feat: /discover command — data stack setup with project_scan tool by anandgupta42 · Pull Request #30 · AltimateAI/altimate-code

anandgupta42 · 2026-03-04T01:44:29Z

Summary

Replaces the AGENTS.md-generating /init command with a comprehensive data stack scanner
New project_scan tool detects dbt projects, warehouse connections (from dbt profiles, Docker, env vars), installed tools, and config files
Updated /init template guides the AI agent through a 5-step setup flow: scan → review → add connections → index schemas → show next steps
Documentation updated across getting-started, warehouse tools, commands, and TUI pages

New Files

File	Purpose
`src/tool/project-scan.ts`	`project_scan` tool with 5 exported detection functions + connection deduplication
`test/tool/project-scan.test.ts`	71 bun:test cases (detectGit, detectDbtProject, detectEnvVars, parseToolVersion, detectDataTools, detectConfigFiles)
`altimate-engine/tests/test_env_detect.py`	24 pytest cases for env var → warehouse mapping

Modified Files

File	Change
`src/tool/registry.ts`	Register `ProjectScanTool`
`src/command/template/initialize.txt`	New 5-step data stack setup prompt
`src/command/index.ts`	Description: "scan data stack and set up connections"
`docs/docs/getting-started.md`	`/init` as first-run command
`docs/docs/data-engineering/tools/warehouse-tools.md`	Added project_scan, warehouse_add, warehouse_remove, warehouse_discover docs
`docs/docs/data-engineering/tools/index.md`	Updated warehouse tools count (2 → 6)
`docs/docs/configure/commands.md`	Documented built-in `/init` and `/review` commands
`docs/docs/usage/tui.md`	Added `/init` to slash command examples

What `project_scan` detects

Category	Method
Git	branch, remote URL via `git` commands
dbt project	Walks up for `dbt_project.yml`, parses manifest
dbt profiles	Bridge call to `~/.dbt/profiles.yml`
Docker DBs	Bridge call for running Postgres/MySQL/MSSQL containers
Existing connections	Bridge call to list configured warehouses
Env vars	Snowflake, BigQuery, Databricks, Postgres, MySQL, Redshift
Schema cache	Bridge call for indexed warehouse status
Data tools	dbt, sqlfluff, airflow, dagster, prefect, soda, sqlmesh, great_expectations, sqlfmt
Config files	.altimate-code/, .sqlfluff, .pre-commit-config.yaml

Test plan

bun test — 1220 pass, 0 fail (full suite including 71 new tests)
pytest tests/test_env_detect.py — 24 pass
mkdocs build --strict — docs build cleanly
Manual: run /init in TUI from a dbt project directory
Manual: run /init in TUI from a bare directory (no dbt)
Manual: verify connections discovered from env vars get offered for setup

🤖 Generated with Claude Code

Replace the AGENTS.md-generating /init command with a comprehensive data stack scanner that detects dbt projects, warehouse connections, Docker databases, installed tools, and config files. The AI agent then walks the user through adding connections, testing them, and indexing schemas. New project_scan tool with 5 exported detection functions: - detectGit: branch, remote URL - detectDbtProject: dbt_project.yml, manifest, packages - detectEnvVars: Snowflake, BigQuery, Databricks, Postgres, MySQL, Redshift - detectDataTools: dbt, sqlfluff, airflow, dagster, prefect, soda, sqlmesh, great_expectations, sqlfmt - detectConfigFiles: .altimate-code/, .sqlfluff, .pre-commit-config.yaml Tests: 71 TypeScript (bun:test) + 24 Python (pytest) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Restore the original /init command (creates AGENTS.md) and move the data stack setup functionality to /discover instead. Updates all docs to reference /discover as the recommended first-run command. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

GitHub Actions checks out in detached HEAD state, so git branch --show-current returns empty. The test now accepts undefined branch in that case. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

In CI detached HEAD, git branch --show-current returns an empty string. Convert empty string to undefined so callers get a clean undefined instead of "". Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

The "detects a git repository" test should only assert isRepo. Branch validation is handled by the dedicated branch test which accounts for CI detached HEAD returning undefined. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

jontsai · 2026-03-04T03:11:04Z

+    stderr: "pipe",
+  })
+  if (isRepoResult.exitCode !== 0) {
+    return { isRepo: false }


thought: detectGit (and detectDataTools) use Bun.spawnSync directly — this couples the tool to the Bun runtime. If there's ever a need to run under Node.js (or test with a different runner), these would need to be swapped for child_process.spawnSync. Fine for now since the project is Bun-only, but worth noting.

