insights/docker-compose.sandbox.yml

#
# Self-contained compose file for isolated dev sandboxes.
# Each sandbox is a complete Insights stack with its own infra.
#
# Extends service definitions from docker-compose.base.yml to stay in sync
# with image versions, healthchecks, and environment defaults.
#
# Managed by bin/sandbox — not intended for direct invocation.
# See `sandbox help` for usage.
#

services:
    # Caddy reverse proxy — routes capture/flags/replay traffic to the
    # appropriate service inside the app container, mirroring the dev proxy
    # in docker-compose.dev.yml. Can't extend base because sandbox routes
    # to a single `app` container (not separate service containers).
    proxy:
        extends:
            file: docker-compose.base.yml
            service: proxy
        restart: on-failure
        ports:
            - '${SANDBOX_PORT:-48001}:8000'
        volumes: []
        environment:
            CADDYFILE: |
                http://localhost:8000 {
                    @replay-capture {
                        path /s
                        path /s/*
                    }

                    @capture-ai {
                        path /i/v0/ai
                        path /i/v0/ai/*
                    }

                    @capture {
                        path /e
                        path /e/*
                        path /i/v0/*
                        path /i/v1/analytics/events
                        path /i/v1/analytics/events/
                        path /batch
                        path /batch*
                        path /capture
                        path /capture*
                    }

                    @capture-logs {
                        path /i/v1/logs
                        path /i/v1/logs/*
                    }

                    @capture-traces {
                        path /i/v1/traces
                        path /i/v1/traces/*
                    }

                    @flags {
                        path /flags
                        path /flags*
                        # Local evaluation (flag definitions) — mirrors cloud Contour routing to Rust
                        path /api/feature_flag/local_evaluation
                        path /api/feature_flag/local_evaluation/
                        path /api/feature_flag/local_evaluation/*
                    }

                    @surveys {
                        path /surveys
                        path /surveys/
                        path /api/surveys
                        path /api/surveys/
                    }

                    @remote-config {
                        path /array/*
                    }

                    @webhooks {
                        path /public/webhooks
                        path /public/webhooks/*
                        path /public/m
                        path /public/m/*
                    }

                    handle @capture-ai {
                        reverse_proxy app:3309
                    }

                    handle @capture {
                        reverse_proxy app:3307
                    }

                    handle @capture-logs {
                        reverse_proxy capture-logs:4318
                    }

                    handle @capture-traces {
                        reverse_proxy capture-logs:4318
                    }

                    handle @replay-capture {
                        reverse_proxy app:3306
                    }

                    handle @flags {
                        reverse_proxy app:3001
                    }

                    handle @surveys {
                        reverse_proxy app:3002
                    }

                    handle @remote-config {
                        reverse_proxy app:3002
                    }

                    handle @webhooks {
                        reverse_proxy app:6738
                    }

                    handle {
                        reverse_proxy app:8000
                    }
                }
        depends_on:
            - app

