)]}' { "log": [ { "commit": "80f77eaba23fc72990576f101ee7c97b96fe7757", "tree": "51cdbcdba7a3d763e7bc18b0fa39301787c95080", "parents": [ "67c3681173e6e5f570d99811d4b205e1c8e312be" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 17:34:28 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 17:34:28 2026 -0400" }, "message": "Add server.json for the MCP Registry\n\nManifest for registry.modelcontextprotocol.io, pointing at the PyPI\npackage. Namespace io.github.apache is verifiable today through the\napache GitHub org; org.apache would need DNS verification of apache.org\nvia ASF Infra.\n\nNot published yet: publishing needs the PyPI release to exist first.\n" }, { "commit": "67c3681173e6e5f570d99811d4b205e1c8e312be", "tree": "efb5648c73f8bfcfb1af12d2d97bd27425e345f7", "parents": [ "bd9e925d642e388cd6e815b9a51347f714d3e8d7" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 17:25:10 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 17:25:10 2026 -0400" }, "message": "Apply ruff format\n\nMechanical only; no behaviour change. The lint job runs\n\u0027ruff format --check .\u0027, which was failing on 11 files, so main would\nhave gone red on its first CI run.\n" }, { "commit": "bd9e925d642e388cd6e815b9a51347f714d3e8d7", "tree": "322e5d58b13d2d1e5e8b4aa644a92b92fbdb1781", "parents": [ "0fde75237d083ad65d839ed96eb7f58d792fafae" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 17:24:06 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 17:24:26 2026 -0400" }, "message": "Fix lint findings under a stricter ruff rule set\n\nReplace two noqa directives with explicit zip(strict\u003dFalse), so the\nintent holds regardless of which rules are enabled.\n\nNarrow is_show_command\u0027s bare except to the exceptions sqlglot can\nraise. Returning False on an unexpected error meant no SHOW filtering,\nwhich fails open; an unexpected error now propagates to run_query\u0027s\nhandler and becomes a rejection.\n" }, { "commit": "0fde75237d083ad65d839ed96eb7f58d792fafae", "tree": "a80c83f588f37807042afefeb12a40306c42bd77", "parents": [ "7ccda928503dc8ca34601e118126e02778d20127", "81c02ae3f7c031a928ee166f030b2a50e2fdf534" ], "author": { "name": "Charles S. Givre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 17:20:53 2026 -0400" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Aug 12 17:20:53 2026 -0400" }, "message": "Merge branch \u0027feat/drill-mcp-server\u0027 into main" }, { "commit": "7ccda928503dc8ca34601e118126e02778d20127", "tree": "c0a60447a9872c2c01d86c89c7f15b58cec90a9d", "parents": [ "1c59dfd66e4070431e4f03fae4513f67a00b5d35", "dcae48eb06651f937e1997c6e2d869e32d91b95d" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 17:14:38 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 17:14:38 2026 -0400" }, "message": "Add Apache Drill MCP server\n\nMCP server exposing nine read-and-manage tools over Apache Drill, with a\ndeny-by-default SQL policy layer between the model and the cluster.\n\n- REST backend by default, optional JDBC backend for Kerberos\n- Writes denied unless a plugin is explicitly allowlisted; ALTER SYSTEM and\n storage-plugin editing are not implemented at all\n- Configurable hidden schemas, enforced in the guard, the metadata tools,\n and by filtering the rows of every SHOW command\n- Secrets redacted from storage-plugin and profile output, not disableable\n- Column discovery branches on storage plugin type, following the\n sqlalchemy-drill dialect\n- 343 tests, no Drill cluster or JVM required\n- CI on Python 3.11-3.13 with ruff linting\n" }, { "commit": "dcae48eb06651f937e1997c6e2d869e32d91b95d", "tree": "c0a60447a9872c2c01d86c89c7f15b58cec90a9d", "parents": [ "fe9465bcb416756ed02bbf60a77ce4348ed0314f" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:44:15 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:44:15 2026 -0400" }, "message": "Add GitHub Actions CI workflow\n\nRun pytest on Python 3.11/3.12/3.13 (no jdbc/kerberos extras, proving\nthe suite needs no JVM) and ruff check/format on pull requests and\npushes to main. Add a CI status badge to the README.\n" }, { "commit": "fe9465bcb416756ed02bbf60a77ce4348ed0314f", "tree": "b052f3842c9cf0976b85984c8ebdbdaa3c6ffe45", "parents": [ "665a7e258910ef5b738ab7fcdfdb2c1d29f37869" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:43:40 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:43:40 2026 -0400" }, "message": "Add ruff for linting and formatting\n\nConfigure ruff (E, F, I, B, UP, C4) with a 100-column line length that\nmatches the codebase\u0027s existing wrapping style, add it to the dev\nextra, and bring the tree into compliance.\n\nTwo zip() calls without strict\u003d (client_jdbc.py, client_rest.py) are\nnoqa\u0027d rather than changed: both are length-matched by construction\n(a JDBC row tuple to its column list; a metadata list already padded\nto len(columns) two lines above), so strict\u003dTrue would only add an\nunreachable error path to reviewed code. One long f-string in\nclient_rest.py\u0027s auth-failure message is noqa\u0027d rather than reflowed\nfor a 4-character overage.\n" }, { "commit": "665a7e258910ef5b738ab7fcdfdb2c1d29f37869", "tree": "534c253a0b5a80b6394c0be088cde1b922f20ea4", "parents": [ "87b61ecacb76f901e62e919912387d3c0625b512" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:34:19 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:34:19 2026 -0400" }, "message": "Fix test env leakage and enforce row cap in the tool layer\n\nAdd an autouse conftest fixture that strips DRILL_* env vars before\neach test, making the suite hermetic regardless of the ambient shell\nenvironment (test_cli_flags_reach_the_config goes through main(),\nwhich legitimately reads os.environ, so per-test env\u003d{} alone cannot\nfix it).\n\nAlso slice DrillTools.run_query\u0027s rows to the effective max_rows\nlimit after hidden-schema filtering. Both backend clients already cap\nrows, but run_query is the last chokepoint before the model, so it\nshould not simply trust what the client returns.