AltimateAI · Sourabhchrs93 · May 14, 2026 · May 14, 2026
@@ -3,3 +3,15 @@ __pycache__/
 *.pyc
 .venv/
 .lycheecache
+
+# Local commit-message scratch (referenced by general.md commit workflow)
+.github/meta/
+
+# Pipeline artifacts (raw release/PR data fetched from private repos)
+data/
+
+# Auto-generated polish drafts that include private PR refs in HTML comments.
+# Curated monthly drafts under generated/<month>/ are tracked explicitly.
+generated/*
+!generated/april-2026/
+!generated/*/
@@ -0,0 +1,11 @@
+---
+title: dbt 1.11 UDF lineage in Datamates MCP
+date: 2026-04-01
+products: [datamates]
+tag: new
+emoji: 🔗
+draft: true
+description: User-defined functions from dbt 1.11 now appear as first-class nodes in lineage graphs served through MCP.
+---
+
+Datamates MCP now parses dbt 1.11 function nodes and serves them as first-class lineage nodes alongside models, sources, and exposures. UDFs registered in a dbt project show up in lineage queries automatically — no new MCP tools, no configuration. Projects on older dbt versions see no change; the parser produces an empty map and falls through.
@@ -0,0 +1,13 @@
+---
+title: dbt Cloud sync, 5x faster
+date: 2026-04-02
+products: [dbt-power-user, datamates]
+tag: improved
+emoji: 🧰
+draft: true
+description: dbt Cloud sync consolidates ~1000 daily tasks per project/environment into a single ingestion run, cutting sync time from ~250 min to under 50.
+---
+
+dbt Cloud sync no longer creates one ingestion task per dbt Cloud run. It now consolidates to a single ingestion per `(project, environment)` per cycle — so a project firing 1000 runs a day produces one ingestion instead of 1000 redundant ones.
+
+A typical 4-worker sync drops from ~250 minutes to under 50.
@@ -0,0 +1,13 @@
+---
+title: LLM Guard blocks prompt injection and poisoning
+date: 2026-04-02
+products: [snowflake-app, databricks-app, datamates]
+tag: improved
+emoji: 🛡️
+draft: true
+description: Studio's agent gateway now blocks nine new injection categories — instruction override, role hijacking, system-prompt extraction, payload obfuscation, multi-turn poisoning, and more.
+---
+
+The Studio agent gateway now blocks nine new categories of prompt-injection and prompt-poisoning attempts before they reach the model: instruction override, role hijacking, system-prompt extraction, security bypass, fake system tokens, payload obfuscation, indirect injection (via tool output or documents), social engineering, and multi-turn poisoning.
+
+Existing AI-model-identification detection continues to work, and the new checks apply to every Studio conversation automatically.
@@ -0,0 +1,11 @@
+---
+title: Remote MCP setup with HTTP and SSE transport options
+date: 2026-04-02
+products: [datamates, altimate-code]
+tag: improved
+emoji: 🔌
+draft: true
+description: The remote MCP setup screen now offers Streamable HTTP and SSE transport tabs, each pre-filled with the right config and auth headers.
+---
+
+The remote MCP setup screen now offers two transports: Streamable HTTP (the default for Claude and most clients) and SSE (for clients that don't speak streamable HTTP — Cursor in particular). Each tab pre-fills the matching config snippet and auth headers, so connecting is a single copy-paste.
@@ -0,0 +1,13 @@
+---
+title: AI code review for GitLab merge requests
+date: 2026-04-04
+products: [altimate-code]
+tag: new
+emoji: 🔎
+draft: true
+description: The Altimate Code CLI fetches a GitLab MR diff, runs the AI review, and posts the result back as MR notes.
+---
+
+Altimate Code's CLI now reviews GitLab merge requests end-to-end. Run `altimate-code gitlab review <mr-url>` and the CLI fetches the diff through GitLab REST API v4, runs the AI review, and posts the results back as MR notes — the same flow that has been available for GitHub since launch.