    app:
        # Base build. When the user has tools.yaml entries, bin/sandbox
        # generates a Dockerfile.user that layers tool installs on top and
        # writes an override (docker-compose.tool-auth.yml) that swaps
        # build.dockerfile to point at it.
        build:
            context: .
            dockerfile: Dockerfile.sandbox
            args:
                SANDBOX_UID: '${SANDBOX_UID:-1000}'
                SANDBOX_GID: '${SANDBOX_GID:-1000}'
        restart: on-failure
        init: true
        # mprocs (used by bin/start) requires a TTY
        tty: true
        stdin_open: true
        # VSCode reads this label when attaching to configure extensions and settings
        labels:
            devcontainer.metadata: >-
                [{"customizations":{"vscode":{"extensions":["ms-python.python","ms-python.vscode-pylance","charliermarsh.ruff","oxc.oxc-vscode","bradlc.vscode-tailwindcss"],"settings":{"python.defaultInterpreterPath":"/cache/python/bin/python"}}}}]
        ports:
            # Vite dev server exposed directly (not proxied)
            - '${SANDBOX_VITE_PORT:-49001}:8234'
            # SSH for IDE remote interpreters (IntelliJ, VSCode Remote-SSH, etc.)
            - '${SANDBOX_SSH_PORT:-50001}:2222'
        volumes:
            # Per-sandbox workspace — populated by bin/sandbox before `compose up`.
            # Destroyed with `sandbox rm`, so unpushed commits die with it.
            - workspace:/workspace
            - ${SANDBOX_CLAUDE_AUTH}:/tmp/claude-auth:ro
            - ${SANDBOX_CLAUDE_JSON:-/dev/null}:/tmp/claude-auth.json:ro
            - ${SANDBOX_HOST_GITCONFIG:-/dev/null}:/tmp/host-gitconfig:ro
            - ${SANDBOX_HOST_GIT_SIGNINGKEY:-/dev/null}:/tmp/host-git-signingkey:ro
            - ${SANDBOX_HOST_ALLOWED_SIGNERS:-/dev/null}:/tmp/host-allowed-signers:ro
            - ${SANDBOX_SSH_AUTHORIZED_KEYS:-/dev/null}:/tmp/sandbox-authorized-keys:ro
            # Forwarded SSH agent. Host path resolved by bin/sandbox.
            - ${SSH_AUTH_SOCK:-/dev/null}:/ssh-agent
            # Shared caches — persist across sandboxes so the second one is fast.
            # uv and pnpm download packages here; subsequent sandboxes get cache hits.
            - uv-cache:/cache/uv
            - pnpm-store:/cache/pnpm
            - cargo-target:/cache/cargo-target
            - python-env:/cache/python
            # JetBrains IDE backend for Gateway (populated by bin/sandbox)
            - jetbrains:/opt/idea
        # Sandbox uses the shared service-connection defaults but NOT .env.development —
        # the latter has host-side URLs (SESSION_RECORDING_V2_S3_ENDPOINT=localhost:8333,
        # OBJECT_STORAGE_PUBLIC_ENDPOINT=localhost:19000) that don't resolve inside the
        # container. Dev runtime knobs (DEBUG, etc.) are set inline below. The user's
        # sandbox.env (resolved to a real path or /dev/null by bin/sandbox) injects
        # their secrets straight into the sandbox environment.
        env_file:
            - .env.services
            - ${SANDBOX_ENV_FILE:-/dev/null}
        environment:
            # --- Sandbox identity ---
            SANDBOX_UID: '${SANDBOX_UID:-1000}'
            SANDBOX_GID: '${SANDBOX_GID:-1000}'
            INSIGHTS_SANDBOX: '1'
            # 1 during a cache-backed `create`: the setup window waits for the
            # background Datastore restore to land before migrating.
            SANDBOX_USE_CACHE: '${SANDBOX_USE_CACHE:-0}'
            SANDBOX_INTENTS: '${SANDBOX_INTENTS:-product_analytics}'
            DEBUG: '1'
            HOST_BIND: '0.0.0.0'
            AUTO_LOGIN: '1'
            # Unique session cookie per sandbox so multiple sandboxes on localhost
            # don't overwrite each other's login sessions.
            # (CSRF cookie is NOT varied — frontend hardcodes 'insights_csrftoken'.)
            SESSION_COOKIE_NAME: 'sessionid_${SANDBOX_PORT:-48001}'
            JS_URL: 'http://localhost:${SANDBOX_VITE_PORT:-49001}'
        depends_on:
            db:
                condition: service_healthy
            redis7:
                condition: service_healthy
            datastore:
                condition: service_healthy
            kafka:
                condition: service_healthy
            objectstorage:
                condition: service_started

    db:
        # Not using extends — base mounts volumes that merge with ours,
        # causing duplicate /docker-entrypoint-initdb.d mounts.
        image: postgres:15.12-alpine
        restart: on-failure
        environment:
            POSTGRES_USER: insights
            POSTGRES_DB: insights
            POSTGRES_PASSWORD: insights
        healthcheck:
            test: ['CMD-SHELL', 'pg_isready -U insights']
            interval: 5s
            timeout: 30s
            retries: 30
            start_period: 10s
        volumes:
            - postgres-data:/var/lib/postgresql/data
            - type: volume
              source: workspace
              target: /docker-entrypoint-initdb.d
              read_only: true
              volume:
                  subpath: docker/postgres-init-scripts
        command: postgres -c max_connections=1000 -c idle_in_transaction_session_timeout=300000

    redis7:
        extends:
            file: docker-compose.base.yml
            service: redis7