\n" }, { "commit": "87b61ecacb76f901e62e919912387d3c0625b512", "tree": "a06a3786555f747da7519ea86186b8ece057250f", "parents": [ "9e879e08c2334310c46847f4828c7d77febe9d55" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:28:16 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:28:16 2026 -0400" }, "message": "chore: remove unused integration pytest marker\n\nNo test uses the integration marker or is gated by -m \u0027not\nintegration\u0027; it registers a category the suite doesn\u0027t have. Removed\nrather than backfilled with integration tests, per scope.\n" }, { "commit": "9e879e08c2334310c46847f4828c7d77febe9d55", "tree": "2fe2f17eecc4c54122287b10f1c0fdf8a1bdf7e0", "parents": [ "80cdb1fbd838111c3586d86a20ab3751151c3a44" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:28:09 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:28:09 2026 -0400" }, "message": "test: isolate ambient environment in test_server/client_rest/client_jdbc\n\nOnly test_config.py isolated ambient environment. The other three test\nfiles called load_config(overrides\u003d...) without env\u003d{}, so an exported\nDRILL_AUTH (or other DRILL_* var) on a contributor\u0027s machine could\nbreak the suite at fixture time -- verified: DRILL_AUTH\u003dbasic\nDRILL_USER\u003dx DRILL_PASSWORD\u003dy in the environment fails 42 tests\nwithout this fix. Matters for an ASF repo where contributors have real\nclusters configured via environment variables.\n" }, { "commit": "80cdb1fbd838111c3586d86a20ab3751151c3a44", "tree": "ee63dde4a74987205ba08e0649d98ec7e6c3a0b7", "parents": [ "0322cf06b5d060c40ebcb5ea1adb6ee7086f71f0" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:27:47 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:27:47 2026 -0400" }, "message": "fix: catch DrillError alongside ConfigError in main()\n\ndrill-mcp --auth kerberos without the kerberos extra died with a raw\ntraceback instead of the clear message client_rest.py raises, because\nbuild_client\u0027s DrillError is only reachable once build_server(config)\nruns the client constructor. Widen main()\u0027s try block to cover that\ncall and its except clause to catch DrillError too.\n" }, { "commit": "0322cf06b5d060c40ebcb5ea1adb6ee7086f71f0", "tree": "1d413cfbf7750dd73a4dfc9491016641b4c51e47", "parents": [ "a5329bb8206a9651f341334a8bbf4f4fc876489b" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:27:38 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:27:38 2026 -0400" }, "message": "fix: drop pydantic\u0027s input_value from ConfigError messages\n\nConfigError(str(exc)) on a pydantic ValidationError embeds\ninput_value\u003d... for every error -- e.g. an unquoted `password: 12345`\nin a YAML config produces \"input_value\u003d12345\" verbatim, which main()\nprints to stderr and any log capturing it. Build the message from\nexc.errors()\u0027s loc/msg fields only, dropping the input value.\n" }, { "commit": "a5329bb8206a9651f341334a8bbf4f4fc876489b", "tree": "e1762d9463422f5d8967c47548ed1df2a268a15c", "parents": [ "d660a67bc273c055850243e25b9a0bd5485e9370" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:27:33 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:27:33 2026 -0400" }, "message": "fix: redact and hidden-schema-filter list_profiles/get_profile output\n\nlist_profiles and get_profile were the only Drill-sourced tool outputs\nbypassing both redact() and hidden-schema filtering. Profiles are\ncluster-wide: they carry other users\u0027 query text (a hidden schema name\ncan leak out there as data, the enumeration path the guard and the\nmetadata-tool filtering were built to close) and a full profile embeds\nDrill\u0027s serialized physical plan, which for JDBC/HTTP plugins can\ncarry plugin configuration.\n\nPass both through redact(); drop list_profiles entries, and refuse\nget_profile, when the profile\u0027s query text case-insensitively names a\nhidden_schemas entry. Updates docs/tools.md, which said get_profile\u0027s\npayload was returned unmodified.\n" }, { "commit": "d660a67bc273c055850243e25b9a0bd5485e9370", "tree": "43a0d4812f55cc69a3f129a7a4cc2a566e604ace", "parents": [ "b3caedc150162804c25f5dbbb0202ca5e3cdb6b4" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:26:56 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:26:56 2026 -0400" }, "message": "clarify: note why test_leaves_innocuous_keys_alone\u0027s input has no embedded credential\n" }, { "commit": "b3caedc150162804c25f5dbbb0202ca5e3cdb6b4", "tree": "f9b1551f2b037ee82f0f46f29c8915eba303e6ef", "parents": [ "a843a7258ecb044aed25a308f60c70575deffa07" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:26:37 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:26:37 2026 -0400" }, "message": "fix: redact secrets embedded inside string values, not just keys\n\nredact() only matched sensitive-looking keys, so a secret embedded\ninside an ordinary-looking value survived untouched -- e.g.\n{\"connection\": \"s3a://AKIA:secret@bucket\"} or a JDBC-style URL with a\npassword\u003d query parameter, both realistic Drill storage-plugin config\nshapes. Scrub userinfo out of URL-shaped values and redact\npassword/secret/api_key/token query parameters after the key check.\n\nUpdates test_leaves_innocuous_keys_alone\u0027s sibling assertion, which\nencoded this bug (connection strings without embedded credentials are\nstill left alone; only embedded credentials are now redacted).\n" }, { "commit": "a843a7258ecb044aed25a308f60c70575deffa07", "tree": "c8966d15264f0535993737abf02f468070aadb66", "parents": [ "df7c9d0d9d89546f5aac5e984360bd44d14ed3b8" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:26:29 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:26:29 2026 -0400" }, "message": "fix: enforce max_rows client-side in RestClient.query\n\nRestClient.query sent autoLimit to Drill and returned payload[\"rows\"]\nunsliced, so the row cap depended entirely on Drill honoring that\nfield. JdbcClient.query already enforces the cap client-side via\nfetchmany(max_rows); slice REST results the same way as defense in\ndepth, keeping autoLimit as the efficient server-side path.