+
+URL parsing handles gitlab.com, self-hosted instances on custom ports, and nested group project paths. A `--no-post-comment` flag dry-runs the review locally, and `--model` picks which model handles the analysis.
@@ -0,0 +1,13 @@
+---
+title: Sign in to Altimate and find project-local dbt profiles
+date: 2026-04-04
+products: [altimate-code]
+tag: new
+emoji: 🔐
+draft: true
+description: A `/login` command for the Altimate provider, plus dbt-standard profile discovery that picks up project-local and `DBT_PROFILES_DIR` paths.
+---
+
+Altimate Code's TUI now includes the Altimate platform as a first-class LLM provider. A new `/login` dialog asks for instance name, API key, and URL, validates the credentials before saving, and re-bootstraps the session immediately — no restart, no manual config edits.
+
+dbt profile discovery also catches up to dbt's standard resolution order. `/discover` now finds profiles in priority order: an explicit path argument, `DBT_PROFILES_DIR`, a project-local `profiles.yml` sitting next to `dbt_project.yml`, and finally `~/.dbt/profiles.yml`. The common CI/CD pattern of committing a project-local profile no longer falls back silently to the global one.
@@ -0,0 +1,13 @@
+---
+title: dbt Cloud environment aliases
+date: 2026-04-06
+products: [dbt-power-user]
+tag: improved
+emoji: 🧰
+draft: true
+description: Map differently-named dbt environments to the same logical environment so warehouse data lines up with model data.
+---
+
+dbt Cloud environments can now carry aliases. If your warehouse refers to "Prod" but your dbt Cloud environment is "Production_5505", set an alias and both names resolve to the same logical environment — no more disjoint joins or missing model rows on the warehouse side.
+
+Multiple aliases per environment are supported; setting none leaves behavior unchanged.
@@ -0,0 +1,13 @@
+---
+title: Studio gets citations, shareable links, and scheduled reports
+date: 2026-04-06
+products: [snowflake-app, databricks-app]
+tag: new
+emoji: ✨
+draft: true
+description: Inline source citations on every Studio answer, shareable session links, and recurring AI reports delivered to email.
+---
+
+Studio answers now cite their sources inline. Click any reference to jump to the underlying table, query, or dashboard the answer was built from — no more taking the AI's word for it.
+
+Sharing a session is now a single dialog: generate a time-limited read-only link and send it to anyone who has access, no copy-pasting screenshots. The new schedule action turns any Studio conversation into a recurring email report — pick a cadence, pick recipients, and the same analysis lands in their inbox each week with fresh data and inline charts.
@@ -0,0 +1,11 @@
+---
+title: Table-level lineage now exports to CSV
+date: 2026-04-06
+products: [datamates]
+tag: new
+emoji: 🔗
+draft: true
+description: Download upstream or downstream table lineage as CSV — same export contract as column lineage, one row per edge.
+---
+
+Table-level lineage now exports to CSV alongside the existing column-level export. Pick any table, choose upstream or downstream, and download the full traversal as a CSV — one row per edge with Source DB / Schema / Table → Target DB / Schema / Table.
@@ -0,0 +1,11 @@
+---
+title: dbt-aware syntax highlighting for SQL+Jinja and YAML+Jinja
+date: 2026-04-07
+products: [dbt-power-user]
+tag: improved
+emoji: 🎨
+draft: true
+description: Dedicated syntax highlighting for dbt SQL+Jinja and YAML+Jinja, so `ref`, `source`, `config`, and Jinja blocks each get their own colors.
+---
+
+dbt Power User now ships dedicated syntax highlighting for dbt SQL+Jinja and YAML+Jinja. `ref`, `source`, `config`, SQL aggregates, window functions, and Jinja `{{ }}` / `{% %}` / `{# #}` blocks each get distinct colors. Schema YAML files highlight Jinja inline alongside YAML, and SQL files no longer fall back to a generic grammar that mis-colored half the keywords.
