Skip to main content
Glama

PhyUDM: Unified Data Model for robot and drone telemetry

License: Apache 2.0 Spec Version Status: Pre-launch

PhyUDM (the Unified Data Model, UDM) is an open, vendor-neutral robot telemetry schema: one event format for mobile robots, AMRs/AGVs, cobots and manipulators, humanoids, drone telemetry, and the fleets they belong to. It serves as a ROS 2 data model that any vendor's stack can map onto, from ROS 2 topics to ROS 2 graph integrity events, and ships as JSON Schemas with a conformance suite, a CLI and an MCP server. Operators, regulators and AI agents get a single common language for physical-world data, regardless of who built the robot.


Why UDM?

Autonomous-system telemetry today is fragmented across vendors and platforms. UDM provides a stable schema for the data every autonomous system produces β€” identity, location, motion, power, safety, perception, manipulation, AI/reasoning, and 17 other domains β€” so the same payload can flow through:

  • compliance and safety auditing (ISO 13482, ANSI/RIA R15.06, ISO/TS 15066, SOC 2, GDPR, …),

  • fleet management and observability backends,

  • LLM agents and autonomous coding pipelines, and

  • digital-twin and simulation environments,

without per-vendor adapters.


Related MCP server: pond

Who this repo is for

1. Spec readers (architects, standards bodies, integrators)

Start with spec/index.md. The navigation hub links to every front-matter, core-schema, domain, reference, and appendix page. The full single-file copy lives at spec/udm-spec.md for offline reading and version-to-version git diff.

Highlights:

2. Autonomous-system implementers (robotics teams, platform builders)

Want to emit UDM from a robot, simulator, or fleet manager? Pair the Core Envelope with the relevant domain pages, then validate your output with the canonical JSON Schema (make validate). The Complete UDM Event Example shows a fully-populated event payload spanning multiple domains.

Phase 2 β€” landed. Canonical JSON Schema (Draft 2020-12) artifacts live under schemas/v0.0.5/; a 152-fixture conformance suite (78 valid + 58 invalid + 15 edge + 1 legacy), plus signing test vectors, lives under conformance/; the versioning contract is at spec/versioning.md. Run make validate to exercise the suite locally.

3. AI / agent developers (LLM tooling, autonomous codegen)

UDM is designed to be the path of least resistance for LLM agents producing or consuming autonomous-system telemetry. The strategic goal is for any agent to be able to:

  1. Discover the schema (via the JSON Schema artifacts and the Model Context Protocol server).

  2. Emit conforming UDM directly, without per-vendor glue.

  3. Validate their output (via the udm CLI).

Available today: the udm CLI (validate, explore and analyse UDM telemetry), the udm-mcp server (Model Context Protocol tools for LLM agents) and the AI cookbook.


Quickstart (spec reader)

git clone https://github.com/Phyware-Labs/PhyUDM.git
cd PhyUDM
$EDITOR spec/index.md           # navigate the split tree
$EDITOR spec/udm-spec.md        # or read the full single-file copy

A minimal UDM event looks like:

{
  "udm_version": "0.0.5",
  "event_id": "01940000-0000-7000-8000-000000000001",
  "event_type": "telemetry_periodic",
  "source_id": "robot-001",
  "source_type": "amr",
  "captured_at": "2026-01-12T10:35:00.123456Z",
  "identity": {
    "source_id": "robot-001",
    "source_type": "amr"
  },
  "location": {
    "frame_id": "map",
    "local": { "x_m": 12.34, "y_m": 56.78, "z_m": 0.0 }
  }
}

See spec/envelope.md for required vs. optional envelope fields and spec/event-example.md for a fully-worked multi-domain example.


Status

Aspect

State

Specification text

v0.0.5 (draft) β€” see CHANGELOG.md

JSON Schema artifacts

βœ… schemas/v0.0.5/ (Phase 2)

Conformance suite

βœ… conformance/ β€” 152 fixtures (Phase 2)

Versioning contract

βœ… spec/versioning.md (Phase 2)

CLI

βœ… crates/udm-cli/

MCP server

βœ… crates/udm-mcp/

Event-store adapters

βœ… crates/udm-eventstore/ trait, in-memory and PhyCloud (read path) adapters

AI cookbook

βœ… docs/ai-cookbook/


Governance

UDM is currently maintained under single-maintainer stewardship by the original authors during the pre-v0.1.0 extraction and stabilization period. Report vulnerabilities as described in SECURITY.md. A formal GOVERNANCE.md will land before the v0.1.0 launch.

The goal is for UDM to become an open de facto standard for autonomous-system telemetry. Cross-vendor governance (e.g., foundation donation) is explicitly out of scope until after v0.1.0 so the surface can stabilize first.


Contributing

Pull requests are welcome. Please read CONTRIBUTING.md before opening a non-trivial change, and sign your commits (git commit -s) per the Developer Certificate of Origin.

This project adopts the Contributor Covenant Code of Conduct. By participating, you agree to abide by its terms.


License

UDM is licensed under the Apache License 2.0 β€” spec text, schema artifacts, and tooling. See NOTICE for attribution and extraction provenance.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      C
      maintenance
      A terminal live-tail and a browser dashboard β€” one process, one event stream, served from localhost. Unified timeline across Claude Code, Codex, Gemini CLI, Cursor, Hermes, and OpenClaw. Token + cost accounting, compaction + anomaly detection, hybrid search, SVG call graphs, monaco-style diff attribution, agent-aware replay ("what would the agent say if I edited the prompt?"), policy editor, MCP s
      8 npm
      16
      MIT
    • A
      license
      A
      quality
      A
      maintenance
      Lossless archive and search for AI agent sessions (Claude Code, Codex, opencode, pi and others), exposed to agents over MCP. Local-first, written in Rust.
      4
      91
      Apache 2.0
    • A
      license
      Not graded
      quality
      B
      maintenance
      Enables natural-language querying and investigation of homelab logs from syslog, Docker, OTLP, and AI transcripts, with SQLite full-text search, timeline correlation, incident clustering, and fleet topology insights.
      4
      AGPL 3.0
    • A
      license
      A
      quality
      B
      maintenance
      Local code-graph engine for AI coding workflows: static analysis of Rust, Cangjie, ArkTS, TypeScript/JavaScript, C/C++, Python, and Shell projects into symbols, imports, call graphs, and quality gates, with pre-commit change review and impact analysis. Read-only by design β€” never executes build scripts or uploads code.
      6
      1
      MIT