OpenTelemetry Plugin for Decent Espresso

Plugin for the Decent Espresso app. The code is largely inspired by the Visualizer Upload plugin.

Overview and Architecture

graph LR
    subgraph Machine[Decent Espresso Machine]
        A1[Espresso App]
        A2[OTLP Plugin]
        A1 -.->|Events| A2
    end

    A2 -->|OTLP/HTTP| B["(Managed) OTel Collector<br/>or EDOT"]
    B -->|Streams| C[Elasticsearch]
    C -->|Visualization & Analysis| D[Kibana]

    style Machine fill:#e8f0fe,stroke:#4e85f4,stroke-width:3px
    style A1 fill:#fff,stroke:#4e85f4,stroke-width:2px
    style A2 fill:#4e85f4,stroke:#333,stroke-width:2px,color:#fff
    style B fill:#00bfb3,stroke:#333,stroke-width:2px,color:#fff
    style C fill:#fed10a,stroke:#333,stroke-width:2px,color:#000
    style D fill:#f04e98,stroke:#333,stroke-width:2px,color:#fff

Loading

Data Flow:

1. Coffee Machine: Decent Espresso with OTLP Plugin (the plugin.tcl file) captures shot data, state changes, and water levels.
2. OpenTelemetry Collector or EDOT: Elastic Distribution for OpenTelemetry is an open source distribution of the OpenTelemetry Collector that receives OTLP logs via HTTP.
3. Elasticsearch: Stores and indexes all telemetry data with custom attributes using Streams.
4. Kibana: Provides dashboards, visualizations, LLM-supported ingestion, and Agents to query the data.

Installation

Copy the plugin.tcl file to your tablet folder de1plus/plugins/otel.

Configuration

• OTLP endpoint: Where an OpenTelemetry Collector or managed endpoint is receiving the data.
• API key: Optional credentials if required by the OTLP endpoint.

To run it in offline mode, start a hotspot on the coffee machine (under Settings, like any Android phone), connect a laptop to it, and point the OTLP endpoint to the laptop's IP address.

Development

• Environment setup
• Symlinked the plugin file into a clone of the de1app repository: ln -s ~/Documents/GitHub/de1app-otel/plugin.tcl ~/Documents/GitHub/de1app/de1plus/plugins/otel/
• Start a local OTel Collector with Elastic (version 9.2+): curl -fsSL https://elastic.co/start-local | sh -s -- --edot
• Send the following request and find the result in Kibana to make sure it's working end to end:

curl -XPOST http://localhost:4318/v1/logs -H "Content-Type: application/json" -d '{
    "resourceLogs": [{
        "resource": {
            "attributes": [{
                "key": "service.name",
                "value": { "stringValue": "my-service" }
            }]
        },
        "scopeLogs": [{
            "scope": {
                "name": "my-logger"
            },
            "logRecords": [{
                "timeUnixNano": "'$(date +%s%N)'",
                "observedTimeUnixNano": "'$(date +%s%N)'",
                "severityText": "INFO",
                "severityNumber": 9,
                "body": {
                    "stringValue": "Hello, OpenTelemetry!"
                }
            }]
        }]
    }]
}'

• Follow the log files in ~/Documents/GitHub/de1app/de1plus/ with tail -f log.txt | grep -i -E "(ERROR|WARNING|otel)". You might have to explicitly flush them with the start / stop button in the app.
• Add an LLM in Kibana like OpenAI for chat-completion (under advanced settings) with model gpt-4.1-mini (for gpt-5-mini you will need a verified organization). Or use the built-in one on Elastic Cloud.
• Make it the default AI connector.
• Enable Agent Builder.
• Enable wired streams on the settings page of Streams (at least in 9.2).
• Enable logs_index locally by adding logs_index: logs in the elasticsearch/otel: exporter (right under password:). On Elastic Cloud, turn on wired streams and configure the fleet policy.
• Partition in the stream based on attributes.log.type for espresso_state-change, espresso_water-level, espresso_data-point, and espresso_shot.
• Start with an LLM suggestion for Grok patterns but tweak it if needed. Example patterns (tweaked):
  • espresso_state-change: \[%{TIMESTAMP_ISO8601:attributes.custom.timestamp}\] state_change from:%{WORD:attributes.custom.from_state}, to:%{WORD:attributes.custom.to_state}
  • espresso_water-level: \[%{TIMESTAMP_ISO8601:attributes.custom.timestamp}\] water_level:%{DATA:attributes.custom.water_level:float}mm, threshold:%{DATA:attributes.custom.water_threshold:float}mm, status:%{LOGLEVEL:severity_text:keyword}
  • espresso_data-point: \[%{TIMESTAMP_ISO8601:attributes.custom.timestamp}\] elapsed:%{DATA:attributes.custom.elapsed:float}, pressure.pressure:%{DATA:attributes.custom.pressure_pressure:float}, pressure.goal:%{DATA:attributes.custom.pressure_goal:float}, flow.flow:%{DATA:attributes.custom.flow_flow:float}, flow.by_weight:%{DATA:attributes.custom.flow_by_weight:float}, flow.by_weight_raw:%{DATA:attributes.custom.flow_by_weight_raw:float}, flow.goal:%{DATA:attributes.custom.flow_goal:float}, temperature.basket:%{DATA:attributes.custom.temperature_basket:float}, temperature.mix:%{DATA:attributes.custom.temperature_mix:float}, temperature.goal:%{DATA:attributes.custom.temperature_goal:float}, totals.weight:%{DATA:attributes.custom.totals_weight:float}, totals.water_dispensed:%{DATA:attributes.custom.totals_water_dispensed:float}, resistance.resistance:%{DATA:attributes.custom.resistance_resistance:float}, resistance.by_weight:%{DATA:attributes.custom.resistance_by_weight:float}, state_change:%{GREEDYDATA:attributes.custom.state_change:float}
  • espresso_shot: \[%{TIMESTAMP_ISO8601:attributes.custom.timestamp}\] profile:title \{%{DATA:attributes.custom.profile}\} author %{DATA:attributes.custom.author} notes \{%{DATA:attributes.custom.notes}\} beverage_type %{WORD:attributes.custom.beverage_type} steps \{.*?\} tank_temperature %{DATA:attributes.custom.tank_temperature:float} target_weight %{DATA:attributes.custom.target_weight:float} target_volume %{DATA:attributes.custom.target_volume:float} %{GREEDYDATA:attributes.custom.more} (this is not complete and could be expanded)

• Create a Data View for logs.*.
• Build and include a dashboard similar to the included export.ndjson example:

• If you want to grant anonymous access to Kibana, add the following role and user (in addition to the kibana.yml configuration):

PUT /_security/role/everyone-read
{
  "indices": [ {
    "names": "*",
    "privileges": [ "read", "view_index_metadata", "read_failure_store", "monitor" ]
  } ]
}
POST /_security/user/anonymous
{
  "password" : "...",
  "roles" : [ "everyone-read", "viewer", "monitoring_user" ],
  "full_name" : "anonymous",
  "email" : "",
  "metadata": {},
  "enabled": true
}

Todo

• Further improve the parsing of espresso shots
• Use wired streams for all setups and add significant events
• Add Alerts and / or SLOs for the water level
• Integrate scale (and grinder?)
• Document how to run a local LLM for offline mode
• Agent Builder integration for coffee stats
• Integrate Workflows into brewing a coffee? Similar to https://github.com/randomcoffeesnob/decent-advanced-rest-api