\n" }, { "commit": "df7c9d0d9d89546f5aac5e984360bd44d14ed3b8", "tree": "2d9508cd433bfdad0891ec3e980f94b2474aed0d", "parents": [ "7424700f53f52e7f836ebc95b32599a866aaface" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:25:59 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:25:59 2026 -0400" }, "message": "fix: strip URL userinfo from REST client error messages\n\nconfig.url is free-form and unvalidated; a password embedded there as\nuserinfo (e.g. http://alice:s3cret@drill:8047) previously reached the\nmodel verbatim through DrillError messages raised by RestClient\n(connection failures, auth failures, JSON-decode errors). Add a\n_safe_url helper, mirroring the defense client_jdbc.py\u0027s _jdbc_url\nalready applies, and use it at every site that echoes config.url.\n" }, { "commit": "81c02ae3f7c031a928ee166f030b2a50e2fdf534", "tree": "abe9aaebc2c011c65ff0279592dc9181447681e1", "parents": [ "7424700f53f52e7f836ebc95b32599a866aaface" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:24:19 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 16:24:19 2026 -0400" }, "message": "Various fixes\n" }, { "commit": "7424700f53f52e7f836ebc95b32599a866aaface", "tree": "3f8187c90ac6efd4bdcb2924c0b8981517583766", "parents": [ "523e7e98211f7d27ccf24da6e3d8d4b3834dfcab" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 15:54:44 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 15:54:44 2026 -0400" }, "message": "docs: set the NOTICE copyright holder\n" }, { "commit": "523e7e98211f7d27ccf24da6e3d8d4b3834dfcab", "tree": "51efe11bf49bd7bf8878539a56803fb3f4d92401", "parents": [ "ee674e7adc6564843c7ca65593c30d25cbff5634" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 15:52:08 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 15:52:08 2026 -0400" }, "message": "docs: remove planning scaffolding, add NOTICE, full README and docs/\n\nRemoves docs/superpowers/ (AI planning/spec material, already archived\noutside git) so it does not ship. Adds a minimal NOTICE alongside\nLICENSE, rewrites README.md as the project\u0027s front door, and adds\ndocs/tools.md and docs/configuration.md as the full reference for the\nnine tools and every config key, derived from the source.\n" }, { "commit": "ee674e7adc6564843c7ca65593c30d25cbff5634", "tree": "34ab1efa60339a247f24e4cf91e3d585a62a36bf", "parents": [ "6ea9a5311c24a5aa9f08a11161b905da6cc02b61" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 15:23:57 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 15:23:57 2026 -0400" }, "message": "fix: correct README config precedence and describe_table claims; tidy imports and typing\n\nREADME stated CLI \u003e config file \u003e environment; the code (and config.py\u0027s\nown docstring) has always applied CLI \u003e environment \u003e config file, later\noverriding earlier -- fixed the sentence and called out the credential\nimplication explicitly. Dropped the pytest -m integration line, since no\ntest currently carries that marker. Corrected the describe_table tool-table\nentry and column-discovery section: DESCRIBE reports nullability, the probe\npath does not and never has claimed to sanitise error text (that claim was\nnever true and the code already propagates Drill\u0027s error unchanged on both\npaths). Moved tests/test_server.py\u0027s mid-file imports to the top block.\nbuild_client is now typed RestClient | JdbcClient via a TYPE_CHECKING-only\nimport, keeping the lazy runtime import (and the JVM-free default install)\nunchanged.\n" }, { "commit": "6ea9a5311c24a5aa9f08a11161b905da6cc02b61", "tree": "85583db612fef79224cc3732a241f5c9aec95fc0", "parents": [ "c3d750904b55c956a3a7144f313fd91954badd54" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 15:20:43 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 15:20:43 2026 -0400" }, "message": "revert: stop suppressing Drill\u0027s error text on the probe path\n\nThe premise for the earlier sanitisation (Drill errors can embed sampled\ncell content) was fabricated and unverified; the Drill maintainer confirmed\nDrill\u0027s errors do not do this. The probe\u0027s failure now propagates Drill\u0027s\nerror text unchanged, same as _describe_columns and fetch_plugin_type, so a\nmissing table, a permissions failure, and a genuine data error are\ndistinguishable again instead of collapsing into one opaque message.\n" }, { "commit": "c3d750904b55c956a3a7144f313fd91954badd54", "tree": "d293e059a446cf2cc14c931c9b38a04b095e1802", "parents": [ "07258a49d7374528613830d8c369d16aa8c678b0" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 15:11:24 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 15:11:24 2026 -0400" }, "message": "fix: raise mcp and pydantic dependency floors to the true minimums\n\nmcp\u003e\u003d1.2.0 permitted resolving to a 1.x release that lacks\nmcp.server.mcpserver.MCPServer entirely, which this code imports; raise\nthe floor to mcp\u003e\u003d2.0 (no upper cap: a future rename should fail loudly\nat import, not be pre-emptively blocked). mcp\u003e\u003d2.0 itself transitively\nrequires pydantic\u003e\u003d2.12.0, so the declared pydantic\u003e\u003d2.6 was never a\nreachable floor in practice; raised to match. sqlglot\u003e\u003d25.0, httpx\u003e\u003d0.27,\nand PyYAML\u003e\u003d6.0 were each verified against their exact declared floor\nand left unchanged.