# Database Migration Management with `agents db`

## Overview

CleverAgents uses [Alembic](https://alembic.sqlalchemy.org/) to manage its
SQLite database schema. The `agents db` command group exposes all migration
operations directly from the command line, giving administrators full control
over schema inspection, upgrades, and rollbacks — without needing to touch
Alembic directly.

This example walks through the complete database migration lifecycle: checking
the current state of a fresh database, viewing the full migration history,
applying all pending migrations, rolling back a single step, and upgrading
again to the latest revision.

## Prerequisites

- CleverAgents installed (`pip install cleveragents`)
- Python 3.13 or higher

## What You'll Learn

- How to **inspect the current migration state** of any database
- How to **browse the full migration history** (41 revisions across the full
  schema evolution)
- How to **apply pending migrations** to bring a database up to `head`
- How to **roll back** one or more revisions safely
- How to **target a specific revision** for precise upgrade or downgrade
- How to use **`--format json`** to integrate migration status into scripts
  and CI/CD pipelines

---

## Part 1: Exploring the Commands

### Step 1: View available `db` subcommands

```bash
$ agents db --help
```

**Output:**
```
 Usage: agents db [OPTIONS] COMMAND [ARGS]...

 Database migration management (Alembic)

╭─ Options ────────────────────────────────────────────────────────────────────╮
│ --help          Show this message and exit.                                  │
╰──────────────────────────────────────────────────────────────────────────────╯
╭─ Commands ───────────────────────────────────────────────────────────────────╮
│ migrate    Generate a new Alembic migration revision (autogenerate).         │
│ upgrade    Apply pending database migrations up to REVISION (default: head). │
│ downgrade  Roll the database back to REVISION.                               │
│ current    Show the current migration revision of the database.              │
│ history    Show the migration revision history.                              │
╰──────────────────────────────────────────────────────────────────────────────╯
```

**What's Happening:**

The `agents db` group provides five subcommands covering the full migration
lifecycle. All commands accept a `--format` flag (`rich`, `json`, `yaml`,
`plain`, `table`) for flexible output.

---

## Part 2: Inspecting a Fresh Database

### Step 2: Check current revision on an uninitialized database

Point CleverAgents at a brand-new SQLite file and check its migration state:

```bash
$ export CLEVERAGENTS_DATABASE_URL="sqlite:////tmp/my-project.db"
$ agents db current
```

**Output:**
```
Current revision : (none)
Pending migrations: 41
  - 001_initial_schema
  - 4b518923afb2
  - c3d9b3d0cf3e
  - a5_001_actions_v3
  - a5_002_lifecycle_plans
  - a5_003_spec_aligned_actions
  - a5_004_spec_aligned_plans
  - b1_001_resource_registry
  - b1_001_resource_links
  - a5_005_rebaseline_plan_phases
  - b0_001_projects
  - c1_001_tool_registry
  - a7_001_session_persistence
  - a6_001_automation_profiles
  - 71cd40eb661f
  - c0_001_skill_registry
  - a7_002_merge_heads
  - c3_001_actor_registry
  - a6_002_drop_automation_level
  - c0_002_merge_skill_registry
  - m4_001_concurrency_locks
  - d0_001_changeset_artifacts
  - d0_002_merge_changeset_and_locks
  - m4_001_decision_tables
  - m4_002_skill_flattened_tools
  - m6_001_checkpoint_metadata
  - c4_001_safety_profile_column
  - m6_002_merge_safety_and_checkpoint
  - m6_003_async_jobs_table
  - m7_001_repo_indexing_tables
  - m6_004_resource_type_inherits
  - m6_004_container_metadata_column
  - m6_005_profile_guards_json
  - m4_003_plan_env_columns
  - m8_001_align_plans_schema
  - m8_001_correction_attempts
  - m5_001_rename_profile_fields
  - m8_002_merge_profile_rename_and_corrections
  - m6_006_estimation_report_json
  - m9_001_session_name_column
  - m9_002_plan_resume_fields
```

**What's Happening:**

On a fresh database, `current revision` is `(none)` because no migrations have
been applied yet. All 41 revisions are listed as pending. The revision IDs
follow a naming convention that encodes the feature milestone (e.g.,
`a5_001_actions_v3` = milestone A5, migration 001, feature "actions v3").

