How to build a changelog with opencode

{
  "mcp": {
    "databased": {
      "type": "remote",
      "url": "https://databasedmcp.com/mcp",
      "oauth": false,
      "headers": {
        "Authorization": "Bearer {env:DATABASED_API_KEY}"
      }
    }
  }
}

Set up a changelog for me in databasedmcp, in one go.

1. Create a database called changelog, described as "Changelog: user-facing changes logged as they land, grouped into releases with drafted notes".
2. Create two tables:
   - releases: id, version (required, unique), released_on (date, defaults to today), status (draft or published, defaults to draft), notes (the drafted release notes in markdown).
   - entries: id, kind (added, changed, fixed or removed), summary (required, one sentence a customer understands, no internal names), area, pr (the pull request number), release_id (references releases, NULL until the change ships), created_at (defaults to now).
3. Write the database a guide that says what each table is for, to log an entry when a change users would notice is merged, and to skip refactors and internal tooling. Include the query for "unreleased": entries with no release yet, grouped by kind. And say how to cut a release: insert it, attach the unreleased entries, read them back grouped by kind, and write the notes to releases.notes with added first, then changed, fixed and removed.

From then on: when we finish something users would notice, log an entry with its kind, a customer-facing summary and the PR. When I cut a release, draft the notes from its entries, and set it to published once I've approved them. Show me the tables when you're finished.

CREATE TABLE releases (
  id INTEGER PRIMARY KEY,
  version TEXT NOT NULL UNIQUE,
  released_on TEXT NOT NULL DEFAULT (date('now')),
  status TEXT NOT NULL DEFAULT 'draft' CHECK (status IN ('draft', 'published')),
  notes TEXT
);

CREATE TABLE entries (
  id INTEGER PRIMARY KEY,
  kind TEXT NOT NULL CHECK (kind IN ('added', 'changed', 'fixed', 'removed')),
  summary TEXT NOT NULL,
  area TEXT,
  pr INTEGER,
  release_id INTEGER REFERENCES releases(id),
  created_at TEXT NOT NULL DEFAULT CURRENT_TIMESTAMP
);

# changelog: entries and releases

- **entries**: one per user-facing change. kind: added, changed, fixed, removed. `summary` is one sentence a customer understands, no internal names. `pr` is the pull request number. `release_id` stays NULL until the change ships.
- **releases**: one per version. `notes` is the drafted release notes in markdown; status: draft, published.

Log an entry when a change users would notice is merged. Skip refactors and internal tooling.

Unreleased: `SELECT id, kind, summary, pr FROM entries WHERE release_id IS NULL ORDER BY kind, id`

To cut a release: insert it, attach the unreleased entries, read them back grouped by kind, and write the notes to `releases.notes`. Added first, then changed, fixed, removed.

INSERT INTO entries (kind, summary, area, pr) VALUES (?, ?, ?, ?)
-- params: ["added","Export any report as a CSV file from the report's menu.","reports",212]

INSERT INTO entries (kind, summary, area, pr) VALUES (?, ?, ?, ?)
-- params: ["fixed","The date picker no longer jumps to January when you change the year.","forms",215]

INSERT INTO releases (version) VALUES (?) RETURNING id
-- params: ["2.4.0"]

UPDATE entries SET release_id = ? WHERE release_id IS NULL
-- params: [1]

UPDATE releases SET notes = ?, status = 'published' WHERE id = ?
-- params: ["## 2.4.0\n\n**Added**\n- Export any report as a CSV file from the report's menu.\n\n**Fixed**\n- The date picker no longer jumps to January when you change the year.",1]

SELECT id, kind, summary, pr FROM entries WHERE release_id IS NULL ORDER BY kind, id

SELECT r.version, r.released_on, sum(e.kind = 'added') AS added, sum(e.kind = 'changed') AS changed, sum(e.kind = 'fixed') AS fixed, sum(e.kind = 'removed') AS removed FROM releases r LEFT JOIN entries e ON e.release_id = r.id WHERE r.released_on >= date('now', 'start of year') GROUP BY r.id ORDER BY r.released_on DESC

SELECT r.version, r.released_on, e.kind, e.summary FROM entries e JOIN releases r ON r.id = e.release_id WHERE e.summary LIKE ? OR e.area LIKE ? ORDER BY r.released_on DESC
-- params: ["%export%","%export%"]