Amin Mousavi 336fc76035 feat(backend): OpenRouter voice providers and per-locale registry
ASR and extraction are separate, independently swappable roles resolved per
locale from config. All three locales point at the same OpenRouter models today
(whisper-1, gemini-3.7-flash); the indirection stays because Persian ASR is the
weakest link and repointing only `fa` must not be a code change.

The model emits a deliberately flat wire shape rather than the internal
discriminated unions — strict json_schema mode has poor union support — and
toVoiceIntent narrows it. That normalizer is total: a missing or malformed
payload yields a shape the resolvers report as unresolved rather than one that
throws.

The prompt supplies catalog codes with labels in the actor's locale, so the
model matches spoken words rather than translating, and carries per-locale
tooth vocabulary. English gets an explicit warning that a bare two-digit number
is ambiguous under Universal numbering, and must not be treated as FDI unless
the speaker said so.

From review of this commit:
- only an actually FDI-shaped code takes the explicit branch; fdi:"6" alongside
  valid arch/side/position used to lose the tooth entirely
- an unrecognised due kind passes through to be flagged, instead of collapsing
  to null and looking like no deadline was ever spoken
- vendor error bodies stay out of the thrown message and the default log level;
  a 4xx can echo the request back, transcript included
- the chat call sets provider.require_parameters so OpenRouter only routes to
  endpoints that honour the JSON schema, rather than ones treating it as a hint
- an unknown locale in VOICE_ENABLED_LOCALES now fails at boot like an unknown
  provider id, instead of silently disabling the microphone everywhere

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-23 23:05:30 +03:30

Dyolink

Monorepo: NestJS backend (backend/), Next.js frontend (frontend/), Docker stack under infrastructure/..

Local development: see backend/README.md and frontend/README.md.

Cursor AI: project conventions for agents are in AGENTS.md, .cursor/rules/, and .cursor/skills/.


Production deploy (Docker Hub + HTTPS + Let's Encrypt)

Full step-by-step guide: infrastructure/DEPLOY.md

Minimal server setup: install Docker, create .env + secrets/, docker login, run one script.

On server (once) In repo / Docker
DNS A record → server IP docker-compose.prod.yml, nginx, certbot
docker login (private Hub) Build & push images from dev machine
secrets/database.env, secrets/backend.env Examples: database.prod.env.example, backend.prod.env.example
infrastructure/.env (DOMAIN, LETSENCRYPT_EMAIL) deploy.prod.env.example

Dev machine — build frontend with the public domain baked in, push to Docker Hub:

./infrastructure/scripts/build-and-push-prod.sh wixur.ir latest

Server — from infrastructure/:

cp deploy.prod.env.example .env          # edit DOMAIN, paths
mkdir -p ../secrets && cp database.prod.env.example ../secrets/database.env
cp backend.prod.env.example ../secrets/backend.env   # set passwords + FRONTEND_URL
docker login
chmod +x scripts/*.sh
./scripts/deploy-prod.sh

SSL is issued automatically via Certbot (scripts/init-letsencrypt.sh). Nginx config is generated from DOMAIN in .env. When you move to another domain (e.g. dyolink.com), update .env + backend.env, re-run init-letsencrypt.sh, and rebuild the frontend image with the new URL.


Deploy on your own server (Docker + Gitea)

High level: build container images → push to a registry → server pulls images and runs Compose. Optionally Gitea Actions automates that on every merge to main / master.

1. One-time server preparation

  1. Install Docker and Docker Compose on the server.

  2. Run Gitea with the container registry enabled (same host/port you use for docker login, e.g. 178.131.50.201:3000).

  3. Copy the repo (or deploy only infrastructure/ + secrets). You need at least:

    • infrastructure/docker-compose.registry.yml
    • infrastructure/nginx/ configs referenced by that compose file
    • infrastructure/database/init.sql if used by your Postgres service
  4. Secrets on the server (never commit real values):

    • Copy infrastructure/database.staging.env.exampledatabase.staging.env (Postgres user/password/db).
    • Copy infrastructure/backend.staging.env.examplebackend.staging.env (e.g. DATABASE_URL, JWT, pointing at the compose Postgres service name).
    • Put both files in one directory on the server, e.g. /opt/dyolink/secrets/.
  5. Registry login from the server (same credentials you use for docker push):

    docker login <registry-host>:<port> -u <user>
    

    For HTTP registries, Docker may require insecure-registries on the daemon.