    datastore:
        extends:
            file: docker-compose.base.yml
            service: datastore
        mem_limit: 8g
        cpus: 2
        environment:
            AWS_ACCESS_KEY_ID: object_storage_root_user
            AWS_SECRET_ACCESS_KEY: object_storage_root_password
        volumes:
            - datastore-data:/var/lib/datastore
            # Shared cache volume for Datastore BACKUP/RESTORE
            - db-cache:/cache/chbackup
            # Directory subpath mounts (supported). Populated before compose up.
            - type: volume
              source: workspace
              target: /idl
              read_only: true
              volume:
                  subpath: insights/idl
            - type: volume
              source: workspace
              target: /docker-entrypoint-initdb.d
              read_only: true
              volume:
                  subpath: docker/datastore/docker-entrypoint-initdb.d
            - type: volume
              source: workspace
              target: /var/lib/datastore/user_scripts
              read_only: true
              volume:
                  subpath: insights/user_scripts
            # Individual Datastore config files can't be volume-subpath mounted:
            # volume subpaths must be directories (files are undocumented and
            # fail for targets that already exist in the image — e.g. config.xml
            # and users.xml). Instead, mount docker/datastore as a directory
            # and copy the files into place in an entrypoint wrapper below.
            - type: volume
              source: workspace
              target: /sandbox-configs
              read_only: true
              volume:
                  subpath: docker/datastore
        entrypoint:
            - /bin/sh
            - -c
            - |
                set -e
                cp /sandbox-configs/config.xml                   /etc/datastore-server/config.xml
                cp /sandbox-configs/users-dev.xml                /etc/datastore-server/users.xml
                cp /sandbox-configs/user_defined_function.xml   /etc/datastore-server/user_defined_function.xml
                cp /sandbox-configs/config.d/default.xml         /etc/datastore-server/config.d/default.xml
                cp /sandbox-configs/config.d/dev-memory.xml      /etc/datastore-server/config.d/dev-memory.xml
                cp /sandbox-configs/config.d/sandbox-backup.xml  /etc/datastore-server/config.d/sandbox-backup.xml
                exec /entrypoint.sh "$@"
            - --
        depends_on:
            - kafka
            - zookeeper

    zookeeper:
        extends:
            file: docker-compose.base.yml
            service: zookeeper

    kafka:
        extends:
            file: docker-compose.base.yml
            service: kafka
        # Lighter resources than dev (1 core, 1G RAM vs 2 cores, 3G)
        command:
            - redpanda
            - start
            - --kafka-addr internal://0.0.0.0:9092,external://0.0.0.0:19092
            - --advertise-kafka-addr internal://kafka:9092,external://localhost:19092
            - --pandaproxy-addr internal://0.0.0.0:8082,external://0.0.0.0:18082
            - --advertise-pandaproxy-addr internal://kafka:8082,external://localhost:18082
            - --schema-registry-addr internal://0.0.0.0:8081,external://0.0.0.0:18081
            - --rpc-addr kafka:33145
            - --advertise-rpc-addr kafka:33145
            - --mode dev-container
            - --smp 1
            - --memory 1G
            - --reserve-memory 200M
            - --overprovisioned
            - --set redpanda.empty_seed_starts_cluster=false
            - --seeds kafka:33145
            - --set redpanda.auto_create_topics_enabled=true

    objectstorage:
        extends:
            file: docker-compose.base.yml
            service: objectstorage

    seaweedfs:
        extends:
            file: docker-compose.base.yml
            service: seaweedfs
        command: ['server', '-s3', '-s3.port=8333', '-filer', '-dir=/data']
        volumes:
            - seaweedfs-data:/data

    elasticsearch:
        extends:
            file: docker-compose.base.yml
            service: elasticsearch

    temporal:
        extends:
            file: docker-compose.base.yml
            service: temporal
        volumes:
            - type: volume
              source: workspace
              target: /etc/temporal/config/dynamicconfig
              read_only: true
              volume:
                  subpath: docker/temporal/dynamicconfig

    temporal-admin-tools:
        extends:
            file: docker-compose.base.yml
            service: temporal-admin-tools
        depends_on:
            - temporal

    temporal-ui:
        extends:
            file: docker-compose.base.yml
            service: temporal-ui
        restart: on-failure
        # No host ports — avoid conflicts between sandboxes
        depends_on:
            temporal:
                condition: service_started
            db:
                condition: service_healthy

