blog-rust-postgres/README.md

blog-rust-postgres

A minimalist blog engine API in Rust: axum + tokio, Postgres via sqlx (runtime queries, no macros, no ORM), argon2 password hashing and HS256 JWTs. Five small source files you can read in one sitting.

Run

git clone https://git.devai.io/templates/blog-rust-postgres.git
cd blog-rust-postgres
docker compose up --build

The API answers on http://localhost:8080 (curl localhost:8080/health → ok). Postgres keeps its state in ./data/postgres; schema.sql is applied on every start (every statement is IF NOT EXISTS), so there is no migrate step.

Without Docker: point DATABASE_URL at any Postgres 13+, copy .env.example to .env (it is loaded on start), then cargo run (Rust 1.94+).

How it works

Method	Path	Auth	Result
GET	/health	—	200 ok
POST	/auth/register	—	{email, password} → 201 {id, email}, 409 if taken
POST	/auth/login	—	{email, password} → 200 {token}, 401 if wrong
GET	/posts	—	200 [{id, title, slug, excerpt, published_at}], published only
GET	/posts/{slug}	—	200 full published post, or 404
POST	/posts	JWT	{title, body} → 201 full post (a draft)
PUT	/posts/{id}	JWT	{title?, body?, published?} → 200 full post
DELETE	/posts/{id}	JWT	204

• Auth — register stores an argon2id hash (passwords of 8+ characters); login returns an HS256 JWT signed with AUTH_SECRET (sub = user id, 7-day expiry). Send it as Authorization: Bearer <token>. The AuthUser extractor in src/auth.rs accepts only HS256, requires exp and sub, and rejects expired tokens.
• Ownership — only a post's author may update or delete it; anyone else gets 403.
• Slugs come from the title ("Hello, World!" → hello-world); a duplicate title gets -2, -3, … Retitling a post regenerates its slug.
• Drafts — new posts are unpublished; PUT {"published": true} publishes. Public endpoints only return published posts. excerpt is the first 200 characters of the body; published_at is stamped the first time a post is published (null until then) and kept through later edits and unpublish/republish; the list is newest first by it.
• Errors are {"error": "message"} with a matching status code.

A full round trip:

curl -s localhost:8080/auth/register -H 'content-type: application/json' \
  -d '{"email":"me@example.com","password":"sup3rsecret"}'
TOKEN=$(curl -s localhost:8080/auth/login -H 'content-type: application/json' \
  -d '{"email":"me@example.com","password":"sup3rsecret"}' | jq -r .token)
ID=$(curl -s localhost:8080/posts -H "Authorization: Bearer $TOKEN" -H 'content-type: application/json' \
  -d '{"title":"Hello, World!","body":"First post."}' | jq -r .id)
curl -s -X PUT localhost:8080/posts/$ID -H "Authorization: Bearer $TOKEN" -H 'content-type: application/json' \
  -d '{"published":true}'
curl -s localhost:8080/posts/hello-world

Layout

src/main.rs    env, router, startup
src/db.rs      Postgres pool + schema.sql on startup
src/auth.rs    register/login, argon2, JWT issue + the AuthUser extractor
src/posts.rs   post handlers, slugs, ownership checks
src/error.rs   {"error": ...} responses and a JSON extractor that uses them
schema.sql     users + posts tables
extras/        drop-in Clerk and Auth0 verifiers, each with swap steps

Local auth lives entirely in src/auth.rs, so it can be replaced wholesale: extras/auth-clerk/ and extras/auth-auth0/ each hold one JWKS-based RS256 verifier and a README with the exact steps.

Deploy

Push to your own GitHub repo and the shipped workflow (.github/workflows/ci.yml) tests the compose stack, publishes the image to GHCR, and — once you set the DEPLOY_HOST / DEPLOY_USER variables and DEPLOY_KEY secret — deploys it to your server over ssh.

---

Part of devai.io — the blog engine series: one API contract, eight backends (blog-{go,rust,zig,python}-{postgres,mongo}) and the blog-react, blog-angular, blog-dart and blog-flutter frontends.