Connecting an own server asked Ollama's /api/tags and gave up on any answer that was not a 404, and every failure, a refused key included, read as "not responding". The model list is now looked for at /api/tags, at the address as entered and at /v1/models, an answer only counts when it is a list, and a refused key or a server that answered without a list say so (#2808). AI parsing through an own server always called Ollama's /api/chat, which an OpenAI-compatible server such as Groq answers with 404. A server without it now gets the same extraction over /chat/completions at the address as entered (#2806).
@trek/shared
Single source of truth for TREK's API contracts, expressed as Zod schemas and consumed by both the server (request validation + inferred DTO types) and the client (typed requests/responses).
This package is part of the incremental NestJS + React 19 migration (see the "Brownfield Rewrite" board). It is intentionally dormant until modules start importing it — adding it changes nothing for users.
Rules
- One folder per domain:
src/<domain>/<domain>.schema.ts(+.spec.ts). - Domain-agnostic building blocks live in
src/common/. - A route is only considered migrated once its contract lives here.
- Schemas are the source of truth; server DTOs and client types are inferred from them
(
z.infer<typeof schema>), never hand-duplicated.
Consumption (dev)
Both apps resolve @trek/shared to this package's TypeScript source:
- Server (
tsx): viapathsinserver/tsconfig.json. - Client (
vite): viaresolve.aliasinclient/vite.config.ts(+pathsfor the type-checker).
Production packaging (Docker / workspace wiring) is introduced in card F2, when the server first depends on this package at runtime. Until then prod builds are untouched.
Not yet here
The canonical error envelope is finalised in card F5 (it must match TREK's current Express error responses byte-for-byte), so it is deliberately not invented in F1.