xrtm 0.3.0

The xRtm Framework (Generative Forecasting).

xRtm: The Generative Forecasting Framework

xRtm is an open-source framework for institutional-grade generative forecasting and agentic reasoning.

Installation

python3.11 -m venv .venv
. .venv/bin/activate
pip install xrtm==0.3.0

This installs the complete framework, including all components below.

Supported Python versions are >=3.11,<3.13. Python 3.13 is intentionally excluded until the dependency stack is validated there.

---

Product shell

The top-level xrtm command is the local-first product cockpit for the stack:

xrtm doctor
xrtm demo --provider mock --limit 2
xrtm profile create local-mock --provider mock --limit 2
xrtm run profile local-mock
xrtm runs list
xrtm runs show <run-id>
xrtm runs compare <run-id-a> <run-id-b>
xrtm runs export <run-id> --output export.json
xrtm perf run --scenario provider-free-smoke --iterations 3 --limit 1
xrtm artifacts inspect runs/<run-id>
xrtm report html runs/<run-id>
xrtm monitor start --provider mock --limit 2
xrtm monitor run-once runs/<run-id>
xrtm tui --runs-dir runs
xrtm web --runs-dir runs

The product shell writes canonical run directories under runs/:

runs/<run-id>/
  run.json
  questions.jsonl
  forecasts.jsonl
  eval.json
  train.json
  provider.json
  events.jsonl
  run_summary.json
  monitor.json
  report.html
  logs/

events.jsonl is a versioned event stream (xrtm.events.v1) and run_summary.json is a compact metrics contract for CLI/TUI/WebUI consumers.

Use --provider local-llm with a local OpenAI-compatible endpoint such as llama.cpp when you want a real local model path.

For installation, local LLM setup, artifact inspection, monitor lifecycle, and troubleshooting, see docs/operator-runbook.md.

---

Ecosystem

Component	Badge	Description
xrtm-forecast		The Inference Engine
xrtm-data		The Snapshot Vault
xrtm-eval		The Judge
xrtm-train		The Training Pipeline

---

Documentation

Full documentation is available at xrtm.org.

---

License

Apache 2.0 — see LICENSE for details.