Introduction

Thadam is the open data plane for robotics. It connects robotic and industrial data sources to modern data platforms with edge-native processing, durable buffering, and flexible sinks.

The name comes from Tamil தடம் (thadam): track, trail, trace, path.

Concretely, Thadam is a single Rust binary that runs next to your data producers — on an industrial PC, a robot’s edge computer, a factory server, or a laptop — and takes responsibility for the whole journey of a telemetry message:

1. Capture. Messages from sources such as MQTT brokers are appended to a local, durable write-ahead log (WAL) and flushed to disk before the source is acknowledged. Acknowledged data survives a process crash when fsync is enabled (claims register).
2. Process. Each sink reads the log independently: messages are decoded, normalized into typed, declared datasets while keeping device identity, time domains, and data quality as first-class columns.
3. Deliver. Processed datasets are written as Parquet segment files and shipped to S3-compatible object storage. ClickHouse is the first analytical database sink (P1, Available, outside the core v0.1 reliability gate). While a downstream system is unavailable, data stays in the local log within the configured disk budget, and delivery resumes when it returns; the delivery guarantees page gives the conditions.

Thadam is open source (Apache-2.0), Rust-first, and designed for the edge: unreliable networks, limited CPU and memory, intermittent power, and long periods of disconnection. Durability is a design input, not a bolt-on: the delivery guarantees Thadam makes (and the ones it explicitly refuses to make) are recorded in ADR-0011 and every public claim is gated on benchmark or spike evidence.

This book

This book is the user documentation for operators and integrators. It covers:

• Getting started: build Thadam and set up a broker.
• Concepts: the event model, the capture-first WAL, per-sink chains, and the delivery guarantees.
• Configuration reference: every configuration key with types, defaults, and constraints.
• Sources and sinks: connector-specific options and operational notes.
• Operations: the data directory, crash recovery, retention, the admin API, metrics, logging, exit codes, and error codes.
• CLI reference: every subcommand and flag.

Design documents — ARCHITECTURE.md, product.md, the ADRs, and the spike reports — are linked from Design and decisions rather than copied, so this book stays about using Thadam.

What Thadam is not

Thadam’s scope is defined as much by what it refuses to be as by what it does. From the product scope (product.md sections 6 and 33), Thadam is deliberately not:

• a robot operating system, a robot control or motion-planning stack, or safety-certified control software — Thadam moves data about robots, it does not drive them;
• a full SCADA system, an HMI, or an asset-maintenance-management product;
• a general-purpose workflow/BPM orchestrator, a data warehouse, a BI tool, or a full ML platform;
• a large distributed stream-processing cluster — Thadam complements Kafka, Flink, and Spark rather than replacing them;
• a video storage or video analytics platform (events may reference large media as object URIs, but media bytes do not travel in the telemetry path);
• a managed cloud platform, a GUI, or proprietary vendor lock-in.

For the current milestone (M1 / MVP v0.1), the reliability gate covers MQTT, the capture WAL, Parquet, S3-compatible storage, and replay. Other sources, sinks, and processing steps are built outside that gate, and some are experimental; the roadmap gives each one’s status. Still not built: Delta Lake, hot configuration reload, and dynamic plugins.

Project status

Thadam is in beta. The status banner in the repository README summarizes what works, and the detailed per-component state is tracked in the readiness tracker. Each page marks its status with the same labels as the homepage: Available In progress Planned. Wherever a feature described here is not yet wired end to end, its page says Planned or In progress explicitly.