2. Manual deploy (build images elsewhere, run on server)

On your dev machine (after successful local builds):

$REG = "<registry-host>:<port>"
$OWN = "<registry-owner>"
$TAG = "manual"

docker build -t "${REG}/${OWN}/dyolink-backend:${TAG}" -t "${REG}/${OWN}/dyolink-backend:latest" ./backend

docker build `
  --build-arg NEXT_PUBLIC_API_URL="http://<your-public-ip>:<nginx-port>/api" `
  --build-arg NEXT_PUBLIC_APP_URL="http://<your-public-ip>:<nginx-port>" `
  --build-arg NEXT_PUBLIC_APP_NAME="Dyolink" `
  -t "${REG}/${OWN}/dyolink-frontend:${TAG}" `
  -t "${REG}/${OWN}/dyolink-frontend:latest" `
  ./frontend

docker push "${REG}/${OWN}/dyolink-backend:${TAG}"
docker push "${REG}/${OWN}/dyolink-backend:latest"
docker push "${REG}/${OWN}/dyolink-frontend:${TAG}"
docker push "${REG}/${OWN}/dyolink-frontend:latest"

On the server, from infrastructure/:

  1. Create deploy.registry.env (see infrastructure/deploy.registry.env.example):

    • REGISTRY_PREFIX=<host>:<port>/<owner> (no http://, no trailing slash)
    • IMAGE_TAG=latest or the tag you pushed
    • STAGING_HTTP_PORT=<host port> (e.g. 8088 — browser uses http://<ip>:8088)
  2. Set DEPLOY_SECRETS_DIR to the absolute path of the folder containing database.staging.env and backend.staging.env (you can export it in the shell or add it to deploy.registry.env if your Compose setup expects it).

  3. Pull and start:

    docker compose -f docker-compose.registry.yml --env-file deploy.registry.env pull backend frontend
    docker compose -f docker-compose.registry.yml --env-file deploy.registry.env up -d
    

The backend container runs prisma migrate deploy on startup (via entrypoint) when NODE_ENV=production, so schema updates apply after you deploy a new image that includes new migrations.

3. Automatic deploy (Gitea Actions)

Workflow file: .gitea/workflows/registry-build-deploy.yml.

Requirements:

  • Gitea Actions enabled for the repository.

  • A self-hosted runner (with Docker) registered to Gitea — the workflow uses runs-on: self-hosted.

  • Windows runners: the workflow uses PowerShell (not Bash). Giteas runner was failing with execvpe(/bin/bash) failed when Bash was routed through WSL without a real /bin/bash. If your runner is Linux, switch .gitea/workflows/registry-build-deploy.yml to defaults.run.shell: bash and use Bash syntax instead.

  • Repository → Actions → Variables (examples):

    • REGISTRY_HOST — e.g. 178.131.50.201:3000
    • REGISTRY_OWNER — image namespace (same as Docker image path after the host), e.g. admin
    • PUBLIC_BASE_URL — URL users open in the browser, e.g. http://178.131.50.201:8088 (no trailing slash)
    • DEPLOY_SECRETS_DIRabsolute path on the runner machine to the folder containing database.staging.env and backend.staging.env
    • Optional: STAGING_HTTP_PORT (defaults to 8088)
  • Repository → Actions → Secrets:

    • REGISTRY_USERNAME
    • REGISTRY_PASSWORD — access token with package read/write (or equivalent)

Trigger: push to main or master, or run the workflow manually (workflow_dispatch).

The pipeline clones from your Gitea instance, builds and pushes backend/frontend images, then on the runner runs docker compose pull and up -d using infrastructure/docker-compose.registry.yml.


Path Role
backend/Dockerfile API image
frontend/Dockerfile Web image
infrastructure/docker-compose.registry.yml Pull-only staging stack (registry images + nginx + postgres)
infrastructure/deploy.registry.env.example Template for deploy.registry.env
Description
The whole dockerized structure of the project repo
Readme 7.8 MiB
Languages
TypeScript 97.5%
Shell 1.1%
CSS 0.8%
JavaScript 0.4%
Dockerfile 0.2%