examining-ado-project

name

examining-ado-project

description

Understand the work done in an ado project/context: recent activity, research underway, study documents, spaces and operations over time, experiments and configs used. Use when the user asks for the latest or recent activity in a project; what research strands, studies, or work are underway; a project or context overview or status; trends across spaces/operations/studies; or to onboard onto an ado project.

Examining an ado Project

End-to-end workflow to summarize all discoveryspaces, operations, and related metadata in the ado project associated to the active context.

• Run all commands from the repository root with uv run (see using-ado-cli).
• The report produced by this skill is stored as a document resource in the active ado metastore context (see Producing a report).
• A project is a namespace for a set of ado resources (spaces, operations, stores). A context is the local config pointing to a project's metastore. Context name equals project name by construction. This definition is sufficient for applying this skill. Only consult the source if you need full schema details: read docs/resources/metastore.md if the source repo is available, otherwise see https://ibm.github.io/ado/latest/resources/metastore/#contexts-and-projects.
• Users may refer to an ado project using either the term "project" or "context"

Tips

• Prefer metastore listing and YAML dumps before heavy uv run ado show data pulls; see query-ado-data.
• For creating document resources that store reports, see resource-yaml-creation — Document.
• For study documents (study-$ID), see create-research-study-document.
• For one space in depth: examining-discovery-spaces.
• For one operation in depth: examining-ado-operations.

Pre-requisites

Check active context is associated with expected project

Context names and project names are identical by construction.

If asked to examine a specific named project:

1. Check the active context name: uv run ado context
   • If it has the correct name, continue to step one
   • If not, run uv run ado get contexts to list all available contexts
     • If one matches, switch to it: uv run ado context $NAME
     • If none match, inform the user that a context for the specified project cannot be found

Workflow

Step 1. Check for existing report

The project report is stored as a document resource whose metadata.name is project_report.

Query the metastore:

uv run ado get document -q 'config.metadata.name=project_report' --details

There may be zero, one, or many documents with that name — each has a unique identifier. If several match, the current report is the most recently created one.

If matches exist

• identify the current document (newest by creation timestamp from --details or from document YAML created / status timestamps).
• Fetch its content (uv run ado get document DOCUMENT_ID -o yaml).
• Proceed to step 2

If no matches exist:

• Proceed to Step 3

Step 2: Determine if there has been recent activity

Here we check whether there has been meaningful activity since the current report was written:

1. Contextualizing Activity
   1. Run uv run ado get docs --details
   2. Find any study docs that were created after the last project report was written
   3. If there are, there has been recent activity in describing the motivations and scope of project
2. Operation Activity
   1. Run uv run ado get operations --details -o stats
   2. Find all the operations that were created after the last report was written
   3. Filter these for ones that are finished AND, if explore operations, that measured entities
   4. If there are, there has been recent experiment and analysis activity

If there has been no recent activity, summarize the existing report for the user and point them to it. Do not continue.

If there has been recent activity, continue to next steps but focus on examining the new activity in order to create an updated report based on the existing one.

Step 3: Get overview of research

First check if there are study documents outlining the research underway in the project

uv run ado get document --details | grep "study-"

For each study found, run

uv run ado get document DOCUMENT_ID -o yaml

to understand the study purpose and to find the labels used to identify resources associated with the study

If there are study documents perform the next steps per study.

Step 4: Examine research activities

Goal: volume of work, recency, and which spaces attract the most operations.

1. Spaces (tabular, with metadata)

   uv run ado get spaces --details
   

   Use age (list is age-sorted, most recent last), name, description, and labels to infer themes and activity.

2. Operations (tabular, with metadata)

   uv run ado get operations --details
   

   Relates operations to target spaces; age and labels summarize recent work.

   Heuristic — operation IDs: many ids encode the operator and a version segment, e.g. OPERATOR_NAME-VERSION-...-UID (exact shape varies). Use this together with uv run ado get operator --details to understand more about the operators used.

3. Operations (with measurement statistics)

   uv run ado get operations -o stats --output-file operations-stats.txt
   

   Adds TOTAL_RESULTS, SUCCESSFUL_RESULTS, FAILED_RESULTS, and MEASURED_ENTITIES columns. For explore operations, compare MEASURED_ENTITIES against the operator's numberEntities to check sampling completeness.