@@ -0,0 +1,13 @@
+---
+title: One-click fix for "dbt works in terminal but not in the extension"
+date: 2026-04-07
+products: [dbt-power-user]
+tag: new
+emoji: 🐍
+draft: true
+description: A "Detect Python from terminal" action finds the Python interpreter where dbt actually lives and points the extension at it.
+---
+
+The most-reported onboarding bug — dbt works in your terminal but not in the extension because VS Code's Python extension picked a different interpreter — now has a one-click fix. The new **Detect Python from terminal** action runs through your login shell to find where `dbt` actually lives and writes that path into the extension's Python override.
+
+The button appears on every Python or dbt error dialog and in the onboarding prerequisites step, so the failure mode that used to require Stack Overflow now resolves in seconds.
@@ -0,0 +1,13 @@
+---
+title: Tokens tab with live conversion and deposit flow
+date: 2026-04-09
+products: [snowflake-app, databricks-app, datamates]
+tag: new
+emoji: 💰
+draft: true
+description: A single Credits page tab that shows monthly allowance, grant balance, and Token Wallet — plus an Add Tokens form with live dollar-to-token conversion.
+---
+
+The Credits page now has a Tokens tab that consolidates monthly allowance, grant balance, and Token Wallet into one view. A progress bar shows what's left of the monthly allowance, and the Token Wallet card shows both dollar and token equivalents so finance and engineering see the same number.
+
+Adding tokens used to mean leaving the page. The new inline form previews the conversion as you type ($10 = 1.2M tokens) and the transaction history below shows every grant, top-up, and consumption event with timestamps and token amounts.
@@ -0,0 +1,13 @@
+---
+title: Chat panel polish — instant open, title bar icon, token usage indicator
+date: 2026-04-11
+products: [datamates]
+tag: improved
+emoji: ⚡
+draft: true
+description: The Altimate MCP chat panel opens with no perceptible delay, lives on the editor title bar, and shows live token usage in the header.
+---
+
+The Altimate MCP chat panel now opens instantly — no waiting on a startup check before the panel renders.
+
+A new Altimate icon lands on the editor title bar for one-click access without the command palette. The header also picks up a compact token usage indicator — usage percentage color-coded against your monthly threshold, or "Unlimited" on unlimited plans. Click it for a detailed popover with allowance, grants, overage, billing period, and wallet balance.
@@ -0,0 +1,13 @@
+---
+title: Generate dbt unit tests from your manifest
+date: 2026-04-13
+products: [altimate-code, dbt-power-user]
+tag: new
+emoji: 🧪
+draft: true
+description: A new `dbt_unit_test_gen` tool inspects compiled SQL and writes dbt unit tests with type-correct mock data, including incremental and ephemeral cases.
+---
+
+Altimate Code now generates dbt unit tests for you. Point it at a model and it reads the manifest and compiled SQL, detects the scenarios that matter (CASE branches, JOINs, GROUP BY, division-by-zero, incremental loads), and writes a `unit_tests:` block in the model's schema YAML with type-correct mock data — happy path, null variants, and boundary cases.
+
+Incremental models also get a prior-state mock so the generated test exercises the merge logic, not just the initial load. Snowflake, Databricks, BigQuery, Redshift, and Postgres are all supported.
@@ -0,0 +1,11 @@
+---
+title: Auto-suspend savings shown as a range
+date: 2026-04-14
+products: [snowflake-app]
+tag: improved
+emoji: 💰
+draft: true
+description: Warehouse auto-suspend savings now report a min/max range that accounts for Snowflake's 30-second polling delay.
+---
+
+Warehouse auto-suspend savings now report a range instead of a single point estimate. Snowflake polls for suspendable warehouses roughly every 30 seconds, so the actual suspend happens between 0 and 30 seconds after the `auto_suspend` timer fires. The recommendation card shows both the best case (immediate suspend) and the expected case (mid-poll, +15s) so the savings number isn't optimistic by 15 seconds of runtime per cycle.
