Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 5 additions & 6 deletions apps/desktop/CHANGELOG.md
Original file line number Diff line number Diff line change
Expand Up @@ -19,12 +19,11 @@ but they share the same engine, so most Git behaviour lands in both at once.
command to run instead.

### Changed
- **You choose the OpenAI model.** A new OpenAI or Azure OpenAI connection no
longer comes with models filled in; each model field lists what the
provider offers for it, and the connection is ready once you pick. A
connection keeps the model it is on — none is swapped for you when OpenAI
retires one. Adding a Custom connection no longer says "Ready to use" before
it has a model.
- **Change a connection's models from the provider's list.** Each model
field lists what the provider offers for that connection; a new connection
still starts on its provider's defaults. Azure OpenAI now starts on
OpenAI's current models instead of gpt-4o. Adding a Custom connection no
longer says "Ready to use" before it has a model.

## [2.9.0] - 2026-11-07

Expand Down
11 changes: 5 additions & 6 deletions apps/desktop/src/renderer/aiSettings.ts
Original file line number Diff line number Diff line change
Expand Up @@ -206,9 +206,9 @@ function buildEditor(editor: HTMLElement, c: AiConnectionView, refresh: () => Pr
// CLI connections have no base URL or API key — they use the local binary's own
// login. They just take optional model overrides (mapped to `--model`).
const urlF = settingsField(l10n.t("API base URL"), c.baseUrl, "https://api.example.com/v1");
// An HTTP connection's models are CHOSEN, never seeded for you where the
// preset has none (OpenAI, Azure, Custom): each field offers what the
// provider lists for this connection (ai:models — live, then known ids).
// An HTTP connection starts on its provider's own defaults (Custom has
// none), and they are yours to change: each field offers what the provider
// lists for this connection (ai:models — live, then known ids).
const choosePlaceholder = l10n.t("Choose from the list, or type a model ID");
const fastF = settingsField(
isCli ? l10n.t("Quick model (optional)") : l10n.t("Fast model"),
Expand Down Expand Up @@ -391,9 +391,8 @@ async function openGallery(body: HTMLElement, refresh: () => Promise<void>): Pro
// one did not, and it is the only way to connect the providers that need no
// key at all. Their users could never lift the Assistant's gate.
announceAiChanged();
// What is left to do. A preset that seeds no model (OpenAI, Azure,
// Custom) needs one chosen — "Ready to use" was said of Custom, which
// was not.
// What is left to do. A preset that seeds no model (Custom) needs one
// chosen — "Ready to use" was said of Custom, which was not.
const needsModel = p.wire !== "cli" && !p.models.fast && !p.models.mid && !p.models.deep;
const todo = p.needsKey
? needsModel
Expand Down
18 changes: 4 additions & 14 deletions apps/desktop/test/aiBridge.settings.test.ts
Original file line number Diff line number Diff line change
Expand Up @@ -117,19 +117,9 @@ test("removing the default connection promotes the next one and deletes its key"
assert.equal(v.defaultId, undefined, "nothing left to default to");
});

/**
* An OpenAI connection with its models picked. The preset seeds none — you
* choose the model — so a test that sends a request picks them first.
*/
async function openAiPicked(bridge: Awaited<ReturnType<typeof makeBridge>>["bridge"]): Promise<string> {
const id = (await bridge.addConnection("openai")).connections[0].id;
await bridge.updateConnection({ id, models: { fast: "gpt-6-luna", mid: "gpt-6.1-terra", deep: "gpt-6.1-sol" } });
return id;
}

test("a key makes a keyed connection usable, and is never shown or written in plain text", async () => {
const { bridge } = await makeBridge();
const id = await openAiPicked(bridge);
const id = (await bridge.addConnection("openai")).connections[0].id;
const v = await bridge.setKey(id, " sk-live-1234567890 ");
assert.equal(v.connections[0].hasKey, true);
assert.equal(v.connections[0].usable, true);
Expand All @@ -151,7 +141,7 @@ test("a key makes a keyed connection usable, and is never shown or written in pl

test("Test: the provider's answer about the connection is a condition; a request it could not read still reports (#28, #29)", async () => {
const { bridge } = await makeBridge();
const id = await openAiPicked(bridge);
const id = (await bridge.addConnection("openai")).connections[0].id;
await bridge.setKey(id, "sk-wrong");
// A refused key, a rate limit: the Test button's whole answer.
for (const status of [401, 429]) {
Expand Down Expand Up @@ -426,7 +416,7 @@ test("Test: a reasoning model that spent the whole cap thinking still answered