---

## Part 3: Browsing Migration History

### Step 3: View the full migration history

```bash
$ agents db history
```

**Output (most recent first):**
```
m9_001_session_name_column -> m9_002_plan_resume_fields (head), Add reversion_count, last_completed_step, last_checkpoint_id columns to v3_plans.
m6_006_estimation_report_json -> m9_001_session_name_column, Add name column to sessions table.
m8_002_merge_profile_rename_and_corrections -> m6_006_estimation_report_json, Add estimation_report_json column to v3_plans.
m5_001_rename_profile_fields, m8_001_correction_attempts, m8_001_align_plans_schema -> m8_002_merge_profile_rename_and_corrections (mergepoint), Merge profile-rename, correction-attempts, and plans-schema heads.
m4_003_plan_env_columns -> m5_001_rename_profile_fields, Rename automation profile task-type threshold columns.
m4_003_plan_env_columns -> m8_001_correction_attempts, Create correction_attempts table.
m4_003_plan_env_columns -> m8_001_align_plans_schema, Align v3_plans table schema with specification DDL.
m6_005_profile_guards_json -> m4_003_plan_env_columns (branchpoint), Add execution_environment and execution_env_priority columns to v3_plans.
m6_004_container_metadata_column -> m6_005_profile_guards_json, Add safety_json and guards_json columns to automation_profiles table.
m6_004_resource_type_inherits -> m6_004_container_metadata_column, Add container_metadata_json column to tool_invocations table.
m7_001_repo_indexing_tables -> m6_004_resource_type_inherits, Add inherits column to resource_types for type inheritance (ADR-042).
m6_003_async_jobs_table -> m7_001_repo_indexing_tables, Add repo_indexes and indexed_files tables for repository indexing.
m6_002_merge_safety_and_checkpoint -> m6_003_async_jobs_table, Add async_jobs table for background job execution.
c4_001_safety_profile_column, m6_001_checkpoint_metadata -> m6_002_merge_safety_and_checkpoint (mergepoint), Merge safety profile and checkpoint metadata heads.
c0_002_merge_skill_registry -> c4_001_safety_profile_column, Add safety_profile_json column to actions table.
m4_002_skill_flattened_tools -> m6_001_checkpoint_metadata, Create checkpoint_metadata table.
m4_001_decision_tables -> m4_002_skill_flattened_tools, Add flattened tool persistence columns to skills table.
d0_002_merge_changeset_and_locks -> m4_001_decision_tables, Add decision tree tables.
d0_001_changeset_artifacts, m4_001_concurrency_locks -> d0_002_merge_changeset_and_locks (mergepoint), Merge changeset artifacts and concurrency locks heads.
c0_002_merge_skill_registry -> d0_001_changeset_artifacts, Add changeset_entries and tool_invocations tables.
c0_002_merge_skill_registry -> m4_001_concurrency_locks, Add concurrency locks table.
a6_002_drop_automation_level, c0_001_skill_registry -> c0_002_merge_skill_registry (branchpoint) (mergepoint), Merge skill registry and actor/automation heads.
c3_001_actor_registry -> a6_002_drop_automation_level, Drop automation_level column from v3_plans.
a7_002_merge_heads -> c3_001_actor_registry, Add yaml_text, schema_version, compiled_metadata to actors table.
71cd40eb661f, a7_001_session_persistence -> a7_002_merge_heads (mergepoint), Merge session persistence and resource/automation heads.
71cd40eb661f -> c0_001_skill_registry, Add skill registry tables (skills, skill_items).
a6_001_automation_profiles, b1_001_resource_links -> 71cd40eb661f (branchpoint) (mergepoint), Merge resource_links and automation_profiles branches
c1_001_tool_registry -> a6_001_automation_profiles, Add automation_profiles table.
c1_001_tool_registry -> a7_001_session_persistence, Add sessions and session_messages tables.
b0_001_projects -> c1_001_tool_registry (branchpoint), Add tool registry tables (tools, tool_resource_bindings, validation_attachments).
a5_005_rebaseline_plan_phases -> b0_001_projects, Add ns_projects and project_resource_links tables.
b1_001_resource_registry -> a5_005_rebaseline_plan_phases, Rebaseline plan phase/state enums for Action + Apply terminal states.
b1_001_resource_registry -> b1_001_resource_links, Add resource_links table for validated DAG parent-child links.
a5_004_spec_aligned_plans -> b1_001_resource_registry (branchpoint), Add resource registry tables (resource_types, resources, resource_edges).
a5_003_spec_aligned_actions -> a5_004_spec_aligned_plans, Create spec-aligned plans and related tables.
a5_002_lifecycle_plans -> a5_003_spec_aligned_actions, Replace actions_v3 with spec-aligned actions table.
a5_001_actions_v3 -> a5_002_lifecycle_plans, Add lifecycle_plans table for v3 plan lifecycle.
c3d9b3d0cf3e -> a5_001_actions_v3, Add actions_v3 table for v3 plan lifecycle.
4b518923afb2 -> c3d9b3d0cf3e, Add actors table for actor registry.
001_initial_schema -> 4b518923afb2, add_debug_attempts_table
<base> -> 001_initial_schema, Initial database schema with all core tables.
```

