OpenResearch/SKILL.md

name, description

name

description

openresearch-cli

Use the `orx` CLI to drive OpenResearch projects from a terminal — browse the experiment tree, runs, logs, artifacts, and the evidence DB; create experiments; launch, wait on, and cancel runs on GPU compute; and chart W&B metrics. Each experiment is a git branch in a local cache-dir clone — reading, diffing, and editing code all happen there with plain git. Read this before driving `orx` programmatically.

OpenResearch CLI (orx)

orx is a command-line client over the OpenResearch API. It authenticates with a personal access token and exposes both read views of a project (experiment tree, runs, logs, artifacts, evidence database) and write actions (create experiments, launch/cancel runs on GPU compute). Code is the one thing orx does not serve: every experiment is a git branch on the project's GitHub repo, and the local clone in ~/.cache/openresearch/repos/<owner>/<repo> is the standard way to read, diff, and edit it (see the orx-git module). Use orx when you need to inspect or drive project state from a shell instead of the web UI.

This overview is deliberately short: it carries the cardinal rules and a command quick-reference, then points at focused modules for everything else. Load a module with orx skill <name> (the live index is printed at the end of orx skill output).

Cardinal rules — read before doing anything else

These four govern everything below. Breaking any one silently invalidates your results — they are not style preferences. The orx-experiment-tree module expands on the why; these are the non-negotiables.

1. Never edit a node once a run has answered it. A node freezes the moment a run establishes its baseline or tests its hypothesis — that includes the root — and freezing is permanent: a disappointing result is still a result. Until then it is provisional: seeding it, fixing its deps, and making it run all happen on its own branch (orx-experiment-tree). To try an idea, branch a child and edit the child.
2. The run command and the environment are a fixed contract — identical on every node. A child inherits its parent's run command verbatim; leave it alone. Do not give nodes different start commands, and do not vary behavior through environment variables or env-prefixed commands (LR=3e-4 python …). The only thing that may differ between nodes is the committed code/config on the node's git branch. orx exp cmd --set is legitimate exactly once: to set the baseline's command when it has none.
3. Vary code, not knobs-in-the-command. Encode hyperparameters in the code/config files and branch a child per variant — never sweep them by editing the run command or passing env vars. Every node runs the same command over different code, so their EVAL.md outputs stay comparable.
4. Grow the tree downward, not sideways. Fan a little within a round (the options of one decision), then descend onto that round's winner for the next round. A root with a long row of direct children and no grandchildren is the failure mode. See "Shape the tree" in the orx-experiment-tree module.

If you're ever tempted to change the command, pass an env var, or pile another node onto the root instead of branching a child, editing its branch, and descending — stop. That's the anti-pattern, not a shortcut.

Setup

orx login          # opens a browser, stores a token at ~/.config/openresearch/credentials.json
orx logout         # remove the stored token

• The API base URL resolves from --api-url → OPENRESEARCH_API_URL → a built-in default. Set OPENRESEARCH_API_URL for non-local use.
• Every command except login, lit, and paper needs a token; if you see Not logged in, run orx login. (lit and paper hit public alphaXiv / OpenAlex / bioRxiv hosts and work without one.)

Command quick-reference

Project-scoped commands take a project id; experiment-scoped commands take an experiment id; run-scoped commands take a run id. Don't mix them — get ids from orx projects, orx experiments, and orx runs respectively. Each group below has a module (orx skill <name>) with the full flags and rules.

Auth

Command	What it does
orx login [--api-url <url>]	Open a browser, do loopback OAuth, store a token.
orx logout	Remove the stored token.

Discover (project- and experiment-scoped)

Command	What it does
orx projects [--all] [--json]	List your projects (id + name + GitHub owner/repo), grouped by org. --all includes archived; --json emits a flat array (incl. each project's paperId) for scripts. Project ids, org ids, and the repo to clone come from here.
orx explore [--json]	List the public project directory (id + name + repo) — projects anyone can view. Drill in with orx project view / orx experiments / orx runs.
orx project view <projectId>	Overview of one project: details, its experiment tree, and its reports. Works on any public project or any private one in your orgs.
orx experiments <projectId>	Print the project's experiments as an indented tree. Experiment ids come from here.
orx runs <projectId> [--experiment <id>]	List runs as a table, newest first. Run ids come from here.
orx env <projectId>	List the names of the env vars a run will see (merged org + project + per-user), each tagged with its source. Names only — values never returned.

