks for raising this. This is a known issue, we've built a query editor and launched it in cloud as beta, with the open-source slightly lagging behind. I'll get it working over the weekend

[draneone]

@mi875
I figured that out, added CLICKHOUSE_RO_USER and CLICKHOUSE_RO_PASSWORD to the enviroment, got most of dashboard to work.

But trace status and trace durations still errors out with:

"{\"error\":\"Error executing query: Code: 60. DB::Exception: Unknown table expression identifier 'traces' in scope traces_data AS (SELECT toStartOfInterval(start_time, toInterval(1, 'MINUTE')) AS time, status, count() AS value FROM traces WHERE (start_time >= _CAST('1757490369.821', 'DateTime64')) AND (start_time <= _CAST('1757493969.821', 'DateTime64')) AND (trace_type = 0) AND (status IN ('', 'error')) GROUP BY time, status ORDER BY time ASC (UNKNOWN_TABLE)\"}"

[dinmukhamedm]

@draneone thanks, that's smart! We'll update everything there as well, I think these queries may use older versions, we'll add new versions in the init script to fix this

[dinmukhamedm]

Hey @mi875 @draneone sorry about the delay. Most of this is fixed in the newest versions. The queries themselves need a change that I will work on soon. For now, you can try manually changing the queries by executing the following in your postgres:

UPDATE dashboard_charts SET query = REGEXP_REPLACE(
  REGEXP_REPLACE(query, 'span_type[\s\n]*=[\s\n]*1', 'span_type = ''LLM'''),
 'trace_type[\s\n]*=[\s\n]*0', 'trace_type = ''DEFAULT'''
) WHERE 1=1;

You can either pnpm i && npx drizzle-kit studio from frontend directory when running docker-compose-local-dev.yml or expose a port on your postgres docker and connect to it from your host using psql

[dinmukhamedm]

Alternatively, if you create new projects, the queries inside them are already corrected. Obviously, you will want to look at the old data, so I'll work on patching the queries inside existing projects