jontsai · 2026-03-04T03:11:23Z

+      type: wh.type,
+      source: "env-var",
+      signal: matchedSignal,
+      config,


suggestion: The config object captures raw env var values including password, access_token, etc. These get returned to the LLM in the tool output. Consider either redacting sensitive fields (replace with "***") or only including non-secret fields in the scan output. The connection setup step (warehouse_add) can read the env vars directly when actually configuring.

jontsai · 2026-03-04T03:11:24Z

+        database: "REDSHIFT_DATABASE",
+        user: "REDSHIFT_USER",
+        password: "REDSHIFT_PASSWORD",
+      },


thought: DATABASE_URL is used as a Postgres signal, but many frameworks (Rails, Django, etc.) set DATABASE_URL for any database type including MySQL, SQLite, etc. This could produce false positives. Consider parsing the URL scheme (postgresql:// vs mysql://) before categorizing, or at least noting the assumption in the output.

jontsai · 2026-03-04T03:11:25Z

+  for (const tool of DATA_TOOL_NAMES) {
+    try {
+      const result = Bun.spawnSync([tool, "--version"], {
+        stdout: "pipe",


nit: The 9 tool version checks run sequentially. Since each has a 5s timeout, worst case is 45s. Could use Promise.all to run them in parallel and cut the scan time significantly.

jontsai

LGTM — solid feature addition with excellent test coverage (71 TS + 24 Python tests).

What's good:

Clean separation of detection functions, each independently testable
Thoughtful connection deduplication logic across sources
Resilient CI handling (detached HEAD edge case)
Well-structured /discover flow that guides users step by step
Docs are comprehensive and match the implementation

Minor items (posted inline):

Env var detection captures secrets (passwords, tokens) in scan output — consider redacting
DATABASE_URL assumed to be Postgres but could be any DB type
Tool version checks are sequential; parallelizing would speed up scans
Bun.spawnSync coupling is fine for now but noted for awareness

None are blockers. Ship it! 🚀

…arallelize tool checks - Redact sensitive env var values (password, access_token, connection_string) in scan output with "***" so secrets are never sent to the LLM - Parse DATABASE_URL scheme (postgresql://, mysql://, etc.) to detect the correct database type instead of assuming Postgres - Parallelize tool version checks with Promise.all instead of sequential loop Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

setu-altimateai

Nice work addressing the previous review comments! The secret redaction and DATABASE_URL scheme parsing in the TypeScript implementation look solid. One issue with the Python test file — see inline comment. Overall LGTM pending that fix.

setu-altimateai · 2026-03-04T03:23:53Z

+            "schema": "SNOWFLAKE_SCHEMA",
+            "role": "SNOWFLAKE_ROLE",
+        },
+    },


issue: This Python reference implementation is now out of sync with the TypeScript version in two ways:

DATABASE_URL is still listed as a postgres signal (line 28), but the TS version removed it from postgres signals and handles it separately with scheme-based type detection (postgresql:// → postgres, mysql:// → mysql, etc.). This means the Python version will always classify DATABASE_URL as postgres regardless of the actual scheme.

No secret redaction — the TS version now masks sensitive keys (password, access_token, connection_string) with "***", but the Python detect_env_connections() returns raw values. If this Python implementation is used later, it would leak secrets.

Since the docstring says "mirrors TypeScript detectEnvVars", these should be kept in sync.

- Remove DATABASE_URL from postgres signals, add scheme-based detection - Add secret redaction (password, access_token, connection_string) - Add tests for DATABASE_URL scheme parsing and deduplication Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

jontsai

All previous comments addressed ✅

Secret redaction with SENSITIVE_KEYS set — passwords, tokens, connection strings all masked as ***
DATABASE_URL scheme parsing maps to correct DB types (postgresql, mysql, redshift, sqlite)
Python test file now in sync with TS: includes SENSITIVE_KEYS, DATABASE_URL_SCHEME_MAP, and redaction tests
Tool version checks parallelized with Promise.all

Clean PR, great test coverage (71 TS + 24 Python). LGTM 🚀

setu-altimateai

LGTM ✅

Final review complete. The previously flagged issue (Python test_env_detect.py out of sync with TS) is resolved:

✅ DATABASE_URL now uses scheme-based detection matching TS (postgresql, postgres, mysql, mysql2, redshift, sqlite)
✅ Sensitive values (password, access_token, connection_string, private_key_path) properly redacted with "***"
✅ Python reference implementation mirrors TS detectEnvVars faithfully
✅ Comprehensive test coverage across all warehouse types
✅ project-scan.ts is well-structured with proper deduplication logic
✅ Docs updated consistently

