bohrium-database

name	bohrium-database
description	Operate Bohrium private databases with bohrium-open-sdk. Use when: the user asks to list private/authorized Bohrium databases, inspect tables or schemas, query/filter/export rows, insert/update/delete records, create tables, or alter table columns using db_ak/table_ak IDs. NOT for: public materials databases such as crystal/electrolyte/positive-electrode databases, internal shrimp PostgreSQL operations, general file storage, datasets, or compute jobs.
metadata	{"openclaw":{"primaryEnv":"BOHR_ACCESS_KEY"}}

SKILL: Bohrium Private Database

Overview

Use bohrium-open-sdk to operate Bohrium private databases that the user can access. "Private databases" here includes databases created by the user, team-authorized databases, and any database visible to the current AK. It does not include the public materials database catalog such as crystal, electrolyte, or positive-electrode databases. Common identifiers:

• db_ak: database access key, for example 669ng
• table_ak: table access key, for example 669ng00

Prefer the bundled database_manager.py for routine operations. Use direct SDK calls only for complex filters, data cleaning, or custom exports.

Authentication and Install

Read BOHR_ACCESS_KEY from the environment. Never print, log, or hardcode it.

test -n "$BOHR_ACCESS_KEY" || echo "BOHR_ACCESS_KEY is missing"
python3 -m pip install "bohrium-open-sdk>=1.0.5" -q

If BOHR_ACCESS_KEY is missing, ask the user to configure their AccessKey before calling the API.

Quick CLI

Run these commands from the skill directory, or replace database_manager.py with the actual script path:

python3 database_manager.py list-dbs
python3 database_manager.py list-tables 669ng          # shows up to 30 fields per table by default
python3 database_manager.py detail 669ng00

Query rows:

python3 database_manager.py query 669ng00 \
  --where '[{"field":"name","op":"LIKE","value":"sample"}]' \
  --order-by createTime --desc --page 1 --page-size 20

Insert rows:

python3 database_manager.py insert 669ng00 \
  --rows '[{"name":"new sample","value_a":1.5,"value_b":2.3}]'

Update and delete preview the target rows by default. Add --yes after confirming:

python3 database_manager.py update 669ng00 \
  --where '[{"field":"name","op":"EQ","value":"old value"}]' \
  --set '{"name":"new value","value_a":99}' \
  --yes

python3 database_manager.py delete 669ng00 \
  --where '[{"field":"name","op":"EQ","value":"delete me"}]' \
  --yes

Create or alter tables:

python3 database_manager.py create-table 669ng "new_table" \
  --schema '[{"title":"sample_name","dataType":"str"},{"title":"mass","dataType":"num"}]'

python3 database_manager.py alter-table 669ng00 \
  --schema '[{"title":"sample_name","dataType":"str"},{"title":"mass","dataType":"num"},{"title":"new_column","dataType":"str"}]' \
  --yes

--rows, --where, --set, and --schema accept either JSON strings or @/path/to/file.json.

Direct SDK Template

import os
from bohrium_open_sdk.db import SQLClient

client = SQLClient("", os.environ["BOHR_ACCESS_KEY"], "https://openapi.dp.tech")

List Databases

data = client.base().Dbs()
for db in data.get("list", []):
    print(db["name"], db["ak"], db.get("createTime", ""))

List Tables and Schema

result = client.db_with_ak("669ng").Tables()
for table in result.get("tables", []):
    fields = ", ".join(f"{f['name']}({f['type']})" for f in table.get("fields", []))
    print(table["name"], table["tableAk"], fields)

detail = client.table_with_ak("669ng00").Detail()
for field in detail.get("fields", []):
    print(field["name"], field["type"])

Query

from bohrium_open_sdk.db import Op, Where

count, rows = (
    client.table_with_ak("669ng00")
    .Where("price", Op.GTE, 10)
    .And(Where("name", Op.LIKE, "sample"))
    .order(key="price", is_asc=False)
    .page(1)
    .page_size(20)
    .Find()
)
print(count, rows[:20])

Full Export

from bohrium_open_sdk.db.func import db_query_full

table = client.db_with_ak("669ng")
table.table_ak = "669ng00"
count, df = db_query_full(table, batch_size=1000)
df.to_csv("bohrium_table_export.csv", index=False)
print(f"exported {count} rows")

Filter Format

database_manager.py --where accepts an array of conditions:

Field	Required	Description
field	Yes	Column name, case-sensitive
op	Yes	Operator, see below
value	Yes	Filter value
join	No	From the second condition onward: AND or OR; default AND

Operators:

Operator	Meaning	Value type
EQ	Equal	string or number
NEQ	Not equal	string or number
GT / GTE	Greater than / greater or equal	number
LT / LTE	Less than / less or equal	number
BETWEEN	Closed range	"1,100" or [1, 100]
LIKE	String contains	string
IN / NIN	In / not in	array
ELEMENTMATCHOFNUM	Numeric array element range	{"min":0,"max":10}

Column Types

When creating or altering tables, schema items use {"title": "column_name", "dataType": "type"}.

Type	Description
str	String
num	Number
smiles	SMILES string
img	Image object, usually uploaded to Tiefblue first
file	File object, usually uploaded to Tiefblue first
file_arr	Array

Altering a table requires the complete new schema. Omitting an existing column usually means deleting it and may lose data.

Safety Rules

• Preview target rows or target schema before deleting, updating, or altering tables.
• For batch inserts, updates, or deletes affecting more than 100 rows, explain the impact and ask for confirmation first.
• If a query returns more than 20 rows, show only the first 20 rows and the total count unless the user asks for export.
• For wide tables, use list-tables --max-fields 30 for an overview, then use detail for the full schema of one table.
• Prefer filters for wide or large tables; unfiltered scans may time out. Use export when the user needs the full table.
• For image and file columns, show short fields such as name; do not show full signed URLs.
• Hide metadata fields such as _id, audit_id, createTime/create_time, updateTime/update_time, authors, owner_id, status, and internal audit/permission fields unless requested.
• Do not expose complete API responses; they may contain internal IDs, temporary paths, or signed links.

Error Handling

Signal	Handling
BOHR_ACCESS_KEY missing	Ask the user to configure AccessKey
401 / 403 / code: 2000	Treat as invalid key or missing permission
Column KeyError, or unexpectedly empty query	Run detail to verify exact column names and types
Table or database not found	Use list-dbs and list-tables to confirm db_ak / table_ak
Network timeout	The helper retries once; if it still fails, narrow the filter/page or use export