\n" }, { "commit": "07258a49d7374528613830d8c369d16aa8c678b0", "tree": "313235137465b322b59895b0203389d4192e814a", "parents": [ "c92b58b09a8b16a38c1452822e52a5764911b3c3" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 14:21:25 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 14:21:25 2026 -0400" }, "message": "docs: correct sqlglot warning-count claim in main() comment\n" }, { "commit": "c92b58b09a8b16a38c1452822e52a5764911b3c3", "tree": "4514aaa43770fa6052916f2a0454e748b292cb80", "parents": [ "64e822dcace43ab4a5599598a858079440e5fb78" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 14:21:12 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 14:21:12 2026 -0400" }, "message": "feat: server wiring, CLI entry point, and documentation\n\nWires DrillTools\u0027 nine bound methods into the installed mcp SDK\u0027s\nMCPServer (mcp.server.fastmcp.FastMCP no longer exists as of mcp\n2.0.0, which is what this environment has installed; MCPServer is its\nreplacement and keeps the same add_tool/_tool_manager/run shape), adds\nthe drill-mcp CLI entry point with eager config validation, and\ndocuments actual server behavior in README.md and drill.example.yaml.\n" }, { "commit": "64e822dcace43ab4a5599598a858079440e5fb78", "tree": "6a971f190bc03512d1f6777f4ef6c18302b244e3", "parents": [ "b4671dc76d025c5b800c1a908d3d36e1acca4f89" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 13:55:37 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 13:55:37 2026 -0400" }, "message": "docs: drop the view lookup from Task 11\n\nThe dialect\u0027s view branch works around its own format_drill_table, which\nsplits a dotted table name. Our quoting helpers produce the same form for\nviews and files, so the lookup was unused surface.\n" }, { "commit": "b4671dc76d025c5b800c1a908d3d36e1acca4f89", "tree": "08b3b2492ba9cf41c69ad768990c956983218005", "parents": [ "83e399c1944dc61a5a036dbcfc94d1428ff128a5" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 13:50:37 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 13:50:37 2026 -0400" }, "message": "fix: close probe error/repr leaks and drop the unused view lookup\n\n- JDBC metadata now reads a DBAPITypeObject\u0027s .values instead of\n stringifying it, matching sqlalchemy-drill\u0027s get_columns.\n- The dynamic-schema probe no longer surfaces Drill\u0027s raw error text on\n failure (it can embed sampled cell content); it also now raises instead\n of silently returning [] when a probe finds zero rows.\n- Removed fetch_view_names: the view/file quoting split it existed to\n support was already proven unnecessary, so it had no caller.\n- QueryResult.metadata is annotated list[str | None].\n" }, { "commit": "83e399c1944dc61a5a036dbcfc94d1428ff128a5", "tree": "804a67fadbc9216e76409e8828186452675c1e2d", "parents": [ "6fd9f1f1b2589ed170b0e631be76784c8c6a2775" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 13:21:48 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 13:21:48 2026 -0400" }, "message": "Various commits\n" }, { "commit": "6fd9f1f1b2589ed170b0e631be76784c8c6a2775", "tree": "ee3d4ebe4c18b06586ed06e7d762e16c5656f63a", "parents": [ "65148076b07da6451087a82e520817d3cf9bb9d2" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 13:14:08 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 13:14:08 2026 -0400" }, "message": "feat: follow the dialect methodology for dynamic-schema plugins\n\nfetch_columns now branches on plugin TYPE: file/mongo/splunk probe with\nSELECT ... LIMIT 1 (schema discovered at read time, DESCRIBE cannot\nanswer), http raises an explanatory error, everything else uses DESCRIBE.\nQueryResult gains metadata (per-column type strings) on both REST and\nJDBC backends. fetch_plugin_type resolves a bare plugin name by matching\nin Python instead of LIKE, avoiding a wildcard-injection surface.\n" }, { "commit": "65148076b07da6451087a82e520817d3cf9bb9d2", "tree": "46dd2391f08621bbdffad20635e93bc0575ec27f", "parents": [ "9e84eebc4cf1f6eb1ffd56bea7aa7169c1b4e233" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 12:57:12 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 12:57:12 2026 -0400" }, "message": "fix: stop classifying SHOW spellings, filter all SHOW output instead\n\nis_show_schemas\u0027s comment-stripping regex was itself bypassable: non-greedy\n/\\*.*?\\*/ matching stops at the first */, and a stray */ with no opener\nstrips nothing, so \u0027SHOW /* /* nested */ */ SCHEMAS\u0027 and \u0027SHOW */ SCHEMAS\u0027\nboth leaked hidden schemas unfiltered. This is the third classifier over the\nSHOW text to fail the same way (raw regex, literal equality, comment\nstripping) because Drill\u0027s SHOW grammar has no dedicated sqlglot node and\neverything after the keyword collapses into one opaque literal with no\nreliable positive signal for \u0027this is a schema listing\u0027.\n\nReplace classification with guard.is_show_command(sql), which checks only\nthe already-parsed Command.this \u003d\u003d \u0027SHOW\u0027 field and filters every SHOW\ncommand\u0027s first column when hidden_schemas is configured. SHOW TABLES/SHOW\nFILES rows are now incidentally filtered too (a table literally named after\na hidden schema would be dropped), accepted as the fail-closed trade-off in\nplace of a classifier that can leak.\n" }, { "commit": "9e84eebc4cf1f6eb1ffd56bea7aa7169c1b4e233", "tree": "709f2eae2fabb230f89118778ec91e4f5818296e", "parents": [ "0d747a21370a62e59b406b3885c991ee4dc1f83b" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 12:50:04 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 12:50:04 2026 -0400" }, "message": "docs: add Task 11 for dynamic-schema plugin metadata\n\nDESCRIBE cannot answer for plugins whose schema is discovered at read\ntime. Follow sqlalchemy-drill\u0027s get_columns methodology: probe with\nLIMIT 1 for file/mongo/splunk, SELECT `**` for mongo, DESCRIBE only\nfor static-schema plugins. HTTP plugins cannot report columns until a\nquery has been run, so say so explicitly rather than returning empty.