**What's Happening:**

Each line shows `<from_revision> -> <to_revision>, <description>`. The
`(head)` marker identifies the latest revision. `(branchpoint)` and
`(mergepoint)` labels show where the migration DAG diverges and converges —
CleverAgents uses parallel migration branches to allow independent feature
development without blocking each other.

### Step 4: Get history as JSON for scripting

```bash
$ agents db history --format json
```

**Output (excerpt):**
```json
{
  "command": "",
  "status": "ok",
  "exit_code": 0,
  "data": [
    {
      "revision": "m9_002_plan_resume_fields",
      "down_revision": "m9_001_session_name_column",
      "description": "Add reversion_count, last_completed_step, last_checkpoint_id columns to v3_plans."
    },
    {
      "revision": "m9_001_session_name_column",
      "down_revision": "m6_006_estimation_report_json",
      "description": "Add name column to sessions table."
    },
    {
      "revision": "m6_006_estimation_report_json",
      "down_revision": "m8_002_merge_profile_rename_and_corrections",
      "description": "Add estimation_report_json column to v3_plans."
    },
    {
      "revision": "m8_002_merge_profile_rename_and_corrections",
      "down_revision": [
        "m5_001_rename_profile_fields",
        "m8_001_correction_attempts",
        "m8_001_align_plans_schema"
      ],
      "description": "Merge profile-rename, correction-attempts, and plans-schema heads."
    }
  ],
  "timing": {
    "duration_ms": 0
  },
  "messages": [
    {"level": "ok", "text": "ok"}
  ]
}
```

**What's Happening:**

The JSON output exposes the full DAG structure: `down_revision` is a list when
a revision has multiple parents (a merge point). This is useful for tooling
that needs to programmatically inspect the migration graph.

