Plug into Galena.

The lookup answers one question about an address: what does the public record say about its water service pipe, and how sure is that? It is the service behind the address check on this site. It reads only, and it gives no model's answer about any single home.

Base address

https://galena.craftpine.com/lookup/v1/

Methods

GET

Key

None

Limits

60 searches and 40 reports a minute from one address

Schedule 1. Endpoints of the lookup

1. GET /lookup/v1/search

   Parameters

   q, the address, required, at most 200 characters. limit, from 1 to 10 results, 5 when not given.

   The answer

   results, a list. Each result has id, label, city, lat, lon, source. Pass its id to report.

2. GET /lookup/v1/report

   Parameters

   id, from a search result. It has one of four forms: kit: and wat: for City of Kitchener and City of Waterloo addresses, inv:<place>: for a utility's inventory, and geo:<lat>,<lon> for a point on the map.

   The answer

   One answer. Its fields: address, verdict, record, building, area, tests, programme, how_to_check, sources, map, geometry, area_map, water, fix, testing. A field with nothing to say for an address is null, as tests is in the example below.

3. GET /lookup/v1/health

   Parameters

   None.

   The answer

   ok, kitchener_addresses, waterloo_addresses, inventory_records, area_maps, water, facts_sha256, problems.

An error is JSON with one field, error, in plain words. The status is 400 for a missing or bad parameter, and 404 for an id that is not in the list. It is 429 past the limit, and 502 when the map search behind the lookup does not answer. For example, a search with no address answers 400 and {"error":"Type an address."}.

Example: search for Kitchener City Hall

curl 'https://galena.craftpine.com/lookup/v1/search?q=200+King+St+W,+Kitchener&limit=1'

{"results":[{"id":"kit:975827","label":"200 King St W, Kitchener, ON","city":"Kitchener","lat":43.451764,"lon":-80.492419,"source":"City of Kitchener address points"}]}

Example: the report for that result, a few of its fields

curl 'https://galena.craftpine.com/lookup/v1/report?id=kit:975827'

"verdict": {"band": "low_chance", "certainty": "low",
  "headline": "This building was built in 1993, after 1980. Lead is unlikely, though no record names the pipe material."}
"record": {"publisher": "City of Kitchener (Kitchener Utilities)", "material": "not recorded",
  "install_year": 1992, "licence": "Open Government Licence - City of Kitchener"}

Fig. 1. Read live on 2026-11-03 17:36 Toronto time. Kitchener City Hall is a public building. Use public buildings in your own examples and tests.

What happens to an address you send

• The lookup does not log a query or an address. Its server writes no access log, and the web front in front of it keeps no error log, where request addresses would appear.
• When the lookup's own lists do not hold an address, it sends the text you typed to OpenStreetMap's Nominatim search, to place it on the map. It does so at most once a second. It keeps the last 5,000 such searches in memory for up to a day, and never on disk.
• Browsers let pages on craftpine.com call the lookup, and no other site. Call it from your own server or script.
• Each answer names its sources, with their licences, in sources.

Galena publishes what it counted and what its checks found, as files anyone can download. Each file names its source and the licence of that source. A cache may keep a file for 5 minutes.

City results

The index is /data/cities/index.json. It holds generated, cities, not_published, method. Each of its 4 cities has id, name, province_or_state, country, lines, decision, headline, bbox, licence, source_url, updated, retrieved, publisher, report, hexes, tiles.

Each city's folder holds four files:

• report.json, the city's result. Its fields: schema, id, name, province_or_state, country, generated, decision, headline, reasons, metrics, sufficiency, expected_lead_per_cell, estimate, counts, frame, labels_rule, note, method, data_card, bbox, clearance, release_bound, plain, false_non_lead, false_non_lead_field_checked, calibration, equity_failures, caveats, repeat.
• hexes.geojson, a GeoJSON FeatureCollection of map cells. Each cell has cell, lines, recorded_lead, recorded_grr, recorded_not_lead, recorded_unknown, share_unknown.
• evaluation.json, the check's full record. Its fields: schema_version, report_id, dataset_card, model_card, split_strategy, metrics, calibration, conformal, bias_diagnostics, equity, red_team, leakage, drift, gate_decision, versions, seeds.
• SHA256SUMS, the sha256 of the other three.