\n" }, { "commit": "0d747a21370a62e59b406b3885c991ee4dc1f83b", "tree": "f51d719045234c1aec9c55c7026fe0deb29404d9", "parents": [ "cedd93d7f80e158401de55cc621c82815c6d903e" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 10:04:22 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 10:04:22 2026 -0400" }, "message": "fix: close SHOW SCHEMAS LIKE bypass and harden filter against bad input\n\nis_show_schemas compared the whole Command literal for equality against\n{SCHEMAS, DATABASES}, but sqlglot\u0027s Command fallback swallows the entire\nremainder of a SHOW statement into that literal, so SHOW SCHEMAS LIKE \u0027...\u0027,\na trailing block comment, or SHOW/**/SCHEMAS all defeated it and let hidden\nschemas through unfiltered. Detection now strips block comments and compares\nonly the first token.\n\nAlso hardens _first_value and list_storage_plugins against non-dict rows/\nentries instead of raising, makes _visible fail closed on None, and covers\nall five management tools (not just cluster_status) against a client\nmissing the REST endpoints.\n" }, { "commit": "cedd93d7f80e158401de55cc621c82815c6d903e", "tree": "261334018fe698efb67a2817f284a5f66fe916d6", "parents": [ "8f413e7b861d0e7593819846a2b1886b30bc166f" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 09:48:36 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 09:48:36 2026 -0400" }, "message": "feat: management tools and parser-based SHOW SCHEMAS filtering\n\nSHOW SCHEMAS/DATABASES rows are filtered on return using a parsed-statement\ncheck (guard.is_show_schemas) rather than a regex over the raw SQL, since a\nleading comment defeats a regex anchor but not the tokenizer.\n" }, { "commit": "8f413e7b861d0e7593819846a2b1886b30bc166f", "tree": "7cd13577d16a5e2a7f07ef96b58bda6d308896a9", "parents": [ "9297918d10c348c0fddbd9fad08e73c62ec96d8f" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 01:34:25 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 01:34:25 2026 -0400" }, "message": "fix: never let unexpected exceptions escape run_query as raw tracebacks\n\nBroaden guard._check\u0027s parse-failure handling from SqlglotError to\nException so a RecursionError from adversarial input (deeply nested\nparens) becomes a PolicyError instead of an uncaught crash. Add a\ncatch-all in DrillTools.run_query around the policy check as a\nsecond line of defense, deliberately without embedding the\nexception text (which could carry a path). Validate sql/max_rows\nargument types at the tool boundary. Strengthen list_tables/\ndescribe_table tests to assert forwarded arguments, fix _visible\u0027s\nfail-open truthiness, and split the tools section comment for\nTask 9.\n" }, { "commit": "9297918d10c348c0fddbd9fad08e73c62ec96d8f", "tree": "a19b785f1d4e432e2b5506326ea21cba059088fc", "parents": [ "3249bfce57dd32b4290edda62896b584a7b22fad" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 01:27:47 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 01:27:47 2026 -0400" }, "message": "feat: query and metadata MCP tools\n" }, { "commit": "3249bfce57dd32b4290edda62896b584a7b22fad", "tree": "4e4bb46ee264237087007bd07a5f3d9ae7e90475", "parents": [ "1058cfeb975470ff9ffcc028287ef99bef1ee077" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 01:09:58 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 01:09:58 2026 -0400" }, "message": "fix: scrub credentials from JDBC error paths, close cursors, drop dead code\n\nTwo criticals from review:\n- _jdbc_url() built the connection string from urlparse().netloc, which\n includes userinfo (user:pass@host); a URL with embedded credentials put\n the password directly into the JDBC connection string and from there\n into any driver error that echoed it. Build from hostname/port only.\n- Driver exceptions can also echo the password we hand to\n jaydebeapi.connect() directly (e.g. an auth-failure message). Add\n _scrub() and apply it to every DrillError built from driver-supplied\n text: connect, query, and close. The prior test asserted the\n unscrubbed message verbatim, so it passed only because the leak was\n present; inverted it to assert the password is absent, and added\n matching cases for the query and close paths.\n\nAlso, from the same review:\n- query() now closes its cursor via contextlib.closing instead of\n leaking a statement handle per call\n- close() clears the stale connection handle in a finally and wraps a\n failing connection.close() in DrillError instead of letting it escape\n raw\n- removed the unreachable `if jaydebeapi is None` branch and the dead\n `except DrillError: raise` in query()\n- JDBC\u0027s truncated flag now matches REST\u0027s max_rows\u003e0 guard\n- test_schemas_uses_information_schema now asserts the emitted SQL,\n not just the mapped result\n- noted that Query is invoked positionally\n" }, { "commit": "1058cfeb975470ff9ffcc028287ef99bef1ee077", "tree": "79e045745a90c34c0635f75fd4cac786da92eb87", "parents": [ "6e63ceaa218672f6d34af978cb5a9506026f497b" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 00:54:07 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 00:54:07 2026 -0400" }, "message": "feat: optional JDBC backend\n\nAdd JdbcClient, delegating query building/row-mapping for metadata to\nthe fetch_* functions extracted from RestClient. Management endpoints\n(storage_plugins, cluster_status, profiles, profile, cancel_query) are\nREST-only and deliberately absent. jaydebeapi is imported lazily and\nmocked in tests, so the suite never starts a JVM.\n" }, { "commit": "6e63ceaa218672f6d34af978cb5a9506026f497b", "tree": "883188e6cd0fcca3c70abcc97e232b21a5fbc438", "parents": [ "549397d2845822f200d24166b7cd1c2285f6ff9f" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 00:53:07 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 00:53:07 2026 -0400" }, "message": "fix: address six minor findings from Task 6 review\n\n- reject a bare \"..