---

## Part 4: Applying Migrations

### Step 5: Upgrade to the latest revision (`head`)

```bash
$ agents db upgrade
```

**Output (stderr shows Alembic progress, stdout shows result):**
```
INFO  [alembic.runtime.migration] Context impl SQLiteImpl.
INFO  [alembic.runtime.migration] Will assume non-transactional DDL.
INFO  [alembic.runtime.migration] Running upgrade  -> 001_initial_schema, Initial database schema with all core tables.
INFO  [alembic.runtime.migration] Running upgrade 001_initial_schema -> 4b518923afb2, add_debug_attempts_table
INFO  [alembic.runtime.migration] Running upgrade 4b518923afb2 -> c3d9b3d0cf3e, Add actors table for actor registry.
INFO  [alembic.runtime.migration] Running upgrade c3d9b3d0cf3e -> a5_001_actions_v3, Add actions_v3 table for v3 plan lifecycle.
INFO  [alembic.runtime.migration] Running upgrade a5_001_actions_v3 -> a5_002_lifecycle_plans, Add lifecycle_plans table for v3 plan lifecycle.
INFO  [alembic.runtime.migration] Running upgrade a5_002_lifecycle_plans -> a5_003_spec_aligned_actions, Replace actions_v3 with spec-aligned actions table.
INFO  [alembic.runtime.migration] Running upgrade a5_003_spec_aligned_actions -> a5_004_spec_aligned_plans, Create spec-aligned plans and related tables.
INFO  [alembic.runtime.migration] Running upgrade a5_004_spec_aligned_plans -> b1_001_resource_registry, Add resource registry tables (resource_types, resources, resource_edges).
INFO  [alembic.runtime.migration] Running upgrade b1_001_resource_registry -> b1_001_resource_links, Add resource_links table for validated DAG parent-child links.
INFO  [alembic.runtime.migration] Running upgrade b1_001_resource_registry -> a5_005_rebaseline_plan_phases, Rebaseline plan phase/state enums for Action + Apply terminal states.
INFO  [alembic.runtime.migration] Running upgrade a5_005_rebaseline_plan_phases -> b0_001_projects, Add ns_projects and project_resource_links tables.
INFO  [alembic.runtime.migration] Running upgrade b0_001_projects -> c1_001_tool_registry, Add tool registry tables (tools, tool_resource_bindings, validation_attachments).
INFO  [alembic.runtime.migration] Running upgrade c1_001_tool_registry -> a7_001_session_persistence, Add sessions and session_messages tables.
INFO  [alembic.runtime.migration] Running upgrade c1_001_tool_registry -> a6_001_automation_profiles, Add automation_profiles table.
INFO  [alembic.runtime.migration] Running upgrade a6_001_automation_profiles, b1_001_resource_links -> 71cd40eb661f, Merge resource_links and automation_profiles branches
INFO  [alembic.runtime.migration] Running upgrade 71cd40eb661f -> c0_001_skill_registry, Add skill registry tables (skills, skill_items).
INFO  [alembic.runtime.migration] Running upgrade 71cd40eb661f, a7_001_session_persistence -> a7_002_merge_heads, Merge session persistence and resource/automation heads.
INFO  [alembic.runtime.migration] Running upgrade a7_002_merge_heads -> c3_001_actor_registry, Add yaml_text, schema_version, compiled_metadata to actors table.
INFO  [alembic.runtime.migration] Running upgrade c3_001_actor_registry -> a6_002_drop_automation_level, Drop automation_level column from v3_plans.
INFO  [alembic.runtime.migration] Running upgrade a6_002_drop_automation_level, c0_001_skill_registry -> c0_002_merge_skill_registry, Merge skill registry and actor/automation heads.
INFO  [alembic.runtime.migration] Running upgrade c0_002_merge_skill_registry -> m4_001_concurrency_locks, Add concurrency locks table.
INFO  [alembic.runtime.migration] Running upgrade c0_002_merge_skill_registry -> d0_001_changeset_artifacts, Add changeset_entries and tool_invocations tables.
INFO  [alembic.runtime.migration] Running upgrade d0_001_changeset_artifacts, m4_001_concurrency_locks -> d0_002_merge_changeset_and_locks, Merge changeset artifacts and concurrency locks heads.
INFO  [alembic.runtime.migration] Running upgrade d0_002_merge_changeset_and_locks -> m4_001_decision_tables, Add decision tree tables.
INFO  [alembic.runtime.migration] Running upgrade m4_001_decision_tables -> m4_002_skill_flattened_tools, Add flattened tool persistence columns to skills table.
INFO  [alembic.runtime.migration] Running upgrade m4_002_skill_flattened_tools -> m6_001_checkpoint_metadata, Create checkpoint_metadata table.
INFO  [alembic.runtime.migration] Running upgrade c0_002_merge_skill_registry -> c4_001_safety_profile_column, Add safety_profile_json column to actions table.
INFO  [alembic.runtime.migration] Running upgrade c4_001_safety_profile_column, m6_001_checkpoint_metadata -> m6_002_merge_safety_and_checkpoint, Merge safety profile and checkpoint metadata heads.
INFO  [alembic.runtime.migration] Running upgrade m6_002_merge_safety_and_checkpoint -> m6_003_async_jobs_table, Add async_jobs table for background job execution.
INFO  [alembic.runtime.migration] Running upgrade m6_003_async_jobs_table -> m7_001_repo_indexing_tables, Add repo_indexes and indexed_files tables for repository indexing.
INFO  [alembic.runtime.migration] Running upgrade m7_001_repo_indexing_tables -> m6_004_resource_type_inherits, Add inherits column to resource_types for type inheritance (ADR-042).
INFO  [alembic.runtime.migration] Running upgrade m6_004_resource_type_inherits -> m6_004_container_metadata_column, Add container_metadata_json column to tool_invocations table.
INFO  [alembic.runtime.migration] Running upgrade m6_004_container_metadata_column -> m6_005_profile_guards_json, Add safety_json and guards_json columns to automation_profiles table.
INFO  [alembic.runtime.migration] Running upgrade m6_005_profile_guards_json -> m4_003_plan_env_columns, Add execution_environment and execution_env_priority columns to v3_plans.
INFO  [alembic.runtime.migration] Running upgrade m4_003_plan_env_columns -> m8_001_align_plans_schema, Align v3_plans table schema with specification DDL.
INFO  [alembic.runtime.migration] Running upgrade m4_003_plan_env_columns -> m8_001_correction_attempts, Create correction_attempts table.
INFO  [alembic.runtime.migration] Running upgrade m4_003_plan_env_columns -> m5_001_rename_profile_fields, Rename automation profile task-type threshold columns.
INFO  [alembic.runtime.migration] Running upgrade m5_001_rename_profile_fields, m8_001_correction_attempts, m8_001_align_plans_schema -> m8_002_merge_profile_rename_and_corrections, Merge profile-rename, correction-attempts, and plans-schema heads.
INFO  [alembic.runtime.migration] Running upgrade m8_002_merge_profile_rename_and_corrections -> m6_006_estimation_report_json, Add estimation_report_json column to v3_plans.
INFO  [alembic.runtime.migration] Running upgrade m6_006_estimation_report_json -> m9_001_session_name_column, Add name column to sessions table.
INFO  [alembic.runtime.migration] Running upgrade m9_001_session_name_column -> m9_002_plan_resume_fields, Add reversion_count, last_completed_step, last_checkpoint_id columns to v3_plans.
Database upgraded to revision: m9_002_plan_resume_fields
```