test("Test explains a rejected key in the provider's words", async () => {
const { bridge } = await makeBridge();
const id = await openAiPicked(bridge);
const id = (await bridge.addConnection("openai")).connections[0].id;
await bridge.setKey(id, "sk-bad");
model.replies.push(new Response("{}", { status: 401 }));
const r = await bridge.test(id);
Expand All @@ -448,7 +438,7 @@ test("with no connection the model picker is empty", async () => {

test("the model picker offers configured models, then live ones, then known ones — once each", async () => {
const { bridge } = await makeBridge();
const id = await openAiPicked(bridge);
const id = (await bridge.addConnection("openai")).connections[0].id;
await bridge.setKey(id, "sk-k");
model.models = new Response(JSON.stringify({ data: [{ id: "gpt-6-luna" }, { id: "gpt-live-1" }, { name: "named-model" }, {}] }));
const ids = (await bridge.listModels(id)).map((m) => m.id);
Expand Down
17 changes: 10 additions & 7 deletions apps/extension/CHANGELOG.md
Original file line number Diff line number Diff line change
Expand Up @@ -7,13 +7,16 @@ All notable changes to **GitStudio** are documented here. This project adheres t
## [Unreleased]

### Changed
- **You choose the OpenAI model.** The OpenAI model settings no longer
default to gpt-4o-mini / gpt-4o, and the AI panel's OpenAI card no longer
comes filled in with gpt-4o: the model field lists what your key can use,
and Connect asks for one until you pick it. A detected list is offered,
never picked from (it took the first, alphabetically, for some keys). A
setup already running on the old defaults keeps them as its own settings,
so nothing switches off on update.
- **Every provider starts on its own default model, and you can change it.**
The model settings used to default to OpenAI's gpt-4o-mini / gpt-4o for
every endpoint, so Groq, OpenRouter or a local server was sent OpenAI's
model. Left empty, they now mean the defaults of the provider your base URL
points at — the same ones the desktop app uses (OpenAI's are now
gpt-6-luna / gpt-6.1-terra / gpt-6.1-sol). Each card in the AI panel
starts on its provider's default and lists the models you can switch to. A
detected list no longer replaces a cloud default (it took the first,
alphabetically — "babbage-002" for an OpenAI key); for Ollama or LM Studio
the default becomes a model the server actually has.

