Run a READ-ONLY analytical SQL SELECT over your organization's data.
For aggregates, GROUP BY, joins, and trends the other tools can't express.
Write one Postgres SELECT (or WITH ... SELECT); up to 200 rows come back. It
runs against these org-scoped read-only views (qualify with the schema; base
tables are unreachable) and cannot write. Each view, then what each of its
columns holds; filter enum columns only on the listed values:
- assistant_analytics.data_items: One row per catalogued data item (a RHEED video, SEM image, tool log, ...). Deleted items are excluded.
data_id — The item's id.
raw_name — The uploaded file's name. It often encodes a growth id, but the sample's own name is physical_samples.name.
char_source_type — What kind of data the item is. rheed is a RHEED video, rheed_image a single RHEED frame, rheed_xscan an XSCAN recording, tool_state an instrument log, recipe a growth recipe file. Values: rheed, rheed_image, rheed_xscan, arpes, xps, unknown, sem, tool_state, optical, raman, photoluminescence, recipe, ellipsometry, sims, oes, xrd, afm, uv_vis, xrr.
pipeline_status — Where the item's analysis stands. success means results exist; quarantined items wait for someone to start their pipeline; stream_* values belong to live RHEED streams. Values: success, pending, running, error, quarantined, stream_active, stream_paused, stream_interrupted, stream_finalizing, stream_error.
data_status — Always active in practice; nothing sets stale today. Values: active, stale.
physical_sample_id — The sample the item belongs to (physical_samples.id); NULL when unassigned.
upload_datetime — When the file reached the platform (UTC). Often months after acquisition for bulk uploads, so never order growths by it.
user_id — The user who uploaded it.
process_step_id — The process step the item was acquired in (process_steps.id); NULL when it has not been grouped into one.
collected_datetime — Acquisition time (UTC); order growths by this. Never NULL: when the upload could not read a time from the file it holds the upload time, so a value within a few seconds of upload_datetime means unknown.
- assistant_analytics.physical_samples: One row per physical sample (one growth or specimen).
id — The sample's id.
name — The sample's name as users entered it, usually a growth id.
user_id — The user who created it.
- assistant_analytics.process_steps: A sample's ordered processing phases (e.g. an STO deposition, then a BTO one), each grouping the data items acquired during it.
id — The step's id; data_items.process_step_id points here.
physical_sample_id — The sample the step belongs to.
step_index — 0-based position in time order. This IS the process step number; never infer one from an instrument channel.
name — The step's display name.
start_unix_ms_utc — Step start, in milliseconds since the Unix epoch (UTC); to_timestamp(start_unix_ms_utc / 1000) gives a timestamp.
end_unix_ms_utc — Step end, in milliseconds since the Unix epoch (UTC).
- assistant_analytics.projects: One row per project, a named group of samples.
id — The project's id.
name — The project's name.
owner_id — The user who owns it.
- assistant_analytics.project_samples: Which samples each project holds. Join projects, project_samples, process_steps and data_items for project rollups.
project_id — The project (projects.id).
physical_sample_id — The sample (physical_samples.id).
- assistant_analytics.tags: The organization's tag names. No view links tags to data items, so query_data cannot filter by tag.
id — The tag's id.
name — The tag's name.
- assistant_analytics.changepoints: One row per detected changepoint: a window of one data item's signal where it changed. Changepoints of deleted items are included; join data_items to drop them.
id — The changepoint's id.
data_id — The data item it was detected in (data_items.data_id).
label_category — The category a person labelled it with (e.g. Roughening); NULL when unlabelled, which most are.
severity — How significant the detector judged it. Values: info, warning, critical.
score — Detection score from 0 to 1; higher is a stronger change.
window_start_elapsed — Window start, in seconds from the start of the item's signal (not a clock time).
window_end_elapsed — Window end, in seconds from the start of the item's signal.
property_name — The signal it was found in, e.g. intensity_profile, detected_rheed_reconstruction, referenced_strain, temperature.
data_modality — The item's modality as free text. Older RHEED rows say rheed_stationary or rheed_rotating instead of rheed, so filter by modality through data_items.char_source_type, not this column.
- assistant_analytics.tool_channels: One row per tool-log channel whose meaning a person has certified, whatever the vendor named it (tc_t, manipulator_temperature and table_heater can all be substrate_temperature). A channel without a row has no certified meaning: never infer one from its name. Deleted items are excluded.
data_id — The tool log the channel is in (data_items.data_id).
property_name — The channel's name as the log has it.
concept — What the channel measures or controls. Values: substrate_temperature, source_temperature, source_shutter, non_source_shutter, substrate_rotation_rate, substrate_rotation_angle, chamber_pressure, other_vacuum, gas_flow, precursor_source_pressure, plasma_power_forward, plasma_power_reflected, cycle_or_step_count, rheed_signal, qcm_rate, qcm_thickness, growth_rate, heater_output, controller_internal, status_flag, time, other, unknown.
is_setpoint — True for a setpoint (what the controller was asked for); exclude these rows for measured values.
element — A source's chemical symbol (Ba, Ti); NULL unless the channel belongs to an elemental source.
precursor — The precursor a line carries (W(CO)6, DETe); NULL otherwise.
gas — The gas a line carries (H2S, O2); NULL otherwise.
sensor — What reads the channel; NULL when not known. Values: thermocouple, pyrometer, capacitance_manometer, ion_gauge, mass_flow_controller, other.
unit — The channel's unit, from the log when it states one, otherwise the certified one; NULL when neither says.
display_name — A readable name for the channel.
min_value — Lowest reading over the whole log, ramps included.
median_value — Median reading over the whole log, ramps included.
max_value — Highest reading over the whole log, ramps included.
n_values — How many readings the stats cover (NaN readings skipped).
query_data