The folder's own SHA256SUMS lists every file in it.

Schedule 2. City results, with each licence as the files state it

Schedule 2. Read live on 2026-11-03 17:36 Toronto time. When this page was built, every file's sha256 matched both its city's SHA256SUMS and the folder's. Check a download the same way: in its folder, run sha256sum -c SHA256SUMS.

Area files

The index is /data/check/index.json, schema galena.check-index/1. It lists 74 areas. Each area file counts the records in map cells (H3 resolution 8, about 0.7 square kilometres a cell). A cell with fewer than 10 records is left out. The index holds schema, generated, code, unit, min_cell, classes, fields, corners, cities; each area in it has id, name, region, country, file, url, sha256, bytes, counts, cells, totals, suppressed, bbox, source, inputs, notes, left_out. In an area file, each cell is a list in the order its fields gives: cell, lon, lat, lines, lead, grr, not_lead, unknown, corners. Positions are in units of 1/100,000 of a degree (scale). The folder's SHA256SUMS lists every file.

Schedule 3. Licences of the area files, as the files state them

Schedule 3. Read live on 2026-11-03 17:36 Toronto time. Each area's own entry in the index names its publisher, dataset and source address beside the licence.

The app at galena.craftpine.com runs on an HTTP API. A free workspace gives you a key for 7 days and a private space for your own file of service lines. With the key you can take the same steps as the app's own pages. You upload the file, match its columns, train and check a model, map the results, plan field checks and export.

Base address

https://galena.craftpine.com/v1/

Key

Authorization: Bearer <key>, or X-API-Key: <key>

Errors

application/problem+json

Schema

galena-openapi.json

Get a key

curl -X POST https://galena.craftpine.com/v1/public/workspaces

This needs no key and no body. The answer holds key, workspace_id, expires_at, scopes, limits and note. The key is shown once and cannot be recovered, so keep it. It carries these scopes, and never admin: platform:read, platform:write, jobs:read, jobs:write, lsl:read, lsl:write, lsl:promote. A request that says it comes from another site's page is refused, so start a workspace from a script or from the app itself. When the workspace ends, an hourly cleanup deletes its data, its files and its key. Only the record that it existed and its audit trail stay.

Every limit is public, at /v1/public/workspaces/limits. Schedule 4 gives them as read on 2026-11-03 17:36 Toronto time. In its words: “Each workspace is a separate tenant with its own encryption key. No other workspace can read its data. The data is deleted when the workspace expires.”

Schedule 4. Limits of a free workspace

A workspace lasts this many days, then it is deleted with its filesexpires_after_days

7

Datasets a workspace can holdmax_datasets

3

Trainings it can start a daymax_trainings_per_day

5

Jobs it can have waiting or running at oncemax_open_jobs

2

Changes it can make a daywrites_per_day

300

Largest body of a change, other than an uploadmax_write_body_bytes

64 KiB65,536 bytes

Largest file it can uploadmax_upload_bytes

64 MiB67,108,864 bytes

Files one key can upload an houruploads_per_key_per_hour

30

Bytes one key can upload a dayupload_bytes_per_key_per_day

512 MiB536,870,912 bytes

Bytes all free workspaces together can upload a daypublic_upload_bytes_per_day

1 GiB1,073,741,824 bytes

Workspaces one address can start a dayworkspaces_per_address_per_day

3

Workspaces that can be started a day, in allworkspaces_per_day

50

Files it can uploadupload_formats

CSV

Calls a minute, for each key, and for each address without a keyRateLimit-Policy header

60 in 60 s

The steps, in order

A workspace key may make only the writes below. Any other write is refused with 403. A POST that creates a job or a record needs an Idempotency-Key header: a value you choose, of letters, digits and . _ : -, up to 255 characters. Long work returns 202 and a job; follow it until its status is succeeded, failed or cancelled. Lists come in pages, with cursor and limit (1 to 100, 50 when not given).

1. 1. Upload your file of service lines, as CSV

   Call

   PUT /v1/incoming/{idem}

   What you send and get back

   The file's bytes, up to 64 MiB. {idem} is an Idempotency-Key you choose. The answer is 201 with incoming and bytes.