**What's Happening:**

Alembic applies all 41 pending migrations in topological order, respecting the
DAG structure. Parallel branches (e.g., `a6_001_automation_profiles` and
`b1_001_resource_links`) are applied before their merge point
(`71cd40eb661f`). The final line confirms the database is now at the `head`
revision: `m9_002_plan_resume_fields`.

### Step 6: Verify the database is fully up to date

```bash
$ agents db current
```

**Output:**
```
Current revision : m9_002_plan_resume_fields
Pending migrations: 0
```

**What's Happening:**

`Pending migrations: 0` confirms the database is at `head`. No further
migrations need to be applied.

### Step 7: Get current state as JSON (for CI/CD integration)

```bash
$ agents db current --format json
```

**Output:**
```json
{
  "command": "",
  "status": "ok",
  "exit_code": 0,
  "data": {
    "current_revision": "m9_002_plan_resume_fields",
    "pending_count": 0,
    "pending_revisions": []
  },
  "timing": {
    "duration_ms": 0
  },
  "messages": [
    {"level": "ok", "text": "ok"}
  ]
}
```

**What's Happening:**

The JSON output is ideal for CI/CD pipelines. You can parse `pending_count` to
gate deployments: if `pending_count > 0`, run `agents db upgrade` before
starting the application.

---

## Part 5: Rolling Back Migrations

### Step 8: Downgrade one step (relative rollback)

```bash
$ agents db downgrade -- -1
```

> **Note:** The `--` separator is required when passing negative revision
> identifiers like `-1` so the shell does not interpret them as flags.

**Output:**
```
INFO  [alembic.runtime.migration] Context impl SQLiteImpl.
INFO  [alembic.runtime.migration] Will assume non-transactional DDL.
INFO  [alembic.runtime.migration] Running downgrade m9_002_plan_resume_fields -> m9_001_session_name_column, Add reversion_count, last_completed_step, last_checkpoint_id columns to v3_plans.
Database downgraded to revision: m9_001_session_name_column
```

**What's Happening:**

The `-1` relative identifier rolls back exactly one migration. The database is
now at `m9_001_session_name_column`. The Alembic log shows the description of
the migration that was reversed.

### Step 9: Check state after rollback

```bash
$ agents db current
```

**Output:**
```
Current revision : m9_001_session_name_column
Pending migrations: 1
  - m9_002_plan_resume_fields
```

**What's Happening:**

The database is now one revision behind `head`. The single pending migration
(`m9_002_plan_resume_fields`) is the one that was just rolled back.

---

## Part 6: Targeted Upgrade and Downgrade

### Step 10: Downgrade to a specific revision by ID

```bash
$ agents db downgrade m9_001_session_name_column
```

**Output:**
```
INFO  [alembic.runtime.migration] Context impl SQLiteImpl.
INFO  [alembic.runtime.migration] Will assume non-transactional DDL.
INFO  [alembic.runtime.migration] Running downgrade m9_002_plan_resume_fields -> m9_001_session_name_column, Add reversion_count, last_completed_step, last_checkpoint_id columns to v3_plans.
Database downgraded to revision: m9_001_session_name_column
```

### Step 11: Upgrade to a specific revision by ID

```bash
$ agents db upgrade m9_002_plan_resume_fields
```

**Output:**
```
INFO  [alembic.runtime.migration] Context impl SQLiteImpl.
INFO  [alembic.runtime.migration] Will assume non-transactional DDL.
INFO  [alembic.runtime.migration] Running upgrade m9_001_session_name_column -> m9_002_plan_resume_fields, Add reversion_count, last_completed_step, last_checkpoint_id columns to v3_plans.
Database upgraded to revision: m9_002_plan_resume_fields
```

**What's Happening:**

Both `upgrade` and `downgrade` accept explicit revision IDs. This is useful
when you need to pin a database to a known-good state, for example during
staged rollouts or when bisecting a regression.