Run evidence (run-scoped) — module orx-evidence

Command	What it does
orx logs <runId> [--head] [--bytes <n>] [--range <s>:<e>]	Read a run's terminal log.
orx search-logs <projectId> "<pattern>" (--run <id> | --experiment <id>) [--max <n>]	Grep run logs for a literal pattern.
orx artifacts <runId> / orx artifact <runId> <key> [--head] [--bytes <n>]	List / read a run's text artifacts.
orx wandb <runId>	List the W&B runs linked to a run.
orx chart wandb <projectId> --metric "<key>" --run <runId>[:label] ...	Render a W&B metric across runs to a PNG.
orx query <projectId> "<sql>"	Run one read-only DuckDB SQL statement against the evidence schema.

Create and run experiments (write) — modules orx-create, orx-compute, orx-git

Command	What it does
orx create-project <orgId> --name "<n>" [--repo <owner/repo>]	Create a project bound to a GitHub repo (or a fresh blank repo).
orx project edit <projectId> [--name "<n>"] [--description "<text>" | --description-stdin]	Edit a project's name and/or description (pass at least one); --description-stdin overwrites the description from stdin (long markdown).
orx create-experiment <projectId> --title "<t>" [...]	Add an experiment node; prints its git branch.
orx compute [--gpu <id>] [--count <n>] [--provider <name>] | --cpu]	List the GPU/CPU compute catalog.
orx instance create <orgId> (--gpu <id> … | --cpu <flavor> …)	Spin up a standalone instance in an org.
orx exp status/cmd/run/cancel/wait <expId>	Inspect, run, cancel, and wait on a single experiment node.
orx exp desc <expId> [--set "<text>" | --stdin]	Read or overwrite the experiment's description.
orx report upload/list/show/download <projectId> …	Publish and read project reports (module orx-reports).

To read or edit a node's code — including diffing what a run changed — use plain git in the cache-dir clone; there is no orx code command. See the orx-git module.

Literature & papers — alphaXiv / OpenAlex / bioRxiv (no login required) — module orx-lit

Use before any web search for academic/research queries (paper, author, blog, model release).

Command	What it does
orx lit "<query>" [--source alphaxiv|openalex|biorxiv] [--limit <n>] [--json]	Full-text search; --source picks the corpus (default alphaxiv; openalex = all fields, biorxiv = biology preprints).
orx paper <id|url> [--source ...] [--full]	Fetch a paper: alphaXiv report/--full text, or OpenAlex/bioRxiv metadata+abstract. Source auto-detected from the id.

Meta

Command	What it does
orx skill [name]	Print this overview + the live module index (no args), or print one module / fetch a deeper reference doc by name.

Modules

The detail lives in focused modules — load one with orx skill <name> (the live list, with one-line descriptions, is printed at the end of orx skill output):

• orx-experiment-tree — the experiment-tree model, the auto-research loop, and orx exp desc.
• orx-create — create a project, seed an empty baseline, add experiment nodes.
• orx-compute / orx-compute-k8s — launch runs on compute; the k8s manifest contract.
• orx-git — read, edit, and diff a node's code with plain git.
• orx-evidence — logs, search-logs, artifacts, W&B charts, and the orx query evidence DB.
• orx-reports — write and publish research reports.
• orx-lit — literature search and paper content; the preferred starting point for academic/research queries (not web search).

Deeper API-served references (the project-query schema and worked examples, the report writing guide) are fetchable too — orx skill lists them at the end when the API is reachable.

Typical workflow

Orienting in a project (read-only discovery):

orx projects                     # find the project id
orx experiments <projectId>      # see the tree, pick an experiment id
orx skill experiment-tree        # the model + the auto-research loop
orx runs <projectId>             # find a run id
orx logs <runId>                 # read its output

To actually drive a project toward a goal — edit each node's code on its git branch and keep the GPU capacity saturated — follow the auto-research loop in the orx-experiment-tree module. Every completed run is a decision point with four moves: repair the same node when a run answered nothing, refill the round with another sibling, promote the winner and descend, or stop.