| English | 中文 |
Use catalog/tools/*.yaml and list-tools/describe-tool for actual IDs, parameters and support. The reviewed catalog has 26 entries, 17 with reviewed run-tool adapters (8 core CLI, 6 Python and 3 R entries). Registration does not establish local dependencies, login or scientific validity. See queries, execution and validation records.
Use the installed project environment from the repository root. Direct CLI queries retain their public detail; runner summaries are sanitized.
| Tool ID | CLI command | Runner | Scope |
|---|---|---|---|
airmosaic_list_datasets |
list-datasets | yes | Acquisition declarations |
airmosaic_describe_dataset |
describe-dataset | yes | Metadata and existing aliases |
airmosaic_search_datasets |
search-datasets | yes | Local metadata search |
airmosaic_check_availability |
check-availability | yes | Presence only; direct CLI can expose local paths |
airmosaic_draft_causal_plan |
draft-causal-plan | yes | Local planning template, no model call or causal conclusion |
airmosaic_list_tools |
list-tools | yes | Categories, filters and static state |
airmosaic_describe_tool |
describe-tool | yes | Manifest, parameters and entrypoint |
airmosaic_check_tool |
check-tool | yes | Static prerequisites, not login validation |
airmosaic_run_tool |
run-tool | no | Runner display entry; recursion is disabled |
The SDK exposes DataCatalogService, DataAccessService, CausalDesignService, ToolRegistryService and ToolRunnerService; see core usage. Local stdio MCP is implemented (setup); REST, automatic skill installation and general cross-tool orchestration remain unimplemented.
See classified usage.
| Tool ID | Runner | Scope |
|---|---|---|
worldpop_download |
yes | Network metadata discovery; file transfer enabled explicitly |
csmar_process_company_info |
yes | Existing basic-company ZIP, pandas/openpyxl |
ihme_gbd_fetch_results |
yes | Existing authorized local browser, network/session flags and reviewed dimensions |
csmar_browser_download |
reference | User’s institutional access and an available browser/manual flow |
yearbook_regional_sector_template |
template | Adapt region/industry headers, encoding, years and output |
yearbook_product_output_template |
template | Adapt products, units and missingness; no separate national-sector implementation |
Nine dataset declarations are present in this version. They describe metadata, not successful acquisition:
| Dataset ID | Existing responsible skill_id | Declared scope |
|---|---|---|
worldpop_population_grid |
worldpop | Population raster |
worldpop_zonal_statistics |
worldpop | Zonal-statistics declaration, no standalone local processor |
csmar_listed_company_basic |
csmar | Listed-company basic information |
csmar_financial_statements |
csmar | Table-specific acquisition requires review |
industrial_yearbook_national_sector |
ind_yearbook | National-sector indicators need implementation/adaptation |
industrial_yearbook_regional_sector |
ind_yearbook | Regional-sector indicators |
industrial_yearbook_product_output |
ind_yearbook | Regional product output |
ihme_gbd_mortality_rates |
ihme-gbd-results | Mortality rates |
ihme_gbd_population |
ihme-gbd-results | Population; review request dimensions |
Detailed inputs/dependencies are in model analysis, machine learning and the corresponding skills.
| Tool ID | Runner | Scope |
|---|---|---|
gcam_extract_market_sector |
yes | Existing market/sector database rows; pandas and Java/BaseX |
gcam_extract_technology_components |
yes | Component rows and summaries; no method parameter |
gcam_aggregate_provinces |
yes | Existing market CSV; review coverage, units and weights |
gcam_inspect_output_structure |
yes | Fixed R queries and declared database layout |
gcam_plot_sector_change |
yes | Required market CSV, fixed 2021/2035 years and injected task output |
gcam_configuration_reference |
reference | Reviewed configuration work, no simulation launcher |
pm25_hia_calculate_template |
template | Health inputs/method need adaptation |
pm25_hia_visualize_template |
template | Map inputs, grid, dependencies and outputs need adaptation |
ml_regression_workflow |
yes | Numeric prediction, predictive explanations and declared-DAG PLR |
mlr3_workflow_reference |
reference | Framework modeling, tuning and prediction guidance |
timexer_workflow_reference |
reference | External forecasting configuration and inference adaptation |
The four-stage guide distinguishes method-only and computational work. Build troubleshooting is guidance rather than a compilation adapter.
Executable JSON examples contain declared parameters; the tool ID is supplied to run-tool. Lists use the adapter’s documented comma-separated strings, and output parameters can usually be omitted for task allocation. Reference/browser plans with reference_tool_id and execution_supported=false are not executable runner parameters.
New tools need a manifest, reviewed policy/adapter, parameter handling, classified guidance and a skill under the registration contract. A static ready state differs from a task’s can_execute result; process success differs from scientific acceptance.
Methods explain research design, skills implement operations, workflows compose data/skills and projects retain concrete study evidence. General automatic orchestration is not implemented.