### Step 12: Upgrade with JSON output (for scripting)

```bash
$ agents db upgrade --format json 2>/dev/null
```

**Output (when already at head — idempotent):**
```json
{
  "command": "",
  "status": "ok",
  "exit_code": 0,
  "data": {
    "status": "ok",
    "current_revision": "m9_002_plan_resume_fields"
  },
  "timing": {
    "duration_ms": 0
  },
  "messages": [
    {"level": "ok", "text": "ok"}
  ]
}
```

**What's Happening:**

`agents db upgrade` is **idempotent**: running it when the database is already
at `head` is a no-op. The JSON output confirms `status: ok` and the current
revision. The `2>/dev/null` suppresses the Alembic INFO log lines from stderr
when you only want the structured result.

---

## Complete Interaction Log

<details>
<summary>Click to see the full session transcript</summary>

```bash
# Set up a fresh test database
$ export CLEVERAGENTS_DATABASE_URL="sqlite:////tmp/my-project.db"

# 1. View available commands
$ agents db --help
 Usage: agents db [OPTIONS] COMMAND [ARGS]...
 Database migration management (Alembic)
 ...

# 2. Check state of fresh database
$ agents db current
Current revision : (none)
Pending migrations: 41
  - 001_initial_schema
  - 4b518923afb2
  ... (41 revisions listed)

# 3. Browse migration history
$ agents db history
m9_001_session_name_column -> m9_002_plan_resume_fields (head), ...
...
<base> -> 001_initial_schema, Initial database schema with all core tables.

# 4. Apply all migrations
$ agents db upgrade
INFO  [alembic.runtime.migration] Running upgrade  -> 001_initial_schema, ...
... (41 migrations applied)
Database upgraded to revision: m9_002_plan_resume_fields

# 5. Confirm fully up to date
$ agents db current
Current revision : m9_002_plan_resume_fields
Pending migrations: 0

# 6. Get current state as JSON
$ agents db current --format json
{"data": {"current_revision": "m9_002_plan_resume_fields", "pending_count": 0, ...}}

# 7. Roll back one step
$ agents db downgrade -- -1
INFO  [alembic.runtime.migration] Running downgrade m9_002_plan_resume_fields -> m9_001_session_name_column, ...
Database downgraded to revision: m9_001_session_name_column

# 8. Check state after rollback
$ agents db current
Current revision : m9_001_session_name_column
Pending migrations: 1
  - m9_002_plan_resume_fields

# 9. Upgrade back to head
$ agents db upgrade head
INFO  [alembic.runtime.migration] Running upgrade m9_001_session_name_column -> m9_002_plan_resume_fields, ...
Database upgraded to revision: m9_002_plan_resume_fields
```

</details>

---

## Key Takeaways

1. **`agents db current`** is your go-to command for checking database health.
   It shows both the current revision and the count of pending migrations.

2. **`agents db history`** reveals the full migration DAG, including
   branchpoints and mergepoints. CleverAgents uses parallel migration branches
   to allow independent feature development.

3. **`agents db upgrade`** (no arguments) always upgrades to `head`. It is
   **idempotent** — safe to run on every application startup.

4. **`agents db downgrade -- -1`** rolls back exactly one step. Use `--` to
   separate the negative relative identifier from CLI flags.

5. **All commands support `--format json`** for scripting and CI/CD
   integration. Parse `pending_count` to gate deployments.

6. **The `CLEVERAGENTS_DATABASE_URL` environment variable** overrides the
   default database path. This is the recommended way to point commands at a
   specific database file.

---

## Try It Yourself

```bash
# Initialize a fresh workspace and inspect its migration state
mkdir ~/my-agents-project && cd ~/my-agents-project
agents project init my-agents-project --yes

# The project init automatically runs all migrations.
# Verify the database is at head:
agents db current

# Browse the schema evolution history:
agents db history

# Simulate a rollback scenario:
agents db downgrade -- -1
agents db current          # Shows 1 pending migration
agents db upgrade          # Re-applies the rolled-back migration
agents db current          # Back to head, 0 pending
```

---

*This example was automatically generated and verified by the CleverAgents UAT system.*
*Feature area: Database migration management | Test cycle: 1*

---
**Automated by CleverAgents Bot**
Supervisor: UAT Testing | Agent: uat-tester