@@ -0,0 +1,13 @@
+---
+title: Subscriptions — webhooks, cloning, team alerts, same-weekday comparisons
+date: 2026-04-14
+products: [snowflake-app, databricks-app, datamates]
+tag: improved
+emoji: 🔔
+draft: true
+description: Webhook delivery, one-click clone for any alert, team-level cost alerts, and same-weekday comparison for warehouse metrics.
+---
+
+Subscriptions picks up four shippable improvements this month. Webhooks join Slack and email as a first-class delivery channel — paste a URL into any rule and get a typed JSON payload with tenant, timestamp, and a deep-link back to the dashboard. Clone duplicates any alert or report you own in one click; the copy lands in the wizard ready to edit.
+
+Team is now a first-class entity in alerting, so a rule like _"total daily cost for team data-products > $100"_ works without per-warehouse plumbing. The new same-weekday comparison mode compares today against the prior week's same day — much better for warehouses with weekly seasonality than the old day-over-day baseline. Tag filters in the wizard switched to server-side search, so high-cardinality tag spaces no longer freeze the picker.
@@ -0,0 +1,13 @@
+---
+title: Team attribution in QTP Slack threads
+date: 2026-04-14
+products: [snowflake-app]
+tag: improved
+emoji: 👥
+draft: true
+description: Query Timeout Prediction Slack threads now show the owning team even when the query tag is missing — using your tenant's ownership rules as a fallback.
+---
+
+When QTP posts a long-running query to Slack, the alert now identifies the owning team even on queries that don't carry a `team` tag. Your tenant's existing ownership rules apply as a fallback whenever the query tag itself is missing.
+
+Tag-derived team always wins when present; the rules engine only fills the gap, so existing tagging conventions keep working unchanged.
@@ -0,0 +1,13 @@
+---
+title: New Cortex AI services and readable warehouse names
+date: 2026-04-15
+products: [snowflake-app]
+tag: new
+emoji: ❄️
+draft: true
+description: Cortex Code CLI, Cortex Agent, and Snowflake Intelligence now appear in AI Services. Cortex Functions usage shows warehouse names instead of numeric IDs.
+---
+
+The AI Services page now tracks three new Snowflake Cortex service types: **Cortex Code CLI**, **Cortex Agent**, and **Snowflake Intelligence**. Each has its own cost summary, usage table, filter set, and place on the summary-page graph, so spend on the newer Cortex surfaces is no longer lumped under generic "Cortex".
+
+Cortex Functions usage now displays warehouse names (`EDS_HUMAN_WH_LARGE`) instead of opaque numeric IDs (`223`), and warehouse names work as a filter dimension. Same data, finally readable.
@@ -0,0 +1,13 @@
+---
+title: CTE Profiler — per-CTE timing and row counts in the editor
+date: 2026-04-15
+products: [dbt-power-user]
+tag: new
+emoji: ⏱️
+draft: true
+description: Run a dbt model and the editor decorates each CTE with its wall-clock time, row count, and a hot/warm/cool heat tier so you can find the slow one without leaving the file.
+---
+
+The dbt Power User extension now profiles every CTE in a model in one click. The CTE Profiler measures per-CTE wall-clock time and row count against your warehouse and decorates the editor inline (`⏱ 1.7s · 100 rows`). Hot CTEs go red, warm yellow, cool grey; a total time and row count lands at the bottom of the file.
+
+No Altimate API key required — the profiler runs through your existing dbt connection.
@@ -0,0 +1,13 @@
+---
+title: Schedule name validation and "Send Now"
+date: 2026-04-15
+products: [snowflake-app, databricks-app]
+tag: improved
+emoji: ⏱️
+draft: true
+description: Auto-generated schedule names now stay under the limit, and a Send Now option fires a scheduled report on demand.
+---
+
+Auto-generated schedule names that ran past the 255-character limit no longer fail silently — names are truncated to fit, and the schedule form blocks names over the limit with a clear error instead of a backend rejection.
+
+The schedule "More options" menu also picks up a **Send Now** action, so a report can be fired on demand without waiting for the next scheduled run.