\" (or empty) dot-segment in quote_identifier, closing\n a directory-traversal gap in fetch_columns\u0027s file-plugin path\n- add direct unit tests for quote_identifier_path/quote_identifier\n rejecting backticks, and rename two tests that claimed coverage they\n did not provide\n- factor response.json() decoding into a _json() helper and guard all\n six sites (query, storage_plugins, cluster_status x2, profiles, profile)\n against a non-JSON body\n- profiles() tolerates a non-dict payload and clamps a negative limit to 0\n- add a direct test for plugin_type\n- drop leftover ^...$ anchors from _IDENTIFIER for consistency with\n _FILE_IDENTIFIER\n" }, { "commit": "549397d2845822f200d24166b7cd1c2285f6ff9f", "tree": "f4d60412457490d383badba47700cfa9a6cedb59", "parents": [ "139df5cf26f1c561ae06217b147c9e59d5b22464" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 00:50:47 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 00:50:47 2026 -0400" }, "message": "refactor: extract metadata fetchers to module-level functions\n\nPull plugin_type/schemas/tables/columns bodies out of RestClient into\nfetch_plugin_type/fetch_schemas/fetch_tables/fetch_columns, which take\na query callable instead of self. RestClient\u0027s methods become one-line\ndelegations. Behavior-preserving: Task 6\u0027s tests pass unchanged.\n" }, { "commit": "139df5cf26f1c561ae06217b147c9e59d5b22464", "tree": "705311b97e166cf04cf979147e1c5badf42e044a", "parents": [ "8f72443e940a348a07341e141dfb4bc7f90f84ce" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 00:48:42 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 00:48:42 2026 -0400" }, "message": "docs: have the JDBC client reuse metadata rather than duplicate it\n\nTask 6\u0027s metadata methods are plain SQL over a query callable, including\nthe file-plugin branching and identifier quoting. Extract them into\nfetch_* functions taking that callable, and have both clients delegate,\nrather than copying ~80 lines of security-relevant logic into a second\nclass. The second consumer is what justifies the extraction.\n" }, { "commit": "8f72443e940a348a07341e141dfb4bc7f90f84ce", "tree": "c504c04818ada7c7185c63607c224c0ac71c795b", "parents": [ "f90261d983c5f07f4351ef33f47d8b4dee7f336a" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 00:26:57 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 00:26:57 2026 -0400" }, "message": "fix: quote file-plugin table names as a single identifier in DESCRIBE\n\nConcatenating schema+table before quote_identifier_path split a dotted\nfilename like sales.csv into separate schema/table segments, producing\nan invalid DESCRIBE. Quote the table with a new quote_identifier() that\nkeeps a dotted filename inside one backtick pair.\n" }, { "commit": "f90261d983c5f07f4351ef33f47d8b4dee7f336a", "tree": "e448c999cbab6a8f51cea28d941f4d828e33815b", "parents": [ "cc158b336765d19bd5716f570afe33c4c01df5ac" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 00:24:34 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Wed Aug 12 00:24:34 2026 -0400" }, "message": "feat: metadata and management endpoints on the REST client\n" }, { "commit": "cc158b336765d19bd5716f570afe33c4c01df5ac", "tree": "e420dcd2dae193d08d47e5d47615fb363d22736b", "parents": [ "a9fe883134ee0d9b83a54caa26933cfb2f3e6387" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 23:38:45 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 23:38:45 2026 -0400" }, "message": "docs: branch metadata lookups on storage plugin type\n\nFile plugins (dfs, s3) do not register contents in INFORMATION_SCHEMA, so\nlist_tables on dfs.tmp would have returned empty and looked like an empty\nworkspace. Use SHOW FILES for file plugins and DESCRIBE for their columns,\nfollowing sqlalchemy-drill. DESCRIBE rather than a SELECT * LIMIT 1 probe,\nso a metadata question never reads user data.\n" }, { "commit": "a9fe883134ee0d9b83a54caa26933cfb2f3e6387", "tree": "725b5d8c6ac45e534359a70ff4474b276c854245", "parents": [ "1b5dabeef025e14a36070722edd77425d247bab7" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 19:07:57 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 19:07:57 2026 -0400" }, "message": "fix: make invalid-credentials detection fail closed, not fail clever\n\nThe previous fix stripped HTML tags with \u003c[^\u003e]+\u003e before matching the\ninvalid-credentials marker, but that regex is greedy across any\n\u0027\u003c...\u003e\u0027 span, not just well-formed tags. A stray unmatched \u0027\u003c\u0027 before\nthe marker (e.g. \u00271 \u003c 2\u0027 in unrelated error-page text) makes the\nsubstitution eat everything up to the next unrelated \u0027\u003e\u0027 in the\ndocument, including the marker itself -- so a genuinely failed login\ncould be silently accepted as successful. Verified this against the\npreviously-committed code before fixing it.\n\nFix: check both the raw body and the tag-stripped body, treating a\nmatch in EITHER as a failure. The union of two detectors can only find\nmore than either alone, so no stripping bug can ever suppress a\ndetection the raw search would have made -- stripping now only adds\nthe tags-inside-the-phrase case it was introduced for, and can no\nlonger cause a false negative on its own. Also tightened _HTML_TAG to\nrequire \u0027/\u0027, a letter, or \u0027!\u0027 after \u0027\u003c\u0027 (\u003c/?[A-Za-z!][^\u003e]*\u003e), so it\napproximates real tag grammar instead of matching any bracketed span;\nkept the union check regardless, since that\u0027s what makes the\nfail-closed property hold independent of how well the tag regex\nbehaves.\n\nAdded five tests: the reviewer\u0027s exact stray-\u0027\u003c\u0027-before-marker case\n(mechanically confirmed to fail against the prior implementation), an\nunclosed \u0027\u003c\u0027 with no matching \u0027\u003e\u0027 anywhere, a plain marker with no\nmarkup, and two fail-closed-direction checks confirming ordinary\nsuccessful logins (plain body, and a body with incidental but\nwell-formed \u0027\u003c\u0027/\u0027\u003e\u0027 characters) still succeed.