dbt-cortex-agent 0.0.9

Agent identity binding, fail-closed evaluation gates, durable failure evidence, skill role consistency, and bounded runtime handling. Qualified through package CI, protected exact-wheel lifecycle proof, and the agentmanagementdbt adopter with both native evaluation suites passing unchanged baselines. PyPI publication is gated by a fresh protected release-wheel proof.

dbt-cortex-agent 0.0.8

Centralizes controlled operational failure handling across deployment, routing, and evaluation without changing Agent DDL or lifecycle behavior. The protected release workflow qualifies the exact wheel before PyPI publication.

dbt-cortex-agent 0.0.7

Corrects the public dbt evaluation macro path, validates evaluation run names before SQL interpolation, preserves failed expected-tool evidence, preflights raw-event collisions, and keeps programming defects distinct from controlled operational failures.~

v0.0.6

Immutable Cortex Agent lifecycle, package-native deployment and evaluation safety, compact runtime protocol handling, and reproducible release qualification.

v0.0.5

Fix: preview commands resolve the supplied connection

Governed dbt projects build profiles.yml from environment variables, and every manifest-dependent command runs a fresh dbt parse. In 0.0.4, execution-context resolution was gated behind --apply, so agent deploy and eval verify previews failed with a missing SNOWFLAKE_ACCOUNT parse error even when --connection was supplied.

An explicitly supplied connection is now resolved into dbt's environment for any command that parses. Resolution reads local Snow CLI configuration and does not contact Snowflake, so previews remain non-mutating; mutation, runtime, and spend still require explicit --apply with target and database allowlists.

Upgrade from 0.0.4 if you use agent deploy or eval verify.

v0.0.4

Package-native Agent workflows

• dbt-cortex-agent agent deploy previews or applies selected Agent deployment: it resolves physical identities, validates every Agent and skill-stage database against explicit allowlists, uploads declared skills, then invokes the dbt dependency closure. The cortex_agent materialization remains the only Agent DDL authority.
• dbt-cortex-agent eval verify materializes the eval model, revalidates the signed plan, runs native evaluation, consumes the exact candidate, and applies intrinsic thresholds or an accepted baseline. Quality failure exits 1; controlled configuration or infrastructure failure exits 2.
• Execution context accepts an explicit --role and resolves Snow CLI credentials only for operations that connect, so previews remain offline.
• WORKLOAD_IDENTITY is recognized alongside SNOWFLAKE_JWT when resolving a connection into the dbt child environment.

Consumers no longer need copied Makefiles or Python sequencers for deployment or evaluation. PyPI 0.0.4 and Git tag v0.0.4 identify one immutable release.

dbt-cortex-agent v0.0.3

Highlights

• Support target-resolved Cortex Agents and related resources across multiple approved databases.
• Resolve Agent smoke, skill operations, and mutation guards from physical resource identity.
• Introduce evaluation-plan schema v2 with explicit Agent, table, stage, result, target, role, and warehouse identity.
• Namespace evaluation candidates, diagnostics, and baselines by target and physical Agent FQN.
• Add protected exact-wheel multi-database live qualification before PyPI publication.

dbt-cortex-agent v0.0.2

Changes

• retry partial native Cortex Agent evaluations under a new immutable run identity
• require complete input and metric coverage before accepting evaluation success
• preserve sanitized expected/actual cardinality diagnostics

Verification

• 192 package tests passed
• Python 3.10-3.13 and dbt 1.10/1.11 compatibility checks passed
• source and wheel distributions built successfully

dbt-cortex-agent v0.0.1

First supported release of dbt-cortex-agent. Agent lifecycle is owned by the dbt cortex_agent materialization; Python tooling is limited to skills, runtime smoke, and evaluation orchestration.

dbt-cortex-agent v0.3.0

Summary

• Ships the dbt package and Python operational CLI as one immutable v0.3.0 product.
• Keeps Cortex Agents as dbt exposures while dbt owns graph contracts, rendering, lifecycle DDL, versions, aliases, grants, and evaluation plans.
• Adds canonical/native-eval projections, staged skill deployment and smoke verification, schema-v2 evaluation artifacts, accepted-baseline gates, explicit paid-evaluation allowlists, and dbt-owned role context.
• Includes public installation, architecture, lifecycle, CI, security, support, compatibility, and upgrade documentation.

Install

Python CLI:

pipx install 'dbt-cortex-agent[runtime]==0.3.0'

dbt package:

packages:
  - git: https://github.com/Jeremy-Demlow/dbt-cortex-agent.git
    revision: v0.3.0

The full ski-resort reference implementation rebuilt four managed Agents, preserved the legacy Agent, passed canonical skill and semantic-tool smokes, and passed both paid evaluation suites against immutable accepted baselines.

.... Generated with Cortex Code~