   For richer stats that also include request-level counts (TOTAL_REQUESTS, FAILED_REQUESTS, SUCCESSFUL_REQUESTS):

   uv run ado show stats operation --output-file operations-fullstats.csv -o csv
   

4. Discovery Spaces (with statistics)

   uv run ado get spaces -o stats --output-file spaces-stats.txt
   

   Adds EXPERIMENTS, OPERATIONS, EXPLORE_OPERATIONS, and MEASURED_ENTITIES columns.

   For richer stats that also include full entity-space coverage columns (SIZE_OF_ENTITY_SPACE, UNMEASURED_ENTITIES, MATCHING_ENTITIES, etc.):

   uv run ado show stats discoveryspace --output-file spaces-fullstats.csv -o csv
   

   Performance note: ado show stats discoveryspace is slower than as it instantiates each . Prefer for a quick overview across many spaces.

Synthesis: cluster mentally (or in notes) by creation time to see bursts of activity; count operations per space from the operations listing to see which spaces are busiest; group activity under active studies when study documents exist.

Step 5: Deeper pass: full YAML and experiments

Goal: experiments/actuators in use, operation parameters, and how much was submitted for measurement (entities selected by explore-style operators—may differ from completed measurements if runs failed).

Dump full resource documents for easier scripted or batched reading. Use --output-file to ensure proper file handling:

uv run ado get spaces -o yaml --output-file spaces.yaml
uv run ado get operations -o yaml --output-file operations.yaml

On large metastores, these files can be very large—prefer uv run ado get … -q … filters, scripts, or the SQL store API (see query-ado-data) before dumping everything.

When interpreting YAML fields, confirm paths against resource schemas:

uv run ado template discoveryspace --include-schema
uv run ado template operation --include-schema

From space YAML: note experiment and actuator identifiers referenced by each space.

Gain further information on the experiments using

# Outputs description of actuators used
uv run ado get actuators --details
# Outputs description of experiments used
uv run ado get experiments --details
# Outputs detailed information on an experiment
# Use to drill down into most used experiments
uv run ado describe experiment $EXPERIMENT_ID

From operation YAML: note actuatorconfigurations used (if any), read parameters (operator-specific), target discoveryspace references, and any fields that indicate entity count / batch / sample configuration for explore operations.

Synthesize:

• What is being measured and with which experiments (domain of the project).
• Which explore (or similar) operations drove the largest submitted entity sets and on which spaces.
• Whether experiment definitions or parameters shift over time (versions, configs).

Step 6: Space relationships and entity-space shape

From step 2, pick a handful of commonly operated on or recent spaces. For each, inspect its fragment in spaces.yaml, focusing on entity space structure (dimensions, bounds, representation).

Find Spaces that match these spaces (refinement, expansion, or parallel configurations)

uv run ado get spaces --matching-space-id SPACE_ID

Use the output to

• Group spaces that match
• Within each group identify subgroups whose spaces are identical
• If multiple subgroups identify the hierarchy between them (broadest to narrowest)
• Within subgroups identify what if measurement space is different between them
• Combine this information with the sequence of space creation to understand how researchers have been evolving the spaces

Optionally complement with one-hop links:

uv run ado show related space SPACE_ID

Step 7: Write the report

Write a concise markdown report. Store it as the content field of a document resource (see resource-yaml-creation — Document).

• Set metadata.name to project_report.
• Set metadata.description to include today's date, e.g. Project overview as of YYYY-MM-DD.
• Omit relatedResources (or leave empty).

Project summary

• If study documents available:
  • Summary of the particular study or studies underway
• If no study documents available or many resource without associated study
  • Domains or problems implied by experiments, actuators, and space descriptions.
• Dominant operation and experiment patterns.

Latest activity

Per-study when possible

• Most recent spaces and operations (from --details listings).
• What the latest work seems focused on (labels, names, target spaces).

Spaces overview

Per-study when possible

• Which spaces are most used and most analyzed.
• How entity spaces and matching-space relationships evolve: expanding, narrowing, or shifting configuration.

Operations overview

Per-study when possible

• Operator mix and whether parameters or operator choice evolve over time.
• Which operations submitted the most entities (from operation YAML/config) and which produced the most results (from the stats).
• What analysis-style operations ran (infer from operator names and parameters).
• Make a note of operations with failed measurements and highlight ones with abnormal failure rates (from the stats).