\n" }, { "commit": "1b5dabeef025e14a36070722edd77425d247bab7", "tree": "23f970486c2c8b37fefe9c4a78261a0c01b3575d", "parents": [ "aa9432ef7f5ca5f14e2936d8f59292b9a8fc9597" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 19:03:15 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 19:03:15 2026 -0400" }, "message": "docs: record the login body check and the reference-implementation decision\n\nDrill\u0027s j_security_check returns 200 with an HTML error page on bad\ncredentials, so a status-only check accepts a wrong password. Records the\ntwo-part check and the evidence from sqlalchemy-drill.\n\nAlso records why PyDrill was evaluated and rejected as a dependency: no\nform login, no Kerberos, last release 2018.\n" }, { "commit": "aa9432ef7f5ca5f14e2936d8f59292b9a8fc9597", "tree": "6fe7b5e9c97ecccf3b70228ab1b2905b397372ac", "parents": [ "fe4ee79d51c1856ce0980affe69a0569ddbf586d" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 19:02:28 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 19:02:28 2026 -0400" }, "message": "fix: harden client_rest per review (fullmatch, non-JSON guard, kerberos/close tests, tag-stripped credential check)\n\n- quote_literal/quote_literal_path use fullmatch instead of match: `$`\n matches before a trailing newline under match(), so \u0027foo\\n\u0027 and\n \u0027dfs.tmp\\n\u0027 silently passed the identifier allowlist despite the\n module\u0027s own comment claiming whitespace is rejected. Not exploitable\n (no quote/backslash/semicolon in the allowed charset), but wrong for\n a trust-boundary validator. Added regression cases to both\n parametrized rejection tests.\n- query() now wraps response.json() in try/except ValueError and\n raises DrillError naming the URL, instead of letting a bare\n JSONDecodeError escape when Drill or a fronting proxy/SSO gateway\n returns a 200 HTML page.\n- truncated is now max_rows \u003e 0 and len(rows) \u003e\u003d max_rows, so\n max_rows\u003d0 no longer reports a false truncation.\n- Dropped dead \u0027if not parts\u0027 check in quote_literal_path (str.split\n never returns an empty list; the any(...) clause already covers it).\n- _INVALID_CREDENTIALS is now matched against the login response body\n with HTML tags stripped first, so a marker split across tags (e.g.\n \u0027Invalid\u003cbr\u003eusername/password credentials\u0027) is still detected\n instead of silently passing as a successful login.\n- Added kerberos auth tests (missing extra -\u003e DrillError naming the\n extra; extra present -\u003e auth object reaches httpx.Client) and a\n close() test, removing the prior pragma: no cover now that both\n branches are exercised.\n- test_gives_up_after_one_retry now asserts login/query call counts,\n so an unbounded retry loop would fail the test instead of hanging.\n- Documented in test_login_rejects_200_with_invalid_credentials_body\n why it deliberately omits a /query.json mock (that omission is what\n makes the test non-vacuous).\n- Moved _error_text above RestClient so Task 6\u0027s metadata/management\n methods can append to the end of the file instead of inserting\n mid-file.\n" }, { "commit": "fe4ee79d51c1856ce0980affe69a0569ddbf586d", "tree": "95e15b9eb60e6c85c3d57b1a389ca3517372b143", "parents": [ "01eac1813273adccf5a1a05364c474c7c6b32c30" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 17:20:19 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 17:20:19 2026 -0400" }, "message": "fix: detect Drill login failure from response body, not status code\n\nDrill\u0027s j_security_check returns HTTP 200 with an HTML error page when\ncredentials are wrong; it does not use a 4xx status for a bad password.\nThe previous status_code\u003e\u003d400 check therefore treated any wrong\npassword as a successful login, leaving the client believing it was\nauthenticated. Now a 200 response is additionally checked for the\n\u0027Invalid username/password credentials\u0027 marker in the body (matched\ncase-insensitively, tolerant of surrounding markup); a non-2xx status\nis reported as an endpoint/connection failure instead.\n\nAdds regression coverage: 200-with-invalid-credentials-body fails\nclosed, 200-with-ordinary-body succeeds, and a 200-with-invalid-\ncredentials response during a 401-triggered re-auth fails closed\nwithout looping or retrying the query. Confirmed the regression test\ngenuinely fails against the prior status-code-only implementation.\n" }, { "commit": "01eac1813273adccf5a1a05364c474c7c6b32c30", "tree": "0f2db3cf483b4e8b04148a05e16db174caa10794", "parents": [ "ee18ced2453e2e6f8b4db2c447d7b756eb9abe6b" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 17:04:09 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 17:04:09 2026 -0400" }, "message": "fix: switch SQL guard from Postgres to native Drill dialect\n\nsqlglot ships a Drill dialect; using Postgres as a stand-in wrongly\nrejected idiomatic Drill queries that sqlglot\u0027s Postgres grammar\ncannot parse, e.g. backtick-quoted file-path queries and\nINFORMATION_SCHEMA lookups.\n" }, { "commit": "ee18ced2453e2e6f8b4db2c447d7b756eb9abe6b", "tree": "ecce208096235793408438cdf12f713802c2e712", "parents": [ "f308c84813e18977e7d0678a3e025a7eeda84284" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 16:43:31 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 16:43:31 2026 -0400" }, "message": "feat: Drill REST client with auth and query execution\n\nFixes two defects found in the task brief\u0027s reference implementation:\n- _login\u0027s URL-based check (\"j_security_check\" in str(response.url))\n false-positives on a successful, unredirected login, since the\n response URL still points at the login endpoint. Status code alone\n is the only reliable signal here.