@@ -0,0 +1,13 @@
+---
+title: Model hovers and Run/Test codelenses inside YAML schema files
+date: 2026-04-15
+products: [dbt-power-user]
+tag: improved
+emoji: 🧰
+draft: true
+description: Hover any model in `schema.yml` to see its description, columns, and types — and run or test it without switching to the SQL file.
+---
+
+YAML schema files in dbt Power User now behave like first-class model surfaces. Hover any model name in a `models:` section or source table in a `sources:` block to see the model's description, columns, and data types as a popup — no need to jump to the `.sql` file to remember what's in it.
+
+Each model entry also gets Run, Test, and Document codelenses inline, so the most common actions are one click away from the schema file you're already editing.
@@ -0,0 +1,13 @@
+---
+title: Data-parity diffs across SQL Server, Fabric, and ClickHouse
+date: 2026-04-21
+products: [altimate-code]
+tag: new
+emoji: 🔁
+draft: true
+description: Altimate Code can now diff data across SQL Server / Azure Fabric and ClickHouse with partition-aware execution and seven Azure AD auth flows.
+---
+
+Altimate Code's `data_diff` tool now handles three more warehouses end-to-end: SQL Server, Azure Fabric, and ClickHouse — all with partition-aware execution so large tables diff in independent chunks instead of one monolithic scan.
+
+Azure AD authentication is supported for SQL Server and Fabric, covering the common service-principal, MSI, and CLI flows.
@@ -0,0 +1,13 @@
+---
+title: Databricks AI Gateway as an LLM provider
+date: 2026-04-21
+products: [altimate-code, databricks-app]
+tag: new
+emoji: 🧱
+draft: true
+description: Use any of 11 Databricks-hosted foundation models — Llama 3.1, Claude, GPT-5, Gemini, DBRX, Mixtral — as the backing model for Altimate Code.
+---
+
+Altimate Code now treats Databricks serving endpoints as a first-class LLM provider. Authenticate with a Databricks personal access token and the provider works against any AWS, Azure, or GCP Databricks workspace.
+
+Eleven foundation models register out of the box: Meta Llama 3.1 (405B / 70B / 8B), Claude Sonnet and Opus, GPT-5 variants, Gemini 3.1 Pro, DBRX Instruct, and Mixtral 8x7B.
@@ -0,0 +1,13 @@
+---
+title: Referral signups with bundled token credit
+date: 2026-04-21
+products: [snowflake-app, databricks-app, datamates]
+tag: new
+emoji: 🎁
+draft: true
+description: Sign up with a referral code to get a free token grant, and create codes from a new admin page with optional expiry and usage caps.
+---
+
+New signups can now redeem a referral code at registration. Valid codes unlock the sign-up form for personal email providers that the company-email gate normally blocks, while disposable and temp-mail domains stay rejected. Successful signups receive a community token grant credited to their wallet on email verification.
+
+Admins manage codes from a new `/referrals` page: generate a code with optional max-uses and expiry, see active/inactive/expired/exhausted status badges at a glance, and click any code to copy it. Codes can be auto-generated with high-entropy IDs or supplied manually for hand-tracked campaigns.
@@ -0,0 +1,13 @@
+---
+title: Databricks Jobs, Spark analysis, and Users deep-dive
+date: 2026-04-22
+products: [databricks-app]
+tag: new
+emoji: 🧱
+draft: true
+description: Run trends, Spark stage-level analysis, user attribution, and a dedicated Databricks summary.
+---
+
+The Databricks App now drills down to the rows that explain a job's cost: a Jobs page with success/failure run trends and per-job duration and cost; a Spark analysis surface with stage-level gantt charts, task drill-downs, and config tabs; and a Users page with per-user cost attribution and a summary chart.
+
+A Databricks-specific Summary page replaces the Snowflake layout for Databricks tenants — workspace cost moves to the top of the sidebar, and the navigation hides Snowflake-only sections that didn't apply. Open a job to follow it from run trends → stages → tasks → individual run, without losing the cost context on the way down.