### Fixed
- **OpenAI's reasoning models work.** The o-series and gpt-5 and later
Expand Down
2 changes: 1 addition & 1 deletion apps/extension/package.nls.cs.json
Original file line number Diff line number Diff line change
Expand Up @@ -154,7 +154,7 @@
"config.gitstudio.ai.commitStyle.enumDescriptions.2": "Rozkazovací předmět plus krátké vysvětlující tělo.",
"config.gitstudio.ai.openai.baseUrl.markdownDescription": "Základní URL pro koncový bod kompatibilní s OpenAI (přidá se `/chat/completions`). Příklady: `https://api.openai.com/v1` (OpenAI), `https://openrouter.ai/api/v1` (OpenRouter), `http://localhost:11434/v1` (Ollama), `http://localhost:1234/v1` (LM Studio).",
"config.gitstudio.ai.openai.modelDeep.description": "ID modelu kompatibilního s OpenAI pro volitelné hluboké uvažování.",
"config.gitstudio.ai.openai.modelFast.description": "ID modelu kompatibilního s OpenAI pro rychlé/levné úlohy (zprávy commitů). U místního serveru použijte jeho název modelu, např. llama3.1 nebo qwen2.5-coder.",
"config.gitstudio.ai.openai.modelFast.description": "ID modelu kompatibilního s OpenAI pro rychlé/levné úlohy (zprávy commitů). Když necháte všechna tři nastavení modelu prázdná, použijí se výchozí modely vašeho poskytovatele. U místního serveru použijte jeho název modelu, např. llama3.1 nebo qwen2.5-coder.",
"config.gitstudio.ai.openai.modelMid.description": "ID modelu kompatibilního s OpenAI pro úlohy střední úrovně (vysvětlení diffu, shrnutí).",
"config.gitstudio.ai.provider.enumDescriptions.0": "Preferovat jazykový model VS Code bez klíče (GitHub Copilot / Cursor), pokud je dostupný; jinak Anthropic (pokud je nastaven klíč), potom koncový bod kompatibilní s OpenAI (pokud je nakonfigurován model).",
"config.gitstudio.ai.provider.enumDescriptions.1": "Použít API jazykového modelu VS Code — modely GitHub Copilot nebo Cursor, bez klíče. Konkrétní model vyberte přes „GitStudio: Select AI Model…“.",
Expand Down
2 changes: 1 addition & 1 deletion apps/extension/package.nls.de.json
Original file line number Diff line number Diff line change
Expand Up @@ -154,7 +154,7 @@
"config.gitstudio.ai.commitStyle.enumDescriptions.2": "Imperativer Betreff plus ein kurzer erklärender Body.",
"config.gitstudio.ai.openai.baseUrl.markdownDescription": "Basis-URL für den OpenAI-kompatiblen Endpunkt (`/chat/completions` wird angehängt). Beispiele: `https://api.openai.com/v1` (OpenAI), `https://openrouter.ai/api/v1` (OpenRouter), `http://localhost:11434/v1` (Ollama), `http://localhost:1234/v1` (LM Studio).",
"config.gitstudio.ai.openai.modelDeep.description": "OpenAI-kompatible Modell-ID für optionales tiefes Reasoning.",
"config.gitstudio.ai.openai.modelFast.description": "OpenAI-kompatible Modell-ID für schnelle/günstige Aufgaben (Commit-Nachrichten). Für einen lokalen Server den Modellnamen verwenden, z. B. llama3.1 oder qwen2.5-coder.",
"config.gitstudio.ai.openai.modelFast.description": "OpenAI-kompatible Modell-ID für schnelle/günstige Aufgaben (Commit-Nachrichten). Alle drei Modell-Einstellungen leer lassen, um die Standardmodelle deines Anbieters zu verwenden. Für einen lokalen Server den Modellnamen verwenden, z. B. llama3.1 oder qwen2.5-coder.",
"config.gitstudio.ai.openai.modelMid.description": "OpenAI-kompatible Modell-ID für mittelstufige Aufgaben (Diff erklären, Zusammenfassungen).",
"config.gitstudio.ai.provider.enumDescriptions.0": "Bevorzugt das schlüssellose VS Code Language Model (GitHub Copilot / Cursor), falls verfügbar; andernfalls Anthropic (falls ein Schlüssel gesetzt ist), danach ein OpenAI-kompatibler Endpunkt (falls ein Modell konfiguriert ist).",
"config.gitstudio.ai.provider.enumDescriptions.1": "Die VS Code Language Model API verwenden — die Modelle von GitHub Copilot oder Cursor, ohne Schlüssel. Ein bestimmtes Modell über „GitStudio: KI-Modell auswählen…“ wählen.",
Expand Down
2 changes: 1 addition & 1 deletion apps/extension/package.nls.es.json
Original file line number Diff line number Diff line change
Expand Up @@ -154,7 +154,7 @@
"config.gitstudio.ai.commitStyle.enumDescriptions.2": "Asunto en modo imperativo más un breve cuerpo explicativo.",