2. Then make the dataset from it

   Call

   POST /v1/datasets

   What you send and get back

   The header Idempotency-Key with the same {idem}, and {"name": "My city", "kind": "service_lines", "format": "csv"}. The answer is a job.

3. 2. Match your columns

   Call

   POST /v1/datasets/{id}/mapping

   What you send and get back

   {"dataset_id": "{id}", "fields": {...}}. fields maps each column name in your file to a Galena field: service_line_id, longitude, latitude, year_built, inventory_category, tract_geoid, block_geoid, notes or material.

4. Follow any job

   Call

   GET /v1/jobs/{id}

   What you send and get back

   The job's id. The answer gives its status, progress, error and result_links.

5. 3. Train and check a model

   Call

   POST /v1/lsl/models

   What you send and get back

   {"algorithm": "lightgbm", "training_snapshot_id": "{dataset id}", "params": {"seed": 1}}. The app offers these algorithm names: year-built-rule, records-at-face-value, logistic-regression, lightgbm, xgboost and catboost. Training also checks the model on lines it did not see.

6. See the models and each check's result

   Call

   GET /v1/lsl/modelsGET /v1/lsl/evaluations/{id}

   What you send and get back

   A model's eval_report_id names its evaluation. Its gate_decision is PASS, FAIL or INSUFFICIENT (not enough data), with the reasons.

7. Promote a model whose check passed

   Call

   POST /v1/lsl/models/{id}:promote

   What you send and get back

   An optional note. A model whose check did not pass is not promoted, and only a promoted model that passed is served.

8. 4. Score the lines and map them

   Call

   POST /v1/lsl/scoresGET /v1/lsl/predictionsGET /v1/lsl/predictions/{service_line_id}GET /v1/lsl/rollupsGET /v1/lsl/estimates

   What you send and get back

   {"model_version_id": "{model id}", "data_snapshot_id": "{dataset id}"}, then read the predictions, one line's prediction, the counts by area (level, such as H3) and the estimates (scope, such as SYSTEM).

9. 5. Plan field checks

   Call

   POST /v1/lsl/inspection-plansGET /v1/lsl/inspection-plans

   What you send and get back

   {"mode": "MIXED", "budget_n": 200, "probability_share": 0.5, "seed": 1}. mode is REPRESENTATIVE, REPLACEMENT, MIXED or LCRI_VALIDATION.

10. 6. Export

    Call

    POST /v1/lsl/exportsGET /files/{key}

    What you send and get back

    {"format": "csv", "model_version_id": "{model id}"}. The app offers csv, geojson, gpkg and lcri_baseline. The finished job links to a signed download, which works until it expires.

A call without a key answers 401, for example {"detail":"Send a bearer token or an X-API-Key.","instance":"/v1/datasets","status":401,"title":"Unauthorized","type":"https://example.invalid/problems/unauthorized"}. Every call takes from a rate limit of 60 calls in 60 seconds, for each key, and for each address that calls without a key. The headers RateLimit-Policy and RateLimit say where you stand.

The schema file

galena-openapi.json describes the API as OpenAPI 3.1. It is generated by the API's own code, for this host. It holds 33 paths and 47 operations, every schema they use is inside the one file, and its server is https://galena.craftpine.com. Version 1.2.0, 343,110 bytes, sha256 98dd5cab656bd1b1c26a3393eff3debc261f1fe36a15d56cfb83ac8213a17895. It also lists operations a workspace key cannot use, such as making API keys or webhooks. The upload, the workspace and the download routes above belong to this host's front and are described on this page, not in the file.

• This website, craftpine.com, has Craftpine's terms of use and its privacy page. Questions about them go to hello@craftpine.com, the address the terms give.
• Each data file names the licence of the data it comes from. For the city results it is in Schedule 2, for the area files in Schedule 3, and for a lookup answer in its sources. Read that licence before you reuse the data.
• Galena's own terms of use and privacy note cover the lookup, the data files, the app's API and its free workspaces. Questions go to hello@craftpine.com.

Each row names the code it was checked against, with the file and the line. It also names the live reads it was checked against, with the status and the time read on 2026-11-03. The app's code, marked app, is as of commit dbb5ce405187. The code of the lookup and the data files, marked lookup, is as of commit 3c6444ac876e. Each is the last commit that changed the files cited. Galena's code is private for now.