vb10/vexana/add

vexana add

Targets vexana ^6.0.0. Read references/vexana-api.md and references/templates.md first; generate code only from those templates.

0. Preconditions

• Find the network setup (AppNetworkManager or a NetworkManager< construction in lib/). If there is none, stop and tell the user to run /vexana:init.
• Take {{E}} from that setup (EmptyModel or the generated error model), {{networkPath}} from its folder, and {{package}} from pubspec.yaml name.
• Features root: the first of lib/features/, lib/feature/, lib/src/features/ that exists, even if it is empty; if none exists, lib/features/. {{featurePath}} is that root without lib/, plus the feature name: features/post.
• If the project uses json_serializable or freezed, still generate from the "Model" template, and tell the user the model was written by hand so they can convert it to their generator.
• Write every file under the project root you are working in, using paths relative to it.

1. Inputs

• JSON: pasted text, a file path, or an existing Dart model class to convert to INetworkModel.
• Read large JSON files only as far as needed to infer types (e.g. the first elements plus a key scan); do not echo them back.
• Optional: endpoint(s) as METHOD /path (GET /posts, GET /posts/{id}), feature name (default: the model name in snake_case), model name (default: from the path's last static segment, singular UpperCamel: /posts → Post).
• Invalid JSON → stop and show the parse error position. Do not guess.

2. Infer types

Apply the "Type inference" table and naming rules in templates.md. For a top-level array, infer from all elements. Collect every Object? and List<Object?>? field, and every key whose Dart field name differs from the key, for the report.

3. Check for collisions

If a class with the same name exists anywhere in lib/, or a target file exists, ask the user (reuse, rename, or skip) before writing anything. Do not write files for that input until they answer.

4. Write files

1. Model file (+ nested classes in the same file) from "Model".
2. If endpoints were given: create <feature>_service.dart from "Service" with only those methods, or add the methods (and interface entries) to the existing service.
3. Save the JSON sample to test/fixtures/<model_snake>.json, values unchanged:

• input is a file → copy it with cp <input> test/fixtures/<model_snake>.json. Never re-type a file's contents: a large API response would be regenerated token by token.
• input was pasted → write it as given.

1. Model test from "Model test" (array form for array samples). Make sure test/helpers/fixture_reader.dart has dropNulls.

5. Verify

dart analyze
flutter test test/{{featurePath}}/

Fix and retry at most twice; then report failures verbatim. Never edit a fixture or weaken an assertion to make a test pass.

6. Report

Files created/changed, the service methods added, values replaced in a fake sample, fields typed Object? / List<Object?>? (the sample did not show their type — ask the user to tighten them), and renamed keys (JSON key → Dart field).