Ship it! 🚀

* feat: replace /init with data stack setup command Replace the AGENTS.md-generating /init command with a comprehensive data stack scanner that detects dbt projects, warehouse connections, Docker databases, installed tools, and config files. The AI agent then walks the user through adding connections, testing them, and indexing schemas. New project_scan tool with 5 exported detection functions: - detectGit: branch, remote URL - detectDbtProject: dbt_project.yml, manifest, packages - detectEnvVars: Snowflake, BigQuery, Databricks, Postgres, MySQL, Redshift - detectDataTools: dbt, sqlfluff, airflow, dagster, prefect, soda, sqlmesh, great_expectations, sqlfmt - detectConfigFiles: .altimate-code/, .sqlfluff, .pre-commit-config.yaml Tests: 71 TypeScript (bun:test) + 24 Python (pytest) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * refactor: restore /init, rename data stack setup to /discover Restore the original /init command (creates AGENTS.md) and move the data stack setup functionality to /discover instead. Updates all docs to reference /discover as the recommended first-run command. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: make detectGit test resilient to CI detached HEAD GitHub Actions checks out in detached HEAD state, so git branch --show-current returns empty. The test now accepts undefined branch in that case. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: treat empty git branch as undefined in detectGit In CI detached HEAD, git branch --show-current returns an empty string. Convert empty string to undefined so callers get a clean undefined instead of "". Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: remove branch type assertion from detectGit repo test The "detects a git repository" test should only assert isRepo. Branch validation is handled by the dedicated branch test which accounts for CI detached HEAD returning undefined. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: address PR review — redact secrets, parse DATABASE_URL scheme, parallelize tool checks - Redact sensitive env var values (password, access_token, connection_string) in scan output with "***" so secrets are never sent to the LLM - Parse DATABASE_URL scheme (postgresql://, mysql://, etc.) to detect the correct database type instead of assuming Postgres - Parallelize tool version checks with Promise.all instead of sequential loop Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: sync Python reference implementation with TypeScript changes - Remove DATABASE_URL from postgres signals, add scheme-based detection - Add secret redaction (password, access_token, connection_string) - Add tests for DATABASE_URL scheme parsing and deduplication Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

anandgupta42 changed the title ~~feat: /init command — data stack setup with project_scan tool~~ feat: /discover command — data stack setup with project_scan tool Mar 4, 2026

anandgupta42 and others added 5 commits March 3, 2026 19:11

fix: make detectGit test resilient to CI detached HEAD

c85804d

GitHub Actions checks out in detached HEAD state, so git branch --show-current returns empty. The test now accepts undefined branch in that case. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

fix: treat empty git branch as undefined in detectGit

2012225

In CI detached HEAD, git branch --show-current returns an empty string. Convert empty string to undefined so callers get a clean undefined instead of "". Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

jontsai reviewed Mar 4, 2026

View reviewed changes

jontsai approved these changes Mar 4, 2026

View reviewed changes

anandgupta42 force-pushed the feat/init-command branch from 8c8d769 to 64d311a Compare March 4, 2026 03:12

setu-altimateai Bot reviewed Mar 4, 2026

View reviewed changes

jontsai approved these changes Mar 4, 2026

View reviewed changes

setu-altimateai Bot approved these changes Mar 4, 2026

View reviewed changes

anandgupta42 merged commit 05b3d42 into main Mar 4, 2026
4 checks passed

kulvirgit deleted the feat/init-command branch March 10, 2026 21:06

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: /discover command — data stack setup with project_scan tool#30

feat: /discover command — data stack setup with project_scan tool#30
anandgupta42 merged 7 commits intomainfrom
feat/init-command

anandgupta42 commented Mar 4, 2026

Uh oh!

jontsai Mar 4, 2026

Uh oh!

jontsai Mar 4, 2026

Uh oh!

jontsai Mar 4, 2026

Uh oh!

jontsai Mar 4, 2026

Uh oh!

jontsai left a comment

Uh oh!

setu-altimateai Bot left a comment

Uh oh!

setu-altimateai Bot Mar 4, 2026

Uh oh!

jontsai left a comment

Uh oh!

setu-altimateai Bot left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

anandgupta42 commented Mar 4, 2026

Summary

New Files

Modified Files

What project_scan detects

Test plan

Uh oh!

jontsai Mar 4, 2026

Choose a reason for hiding this comment

Uh oh!

jontsai Mar 4, 2026

Choose a reason for hiding this comment

Uh oh!

jontsai Mar 4, 2026

Choose a reason for hiding this comment

Uh oh!

jontsai Mar 4, 2026

Choose a reason for hiding this comment

Uh oh!

jontsai left a comment

Choose a reason for hiding this comment

Uh oh!

setu-altimateai Bot left a comment

Choose a reason for hiding this comment

Uh oh!

setu-altimateai Bot Mar 4, 2026

Choose a reason for hiding this comment

Uh oh!

jontsai left a comment

Choose a reason for hiding this comment

Uh oh!

setu-altimateai Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

What `project_scan` detects