feat: add 12 more databases through configuration only

[Chat2DB-Pro]

Third batch on the config-only path, following #2633 and #2644. No new Java plugin code — all entries live in generic.json.

Bring their own driver

Database	Driver class	Port
SAP HANA	com.sap.db.jdbc.Driver	30015
MonetDB	nl.cwi.monetdb.jdbc.MonetDriver	50000
Altibase	Altibase.jdbc.driver.AltibaseDriver	20300
Ocient	com.ocient.jdbc.JDBCDriver	4050
InterSystems IRIS	com.intersystems.jdbc.IRISDriver	1972
NuoDB	com.nuodb.jdbc.Driver	—
OpenSearch	org.opensearch.jdbc.Driver	9200
DolphinDB	com.dolphindb.jdbc.Driver	8848
Db2 for IBM i	com.ibm.as400.access.AS400JDBCDriver	—

Reuse a jar we already bundle

Database	Reuses	Why	Port
GreptimeDB	MySQL driver	speaks the MySQL wire protocol	4002
Apache Spark	Hive driver	speaks HiveServer2	10000
TimechoDB	IoTDB driver	is an IoTDB distribution	6667

Two entries get non-default handling: OpenSearch is routed to the Elasticsearch syntax plugin, and TimechoDB inherits the IoTDB entry's quoting rules (identifierQuotes: none, table-only qualification) because dots are structural in its path-based references and quoting any part would break the SQL.

How the values were established

Same method as the previous two batches — nothing written from memory:

• Driver classes — every jar was downloaded from Maven Central and its META-INF/services/java.sql.Driver read directly. MonetDB ships no service file, so its class was confirmed by listing the jar contents; the class matches what DBeaver declares.
• Coordinates — found by reverse-searching Maven Central on the fully-qualified driver class (fc: queries) rather than guessing artifact names.
• Ports — from DBeaver's own plugin.xml driver metadata (defaultPort).

Deliberate omissions

• Apache Kyuubi — the published kyuubi-hive-jdbc jar is unshaded and cannot load standalone. Shipping it would hand users a driver that fails at connect time.
• Netezza, SQream, Timeplus — no driver on Maven Central (zero results).
• Athena, Cloudera Impala — vendor-distributed drivers, not on Maven Central.

These belong to the "user supplies their own jar" path rather than the config-only path.

Progress against DBeaver

A gap analysis against DBeaver's repository (183 driver declarations across 75 ext plugins → 125 distinct products) put us at 44 supported. After #2644 and this PR we are at 69, with the gap down from 82 to 57.

What remains is mostly not reachable by configuration alone: ~23 commercially licensed drivers, ~11 non-JDBC engines needing dedicated plugins, and a set of cloud services requiring SDKs and credentials. The cheap wins from this mechanism are largely spent; the highest-leverage next step is a "bring your own driver jar" flow, which would unlock the licensed group in one change.

Verification

• generic.json grows from 26 to 38 entries; additive diff only, no reformatting
• generic plugin unit tests pass, including GenericPluginConfigLoadTest which parses the file
• Live connections were not exercised; these are configuration entries and need the respective engines to test end to end

🤖 Generated with Claude Code

https://claude.ai/code/session_01Rq65mLom8mVWv1F6VzejN4

[github-actions]

Reviewed 9014d9: 0 critical, 1 high, 0 medium findings. Request changes because the OpenSearch default JDBC URL is not accepted by the configured driver, making the newly added connection path unusable with its shipped template. Coverage was limited to the single GitHub-provided configuration diff plus PR metadata and review history; no tests or runtime checks were run.

Warning

Firewall blocked 1 domain

The following domain was blocked by the firewall during workflow execution:

• awmgmcpg

To allow these domains, add them to the network.allowed list in your workflow frontmatter:

network:
  allowed:
    - defaults
    - "awmgmcpg"

See Network Configuration for more information.

Generated by AI Pull Request Reviewer for #2648 · 7.12 AIC · ⌖ 2.8 AIC · ⊞ 17.2K

[Chat2DB-Pro]

Fixed in 1a4c4bd15, but the finding needs correcting on the reason.

The stated reason does not hold. I ran the driver's own UrlParser.parseProperties against the shipped template:

jdbc:opensearch://localhost:9200         -> {port=9200, host=localhost, usessl=false}
jdbc:opensearch://https://localhost:9200 -> {port=9200, host=localhost, usessl=true}
jdbc:opensearch://http://localhost:9200  -> {port=9200, host=localhost, usessl=false}

The scheme is optional and defaults to plain HTTP — which is also why DBeaver brackets it as [{scheme}://] in its own sample URL. So the URL was not the problem.

The conclusion was right, for a different reason. Loading the driver and calling getConnection shows it fails before reaching the network:

NoClassDefFoundError: com/amazonaws/auth/AWSCredentialsProvider

The published jar is unshaded. That is a real defect and the entry had to go.

The check was worth generalising. I ran the same load-and-connect probe across all 12 entries in this PR, which found two more the review did not:

Entry	Result
OpenSearch	fails on AWS SDK
Ocient	fails on log4j-core
DolphinDB	fails on com.xxdb.data.Vector
HANA, MonetDB, Altibase, IRIS, NuoDB, Db2 for IBM i	reach the network
GreptimeDB, Apache Spark, TimechoDB	reuse already-bundled drivers

My original verification only confirmed each jar declares a java.sql.Driver service, which does not prove the driver can load. That is exactly the defect Apache Kyuubi was excluded for in this same PR, so I applied the same standard.

Resolution:

• DolphinDB kept — its closure is just two jars (dolphindb-javaapi, slf4j-api), added via the existing multi-jar mechanism (same shape as the Derby entry). Re-probed: reaches the network.
• OpenSearch and Ocient removed — AWS SDK and netty/commons-text closures respectively, not worth pinning by hand.

I also probed the drivers in #2649 and the already-merged #2644; all of those reach the network, so no further entries are affected.

[github-actions]

Reviewed 1a4c4bd: 0 critical, 0 high, 0 medium findings. The prior blocking finding is superseded because the OpenSearch entry is no longer present in the current diff; no merge-blocking issue was found in the remaining changed lines. Merge recommendation: proceed. Coverage was limited to the single current configuration patch plus pull request metadata and review history; no tests or runtime checks were run.

Warning

Firewall blocked 1 domain

The following domain was blocked by the firewall during workflow execution:

• awmgmcpg

To allow these domains, add them to the network.allowed list in your workflow frontmatter:

network:
  allowed:
    - defaults
    - "awmgmcpg"

See Network Configuration for more information.

Generated by AI Pull Request Reviewer for #2648 · 6.27 AIC · ⌖ 3.04 AIC · ⊞ 17.2K

Superseded by updated review from same workflow.