\n- test_connection_failure_message_names_the_url_not_the_password only\n mocked /query.json; since auth\u003dbasic logs in first, the unmocked\n /j_security_check route raised respx\u0027s AllMockedAssertionError\n instead of exercising the intended DrillError path. Now mocks both\n endpoints, matching how a real outage would behave.\n" }, { "commit": "f308c84813e18977e7d0678a3e025a7eeda84284", "tree": "6585632bea3071009760c9e3a086eebcbe4ea2e1", "parents": [ "1ca6f4dd1936be6b48578f045a73a45f38e702ac" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 16:32:30 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 16:32:30 2026 -0400" }, "message": "fix: add session key pattern and tests\n\n- Add session[._-]?key alternative to pattern for AWS STS-style credentials\n- Add tests for sessionKey, session_key, session-key, and X-Session-Key variants\n- All 23 redact tests pass, no regressions in full suite (116/116)\n" }, { "commit": "1ca6f4dd1936be6b48578f045a73a45f38e702ac", "tree": "727a1d992b69daab77c4cda5ff7ba17635ac0f4c", "parents": [ "4d622b946af86c0b61301314df0a67f52c9e1062" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 16:20:43 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 16:20:43 2026 -0400" }, "message": "fix: restore broad credential pattern and fix test contradiction\n\n- Restore broad \u0027credential\u0027 pattern (remove negative lookahead that excluded credentialsJson, credentialsB64, etc)\n- Fix test_recurses_into_nested_dicts to use neutral \u0027aws\u0027 container instead of credentialsProvider\n- Add tests for credentialsProvider, credentialsJson, credentialsB64, credentialsProviderType (GCS key shapes)\n- Expand pattern to include authorization, passphrase, keytab, principal\n- Add tests for new pattern additions\n- Add tuple return type test for complete coverage\n" }, { "commit": "4d622b946af86c0b61301314df0a67f52c9e1062", "tree": "f6685005d397837a959f70ee9eb91c077f562247", "parents": [ "58cf48c7bf321147d25b9115b473cb9de527a12b" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 16:12:32 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 16:12:32 2026 -0400" }, "message": "feat: recursive secret redaction\n" }, { "commit": "58cf48c7bf321147d25b9115b473cb9de527a12b", "tree": "4e202cacb55899c2cb8930d45a65b28906dd3572", "parents": [ "3f5b01b2c734d67dd047c588afb501d502f48583" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 15:58:18 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 15:58:18 2026 -0400" }, "message": "test: hidden schema enforcement in the SQL guard\n" }, { "commit": "3f5b01b2c734d67dd047c588afb501d502f48583", "tree": "7b068a44e847f4d21bce04bd0bf8cb55abd5d850", "parents": [ "0413ffd40a25f6cd74bae14822320efe039cdbf5" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 15:53:52 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 15:53:52 2026 -0400" }, "message": "chore: add ASF license headers and LICENSE file\n" }, { "commit": "0413ffd40a25f6cd74bae14822320efe039cdbf5", "tree": "0c2adc3f02a0d767aa2abaa7aa138f6036c8e488", "parents": [ "47d99a40b5158aea0780be1f4725e108b2db8de8" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 15:47:28 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 15:47:28 2026 -0400" }, "message": "fix: close guard gaps found in review — TokenError, embedded writes, EXPLAIN recursion\n" }, { "commit": "47d99a40b5158aea0780be1f4725e108b2db8de8", "tree": "18e8018f180ad0b2947a219fdbfd70cef3737a13", "parents": [ "2d7855ad600ccadb6d4aa67f6bbd68029bfccfa8" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 15:34:29 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 15:34:29 2026 -0400" }, "message": "feat: SQL guard with deny-by-default write policy\n" }, { "commit": "2d7855ad600ccadb6d4aa67f6bbd68029bfccfa8", "tree": "21a72104e56122912df68ecc72827406ca2df60b", "parents": [ "9ac1e364c677fd94f317786fd416f7f74b286923" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 15:28:24 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 15:28:24 2026 -0400" }, "message": "test: isolate config tests from ambient DRILL_* env vars, add numeric-string coercion test\n" }, { "commit": "9ac1e364c677fd94f317786fd416f7f74b286923", "tree": "8e760eae7dc9953b4484c25583a0be1bb22b1789", "parents": [ "1c59dfd66e4070431e4f03fae4513f67a00b5d35" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 15:02:15 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 15:02:15 2026 -0400" }, "message": "feat: package skeleton and configuration loading\n" }, { "commit": "1c59dfd66e4070431e4f03fae4513f67a00b5d35", "tree": "e046510babab9b09abf5a4ca62ddf0d316be7fb0", "parents": [ "75ee75aa37126c7a564ccbd532ac9cfff0a83bf4" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 14:53:12 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 14:54:15 2026 -0400" }, "message": "Fix stale symbol reference in Task 6\n\n" }, { "commit": "75ee75aa37126c7a564ccbd532ac9cfff0a83bf4", "tree": "98c9963d193c03befec395edf8446034ce8aa56a", "parents": [ "68328ed6f0a447a34cce5ff9cae4d8e92cb2797a" ], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 14:52:44 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 14:54:15 2026 -0400" }, "message": "Add implementation plan for the Drill MCP server\n\nTen TDD tasks: config, SQL guard, redaction, REST client, JDBC client,\ntool layer, and wiring. Each task ends with a runnable test cycle.\n\n" }, { "commit": "68328ed6f0a447a34cce5ff9cae4d8e92cb2797a", "tree": "4b09153370a3b461396c6a82e381f40031a44b80", "parents": [], "author": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 14:36:15 2026 -0400" }, "committer": { "name": "cgivre", "email": "cgivre@gmail.com", "time": "Tue Aug 11 14:54:15 2026 -0400" }, "message": "Add design spec for Apache Drill MCP server\n\nPython MCP server over stdio with a REST backend by default and an\noptional JDBC backend for Kerberos. Deny-by-default SQL policy enforced\nwith sqlglot, per-plugin CTAS allowlist, configurable hidden schemas,\nand secret redaction on storage plugin output.\n\n" } ] }