feat(drivers): support dataplane custom driver management

[ets]

Summary

This PR introduces a complete custom dataplane driver management workflow to the Qualytics CLI, enabling operators to package and deploy third-party JDBC drivers into the Qualytics platform without manual YAML authoring.

New Commands

Command	Description
qualytics drivers generate	Introspects a JDBC driver JAR and emits a complete, platform-aligned DriverDefinition YAML
qualytics drivers package	Bundles all generated driver YAMLs into a custom-drivers.jar loadable by the Qualytics dataplane

How It Works

drivers generate

• Accepts a JDBC JAR path, a JDBC URL, optional credentials, and extra connection properties
• Compiles and runs a Java probe at runtime to interrogate the driver: metadata, supported features, connection capabilities, and SQL dialect hints
• Auto-detects the dialectClass via JAR ServiceLoader inspection and a catalog of known Spark built-ins — no manual lookup required
• Derives the correct dataSizeLimit for known 32-bit JDBC drivers (Redshift, SQL Server, DB2 → INT_MAX; all others → LONG_MAX)
• Emits a fully-populated DriverDefinition YAML to dist/META-INF/jdbc-drivers/<prefix>.yaml and maintains an index file
• Optionally invokes an AI assistant to resolve any remaining # TODO fields in the YAML before writing

drivers package

• Reads the index of generated driver YAMLs from the dist tree
• Zips the entire dist/ structure into a single custom-drivers.jar
• The resulting JAR can be deployed alongside the corresponding JDBC JARs for the Qualytics platform to pick up at startup

CLI Integration

• A new drivers command group is registered under the top-level qualytics command
• The group is surfaced as a single top-line entry in the CLI help — keeping the main help clean while grouping related subcommands logically

Changes

File	Change
qualytics/cli/generate_driver.py	New module: Java probe, YAML builder, dialect detection, LLM field resolver, generate and package commands (~1,600 lines)
qualytics/qualytics.py	Register drivers command group

Test plan

• Run qualytics drivers generate --jar <path>.jar --url jdbc:postgresql://... --user ... --password ... — confirm YAML written to dist/META-INF/jdbc-drivers/postgresql.yaml with correct dialectClass and dataSizeLimit
• Run qualytics drivers generate against a Redshift JAR — confirm dataSizeLimit: INT_MAX with correct comment
• Run qualytics drivers package after generating — confirm custom-drivers.jar created containing META-INF/jdbc-drivers/ tree
• Run qualytics drivers --help — confirm both generate and package subcommands are listed
• Run qualytics --help — confirm drivers appears as a single top-level entry

🤖 Generated with Claude Code

[greptile-apps]

Greptile Summary

This PR fixes a value/comment mismatch in generate_driver.py where the YAML template was emitting dataSizeLimit: LONG_MAX for Redshift, SQL Server, and DB2 even though those drivers use 32-bit JDBC interfaces that require INT_MAX. The generator now auto-selects the correct constant and matching comment for those driver prefixes at generation time.

• The core intent of the fix is correct and well-motivated.
• The prefix-detection predicate on line 871 uses substring matching (p in prefix.lower()) rather than exact set-membership (prefix.lower() in _int_max_prefixes), which could produce false positives for custom or future driver prefixes that embed "db2", "redshift", or "sqlserver" as substrings.
• dataSizeLimit is appended to todo_fields unconditionally (line 879) even when the value is auto-selected as INT_MAX; all other auto-detected fields use detected_fields, making this inconsistent.
• The LICENSE copyright year change (2023→2026) is unrelated to the stated fix.

Confidence Score: 4/5

Safe to merge; the core fix is correct and both issues identified are low-risk

No real-world JDBC prefix would trigger the substring false-positive today, and the todo_fields inconsistency has no user-visible impact since the display layer uses its own always_todo list. The fix correctly resolves the stated mismatch between emitted value and comment for 32-bit drivers.

qualytics/cli/generate_driver.py — specifically the prefix-detection predicate on line 871

Important Files Changed

Filename	Overview
qualytics/cli/generate_driver.py	New generate-driver CLI file; INT_MAX/LONG_MAX auto-selection fix is correct but uses substring matching that could produce false positives for non-standard JDBC prefixes
LICENSE	Copyright year updated 2023→2026; unrelated to the stated fix but correct
qualytics/qualytics.py	Wires drivers_app into the main CLI app; no issues

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[generate-driver invoked] --> B[_extract_prefix\nextract jdbc: scheme token]
    B --> C{prefix in 32-bit set?\nredshift / sqlserver / db2}
    C -- Yes --> D[dataSizeLimit = INT_MAX\ncomment: older 32-bit driver]
    C -- No --> E[dataSizeLimit = LONG_MAX\ncomment: TODO review]
    D --> F[_build_yaml assembles YAML]
    E --> F
    F --> G[Write .yaml file to disk]
    G --> H[_collect_todo_fields\nscan YAML for TODO: markers]
    H --> I{LLM-assist\navailable?}
    I -- Yes --> J[LLM fills remaining TODO fields]
    I -- No --> K[Prompt user to review manually]

Loading

_{Reviews (1): Last reviewed commit: "fix(generate-driver): use INT_MAX for Re..." | Re-trigger Greptile}