"config.gitstudio.ai.openai.baseUrl.markdownDescription": "URL base para el endpoint compatible con OpenAI (se añade `/chat/completions`). Ejemplos: `https://api.openai.com/v1` (OpenAI), `https://openrouter.ai/api/v1` (OpenRouter), `http://localhost:11434/v1` (Ollama), `http://localhost:1234/v1` (LM Studio).",
"config.gitstudio.ai.openai.modelDeep.description": "ID del modelo compatible con OpenAI para razonamiento profundo opcional.",
"config.gitstudio.ai.openai.modelFast.description": "ID del modelo compatible con OpenAI para tareas rápidas/económicas (mensajes de commit). Para un servidor local usa su nombre de modelo, p. ej. llama3.1 o qwen2.5-coder.",
"config.gitstudio.ai.openai.modelFast.description": "ID del modelo compatible con OpenAI para tareas rápidas/económicas (mensajes de commit). Deja vacíos los tres ajustes de modelo para usar los modelos predeterminados de tu proveedor. Para un servidor local usa su nombre de modelo, p. ej. llama3.1 o qwen2.5-coder.",
"config.gitstudio.ai.openai.modelMid.description": "ID del modelo compatible con OpenAI para tareas de nivel medio (explicar diff, resúmenes).",
"config.gitstudio.ai.provider.enumDescriptions.0": "Prefiere el Modelo de lenguaje de VS Code sin clave (GitHub Copilot / Cursor) cuando esté disponible; si no, Anthropic (si hay una clave configurada), y luego un endpoint compatible con OpenAI (si hay un modelo configurado).",
"config.gitstudio.ai.provider.enumDescriptions.1": "Usa la API del Modelo de lenguaje de VS Code — los modelos de GitHub Copilot o Cursor, sin clave. Elige un modelo específico con “GitStudio: Select AI Model…”.",
Expand Down
2 changes: 1 addition & 1 deletion apps/extension/package.nls.fr.json
Original file line number Diff line number Diff line change
Expand Up @@ -154,7 +154,7 @@
"config.gitstudio.ai.commitStyle.enumDescriptions.2": "Sujet à l'impératif suivi d'un court corps explicatif.",
"config.gitstudio.ai.openai.baseUrl.markdownDescription": "URL de base pour le point de terminaison compatible OpenAI (`/chat/completions` est ajouté). Exemples : `https://api.openai.com/v1` (OpenAI), `https://openrouter.ai/api/v1` (OpenRouter), `http://localhost:11434/v1` (Ollama), `http://localhost:1234/v1` (LM Studio).",
"config.gitstudio.ai.openai.modelDeep.description": "ID du modèle compatible OpenAI pour le raisonnement approfondi optionnel.",
"config.gitstudio.ai.openai.modelFast.description": "ID du modèle compatible OpenAI pour les tâches rapides et peu coûteuses (messages de commit). Pour un serveur local, utilisez son nom de modèle, par ex. llama3.1 ou qwen2.5-coder.",
"config.gitstudio.ai.openai.modelFast.description": "ID du modèle compatible OpenAI pour les tâches rapides et peu coûteuses (messages de commit). Laissez les trois paramètres de modèle vides pour utiliser les modèles par défaut de votre fournisseur. Pour un serveur local, utilisez son nom de modèle, par ex. llama3.1 ou qwen2.5-coder.",
"config.gitstudio.ai.openai.modelMid.description": "ID du modèle compatible OpenAI pour les tâches intermédiaires (explication de diff, résumés).",
"config.gitstudio.ai.provider.enumDescriptions.0": "Préférer le modèle de langage intégré à VS Code sans clé (GitHub Copilot / Cursor) s'il est disponible ; sinon Anthropic (si une clé est définie), puis un point de terminaison compatible OpenAI (si un modèle est configuré).",
"config.gitstudio.ai.provider.enumDescriptions.1": "Utiliser l'API Language Model de VS Code — les modèles de GitHub Copilot ou de Cursor, sans clé. Choisissez un modèle précis via « GitStudio : Sélectionner un modèle d'IA… ».",
Expand Down
2 changes: 1 addition & 1 deletion apps/extension/package.nls.it.json
Original file line number Diff line number Diff line change
Expand Up @@ -154,7 +154,7 @@
"config.gitstudio.ai.commitStyle.enumDescriptions.2": "Oggetto imperativo più un breve corpo esplicativo.",
"config.gitstudio.ai.openai.baseUrl.markdownDescription": "URL di base per l'endpoint compatibile OpenAI (viene aggiunto `/chat/completions`). Esempi: `https://api.openai.com/v1` (OpenAI), `https://openrouter.ai/api/v1` (OpenRouter), `http://localhost:11434/v1` (Ollama), `http://localhost:1234/v1` (LM Studio).",
"config.gitstudio.ai.openai.modelDeep.description": "ID del modello compatibile OpenAI per il ragionamento approfondito opzionale.",