    otel-collector:
        extends:
            file: docker-compose.base.yml
            service: otel-collector
        restart: on-failure
        environment:
            # Scope container log collection to this sandbox's compose project
            COMPOSE_PROJECT: '${COMPOSE_PROJECT_NAME}'
        volumes:
            - type: volume
              source: workspace
              target: /etc/otel-collector-config.yaml
              read_only: true
              volume:
                  subpath: otel-collector-config.dev.yaml
        depends_on:
            - capture-logs
        networks:
            - default

    jaeger:
        extends:
            file: docker-compose.base.yml
            service: jaeger
        restart: on-failure
        # No host ports — avoid conflicts between sandboxes
        networks:
            - default

    capture-logs:
        extends:
            file: docker-compose.base.yml
            service: capture-logs
        depends_on:
            - kafka
        networks:
            - otel_network
            - default

    kafka_ui:
        extends:
            file: docker-compose.base.yml
            service: kafka_ui
        # No host ports — avoid conflicts between sandboxes
        depends_on:
            - kafka

    flower:
        extends:
            file: docker-compose.base.yml
            service: flower
        # No host ports — avoid conflicts between sandboxes

    maildev:
        extends:
            file: docker-compose.base.yml
            service: maildev
        # No host ports — avoid conflicts between sandboxes

    webhook-tester:
        image: tarampampam/webhook-tester:2.2.6
        restart: on-failure
        environment:
            - AUTO_CREATE_SESSIONS=true

    etcd:
        image: quay.io/coreos/etcd:v3.6.8
        restart: on-failure
        environment:
            ETCD_NAME: etcd0
            ETCD_ADVERTISE_CLIENT_URLS: http://etcd:2379
            ETCD_LISTEN_CLIENT_URLS: http://0.0.0.0:2379
        # No host ports — avoid conflicts between sandboxes
        healthcheck:
            test: ['CMD', 'etcdctl', 'endpoint', 'health']
            interval: 5s
            timeout: 5s
            retries: 10

    localstack:
        extends:
            file: docker-compose.base.yml
            service: localstack
        restart: on-failure
        # No host ports — avoid conflicts between sandboxes

    objectstorage-azure:
        extends:
            file: docker-compose.base.yml
            service: objectstorage-azure
        # No host ports — avoid conflicts between sandboxes

    duckgres:
        extends:
            file: docker-compose.base.yml
            service: duckgres
        # No host ports — avoid conflicts between sandboxes
        volumes:
            - type: volume
              source: workspace
              target: /app/config.yaml
              read_only: true
              volume:
                  subpath: devenv/duckgres.yaml
            - duckgres-data:/app/data

    redis-cluster:
        image: redis:7.2
        restart: on-failure
        # No host ports — avoid conflicts between sandboxes
        command: >
            bash -c "
            redis-server
            --port 6379
            --cluster-enabled yes
            --cluster-config-file /data/nodes.conf
            --cluster-node-timeout 5000
            --cluster-announce-ip 127.0.0.1
            --cluster-announce-port 6379
            --appendonly yes &
            sleep 1 &&
            redis-cli -p 6379 cluster addslots $$(seq 0 16383) &&
            tail -f /dev/null
            "
        volumes:
            - redis-cluster-data:/data
        healthcheck:
            test: ['CMD', 'redis-cli', '-p', '6379', 'cluster', 'info']
            interval: 2s
            timeout: 3s
            retries: 10

networks:
    otel_network:
        driver: bridge

volumes:
    # Per-sandbox workspace holding the cloned git tree. Lifecycle ties to the
    # compose project — `compose down -v` removes it, taking any unpushed work
    # with it. Populated by bin/sandbox before `compose up`.
    workspace:
    postgres-data:
    redpanda-data:
    datastore-data:
    seaweedfs-data:
    redis-cluster-data:
    duckgres-data:
    python-env:
    uv-cache:
        name: sandbox-uv-cache
        external: true
    pnpm-store:
        name: sandbox-pnpm-store
        external: true
    db-cache:
        name: sandbox-db-cache
        external: true
    cargo-target:
        name: sandbox-cargo-target
        external: true
    jetbrains:
        name: ${SANDBOX_IDE_VOLUME:-sandbox-intellij}
        external: true