Record human approval for exactly this job request; returns the token submit_job requires. The calling harness must show the trust summary (snippet, versions, cost, outputs) to a human before calling this.

Build a complete Tapis job request from the app definition.

Mirrors ds.jobs.generate; returns the dict for inspection. Nothing is submitted. SimCenter apps (quoFEM) take no script_filename; dapi wires scInput.json and the driver automatically.

Compile a DAG of job requests without running it.

tasks: [{"task_id", "job", "depends_on": [...], "input_from": {"task_id", "suffix"} (optional)}] Returns the compiled pipeline: deterministic archives, resolved edges.

Every calibration-stage method runnable on DesignSafe, with its engine, what it returns, and its tested/candidate status.

What the grounding index covers and where it came from: each logical source (UW notebooks, dapi, the ds-workflows book), how it resolved (local checkout, fetched cache, or absent), how many passages it contributes, and when the index was built. An absent source means degraded grounding; say so rather than answering from partial knowledge, and fetch_corpus() can fill the gap live.

The app's real interface from Tapis: inputs, parameters, defaults.

Parameters, sensitivities, and the calibration sequence for a constitutive model, transcribed from its manual with page citations.

Use this before planning a calibration: it answers which parameters exist, what data informs each one, and in what order to calibrate.

Estimated SU cost: nodes x hours, per the DesignSafe job-resources guidance.

Fetch grounding sources live from their canonical GitHub repos into corpus/, for deployments without local checkouts.

source: fetch just one logical source by name; empty fetches every GitHub-backed source. CommunityData is not fetched here (needs Tapis auth; see corpus_status for the command). Run reindex after.

List the job archive, or return a small text file's content.

Current Tapis status for a submitted job.

Names of the app templates that ship with dapi.

[schema introspected from dapi 0.6.1 (apps.templates); this tool is generated, not maintained by hand]

List jobs with optional filtering.

Fetches jobs from Tapis ordered by creation date (newest first). Filters are applied client-side.

Args: app_id (str, optional): Filter by application ID. status (str, optional): Filter by job status (e.g., "FINISHED"). Case-insensitive. limit (int, optional): Maximum jobs to fetch. Defaults to 100. output (str, optional): Output format. "df" for pandas DataFrame (default), "list" for list of dicts, "raw" for TapisResult objects. verbose (bool, optional): Print job count. Defaults to False. list_type (str, optional): "MY_JOBS" (default) for jobs you own, "SHARED_JOBS" for jobs shared with you, "ALL_JOBS" for both.

Returns: Depends on output: DataFrame, list of dicts, or list of TapisResult objects.

Raises: JobMonitorError: If the Tapis API call fails.

Example: >>> df = ds.jobs.list(app_id="matlab-r2023a", status="FINISHED") >>> jobs = ds.jobs.list(output="list") >>> raw = ds.jobs.list(limit=10, output="raw")

[schema introspected from dapi 0.6.1 (jobs.list); this tool is generated, not maintained by hand]

List batch queues available on a Tapis execution system.

Args: system_id (str): The ID of the execution system (e.g., "stampede3"). output (str, optional): "df" for DataFrame (default), "raw" for Tapis objects.

Returns: DataFrame or List: Queues with name, maxNodes, maxMinutes, etc.

Example: >>> ds.systems.queues("stampede3")

[schema introspected from dapi 0.6.1 (systems.queues); this tool is generated, not maintained by hand]

List Tapis systems you have access to.

Filters out internal and project-specific systems by default.

Args: category (str, optional): "hpc" for execution systems, "storage" for storage systems, "all" for everything, None for HPC + storage (default). output (str, optional): "df" for DataFrame (default), "list" for dicts.

Returns: DataFrame or List[Dict]: Systems with id, host, category, authn, credentials.

Example: >>> ds.systems.list() # HPC + storage >>> ds.systems.list("hpc") # HPC only with credential status >>> ds.systems.list("storage") # Storage only >>> ds.systems.list("all") # Everything including internal

[schema introspected from dapi 0.6.1 (systems.list); this tool is generated, not maintained by hand]

Which OpenSees to use when: the decision matrix from the training deck.

Returns the variant table (variant -> app_id -> when), the full scope x platform x interface matrix with its legend, and the path of the original slide image for display to the user.

Plan a calibration study from the method matrix.

Pass what you know; unanswered facts come back as open questions. uncertainty_required: must the result carry parameter uncertainty into a downstream prediction (posterior), or is a best-fit point enough? quofem_wrappable: can quoFEM drive the model (OpenSees main script with parameter placeholders and a scalar-QoI postprocessor)? screened: has a sensitivity study already reduced the parameter set?

Plan an OpenSees/quoFEM simulation from the decision matrix.

Pass the facts you know; anything left None is inferred from the request only when unambiguous, and otherwise returned in open_questions for you to ask the user. The decision is the matrix's, not yours: do not override app_id, and compose the cells from the returned snippet, not from memory.

model_language: "tcl" | "python" parallelism: "serial" | "domain-decomposition" | "single-domain-parallel-solver" | "many-independent-cases"

Rebuild the index after mirroring new notebooks or references.

Search every local notebook, model, and script for aligned passages.

Returns the matching text itself with its source path and the index build stamp, so the caller can ground a workflow in what the community actually wrote and knows how fresh that knowledge is. Call corpus_status() to see which sources the index covers; use search_docs for documentation-only search.

Search the documentation corpus: the dapi user guide, the ds-workflows book, SimCenter quoFEM docs, and reference manuals.

Grounding only, never the source of orchestration code; snippets remain the only source an agent may compose runs from. Results carry source and index stamp. Backend today is a local index over live-fetched docs; the Ask AI knowledge graph replaces it later behind this same tool.

Find tested, version-pinned workflow snippets matching a query.

Snippets are executed, self-checking notebooks with pinned app and dapi versions; only entries whose test passed are in the corpus.

Upload a local folder once and return the tapis:// URI to run from.

Submit a validated job. Refuses without the approval token minted by approve_submission for this exact request.

What this server supports, by scientific domain.

Consult this before promising anything. Every capability names how it is served and whether it is tested; 'candidate' and 'untested' entries need a human-run graduation before unattended use. Requests outside this map should be declined by pointing at out_of_scope.

Schema and sanity checks plus input existence, before any SU is spent.

Emit the provenance manifest that makes the run reproducible.

Refuses incomplete provenance: the job must be a fully built request (app, name, resources, inputs) and the uuid must belong to a job this server submitted. A manifest that cannot reproduce the run is worse than none.