"config.gitstudio.ai.openai.modelFast.description": "ID del modello compatibile OpenAI per attività rapide/economiche (messaggi di commit). Per un server locale usa il nome del suo modello, ad es. llama3.1 o qwen2.5-coder.",
"config.gitstudio.ai.openai.modelFast.description": "ID del modello compatibile OpenAI per attività rapide/economiche (messaggi di commit). Lascia vuote tutte e tre le impostazioni del modello per usare i modelli predefiniti del tuo provider. Per un server locale usa il nome del suo modello, ad es. llama3.1 o qwen2.5-coder.",
"config.gitstudio.ai.openai.modelMid.description": "ID del modello compatibile OpenAI per attività di livello medio (spiegazione diff, riepiloghi).",
"config.gitstudio.ai.provider.enumDescriptions.0": "Preferisci il Language Model di VS Code senza chiave (GitHub Copilot / Cursor) quando disponibile; altrimenti Anthropic (se è impostata una chiave), poi un endpoint compatibile OpenAI (se è configurato un modello).",
"config.gitstudio.ai.provider.enumDescriptions.1": "Usa l'API Language Model di VS Code — i modelli di GitHub Copilot o Cursor, senza chiave. Scegli un modello specifico tramite “GitStudio: Select AI Model…”.",
Expand Down
2 changes: 1 addition & 1 deletion apps/extension/package.nls.ja.json
Original file line number Diff line number Diff line change
Expand Up @@ -154,7 +154,7 @@
"config.gitstudio.ai.commitStyle.enumDescriptions.2": "命令形の件名と短い説明の本文。",
"config.gitstudio.ai.openai.baseUrl.markdownDescription": "OpenAI互換エンドポイントのベースURL(`/chat/completions` が付加されます)。例: `https://api.openai.com/v1` (OpenAI)、`https://openrouter.ai/api/v1` (OpenRouter)、`http://localhost:11434/v1` (Ollama)、`http://localhost:1234/v1` (LM Studio)。",
"config.gitstudio.ai.openai.modelDeep.description": "オプトインの深い推論用のOpenAI互換モデルID。",
"config.gitstudio.ai.openai.modelFast.description": "高速・低コストなタスク(コミットメッセージ)用のOpenAI互換モデルID。ローカルサーバーの場合は、そのモデル名(例: llama3.1 や qwen2.5-coder)を使用してください。",
"config.gitstudio.ai.openai.modelFast.description": "高速・低コストなタスク(コミットメッセージ)用のOpenAI互換モデルID。3つのモデル設定をすべて空のままにすると、プロバイダーの既定のモデルが使われます。ローカルサーバーの場合は、そのモデル名(例: llama3.1 や qwen2.5-coder)を使用してください。",
"config.gitstudio.ai.openai.modelMid.description": "中位タスク(差分の説明、要約)用のOpenAI互換モデルID。",
"config.gitstudio.ai.provider.enumDescriptions.0": "利用可能な場合はキー不要のVS Code言語モデル(GitHub Copilot / Cursor)を優先し、なければAnthropic(キーが設定されている場合)、次にOpenAI互換エンドポイント(モデルが設定されている場合)を使用します。",
"config.gitstudio.ai.provider.enumDescriptions.1": "VS Code言語モデルAPIを使用します — GitHub CopilotまたはCursorのモデルで、キーは不要です。「GitStudio: AIモデルを選択…」から特定のモデルを選べます。",
Expand Down
2 changes: 1 addition & 1 deletion apps/extension/package.nls.json
Original file line number Diff line number Diff line change
Expand Up @@ -154,7 +154,7 @@
"config.gitstudio.ai.commitStyle.enumDescriptions.2": "Imperative subject plus a short explanatory body.",
"config.gitstudio.ai.openai.baseUrl.markdownDescription": "Base URL for the OpenAI-compatible endpoint (`/chat/completions` is appended). Examples: `https://api.openai.com/v1` (OpenAI), `https://openrouter.ai/api/v1` (OpenRouter), `http://localhost:11434/v1` (Ollama), `http://localhost:1234/v1` (LM Studio).",
"config.gitstudio.ai.openai.modelDeep.description": "OpenAI-compatible model ID for opt-in deep reasoning.",
"config.gitstudio.ai.openai.modelFast.description": "OpenAI-compatible model ID for fast/cheap tasks (commit messages). For a local server use its model name, e.g. llama3.1 or qwen2.5-coder.",
"config.gitstudio.ai.openai.modelFast.description": "OpenAI-compatible model ID for fast/cheap tasks (commit messages). Leave all three model settings empty to use your provider's default models. For a local server use its model name, e.g. llama3.1 or qwen2.5-coder.",
"config.gitstudio.ai.openai.modelMid.description": "OpenAI-compatible model ID for mid-tier tasks (explain diff, summaries).",
"config.gitstudio.ai.provider.enumDescriptions.0": "Prefer the zero-key VS Code Language Model (GitHub Copilot / Cursor) when available; otherwise Anthropic (if a key is set), then an OpenAI-compatible endpoint (if a model is configured).",
"config.gitstudio.ai.provider.enumDescriptions.1": "Use the VS Code Language Model API — GitHub Copilot or Cursor's models, with no key. Pick a specific model via “GitStudio: Select AI Model…”.",
Expand Down
Loading
Loading