2026-07-02 20:15:00 +00:00
13 changed files with 924 additions and 80 deletions
@@ -48,7 +48,17 @@
      "Bash(curl -sI \"https://huggingface.co/Qwen/Qwen2.5-1.5B-Instruct-GGUF/resolve/main/qwen2.5-1.5b-instruct-q4_k_m.gguf\")",
      "Bash(curl -s \"https://huggingface.co/api/models/Qwen/Qwen2.5-1.5B-Instruct-GGUF?blobs=true\")",
      "Bash(grep -n 'from \"\\\\./lib/git\"\\\\|from \"\\\\./lib/types\"\\\\|^    commit,$' src/App.svelte)",
-      "Bash(kill 24343 24363 24375 24376)"
+      "Bash(kill 24343 24363 24375 24376)",
+      "Bash(curl -s \"https://huggingface.co/api/models/Qwen/Qwen2.5-0.5B-Instruct-GGUF?blobs=true\")",
+      "Bash(curl -s \"https://huggingface.co/api/models/bartowski/Llama-3.2-3B-Instruct-GGUF?blobs=true\")",
+      "Bash(curl -s \"https://huggingface.co/api/models/bartowski/Llama-3.2-3B-Instruct-GGUF\")",
+      "Bash(curl -s \"https://huggingface.co/api/models/Qwen/Qwen2.5-3B-Instruct-GGUF\")",
+      "Bash(curl -s \"https://huggingface.co/api/models/Qwen/Qwen2.5-3B-Instruct-GGUF?blobs=true\")",
+      "Bash(curl -s \"https://huggingface.co/api/models/Qwen/Qwen2.5-3B-Instruct\")",
+      "Bash(: *)",
+      "Bash(exit 0 *)",
+      "Bash(rustc -O sanitize_test.rs -o sanitize_test)",
+      "Bash(./sanitize_test)"
    ]
  }
 }
@@ -1028,7 +1028,9 @@ name = "commit_ai"
 version = "0.1.0"
 dependencies = [
 "mistralrs",
+ "reqwest 0.12.28",
 "serde",
+ "serde_json",
 "tokio",
 ]

@@ -8,3 +8,5 @@ edition = "2024"
 mistralrs = "0.8"
 tokio = { version = "1", features = ["sync"] }
 serde = { version = "1", features = ["derive"] }
+serde_json = "1"
+reqwest = { version = "0.12", default-features = false, features = ["json", "rustls-tls"] }
@@ -0,0 +1,203 @@
+use std::time::Duration;
+
+use serde::{Deserialize, Serialize};
+
+use crate::{build_messages, sanitize_message};
+
+// Großzügig bemessen, damit ein ausführlicher Body mit Stichpunkten nicht abgeschnitten wird.
+const DEFAULT_MAX_TOKENS: u32 = 1500;
+// Ohne Timeout würde ein hängender Endpoint den "AI"-Button dauerhaft blockieren.
+const REQUEST_TIMEOUT: Duration = Duration::from_secs(60);
+
+fn http_client() -> Result<reqwest::Client, String> {
+    reqwest::Client::builder()
+        .timeout(REQUEST_TIMEOUT)
+        .build()
+        .map_err(|err| format!("HTTP-Client konnte nicht erstellt werden: {err}"))
+}
+
+#[derive(Serialize)]
+struct OpenAiMessage {
+    role: &'static str,
+    content: String,
+}
+
+#[derive(Serialize)]
+struct OpenAiRequest {
+    model: String,
+    messages: Vec<OpenAiMessage>,
+    temperature: f32,
+}
+
+#[derive(Deserialize)]
+struct OpenAiResponseMessage {
+    content: Option<String>,
+}
+
+#[derive(Deserialize)]
+struct OpenAiChoice {
+    message: OpenAiResponseMessage,
+}
+
+#[derive(Deserialize)]
+struct OpenAiResponse {
+    #[serde(default)]
+    choices: Vec<OpenAiChoice>,
+}
+
+async fn openai_compatible_request(
+    url: String,
+    bearer: Option<&str>,
+    model: &str,
+    diff: &str,
+    notes: Option<&str>,
+) -> Result<String, String> {
+    let (system, user) = build_messages(diff, notes)?;
+    let body = OpenAiRequest {
+        model: model.to_string(),
+        messages: vec![
+            OpenAiMessage { role: "system", content: system },
+            OpenAiMessage { role: "user", content: user },
+        ],
+        temperature: 0.3,
+    };
+
+    let client = http_client()?;
+    let mut request = client.post(url).json(&body);
+    if let Some(key) = bearer.filter(|k| !k.trim().is_empty()) {
+        request = request.bearer_auth(key);
+    }
+
+    let response = request
+        .send()
+        .await
+        .map_err(|err| format!("Anfrage an das KI-Modell fehlgeschlagen: {err}"))?;
+    let status = response.status();
+    let text = response
+        .text()
+        .await
+        .map_err(|err| format!("Antwort konnte nicht gelesen werden: {err}"))?;
+
+    if !status.is_success() {
+        return Err(format!("API-Fehler ({status}): {text}"));
+    }
+
+    let parsed: OpenAiResponse = serde_json::from_str(&text)
+        .map_err(|err| format!("Antwort konnte nicht verarbeitet werden: {err}"))?;
+
+    parsed
+        .choices
+        .into_iter()
+        .next()
+        .and_then(|choice| choice.message.content)
+        .map(|content| sanitize_message(&content))
+        .filter(|content| !content.is_empty())
+        .ok_or_else(|| "Das Modell hat keine Antwort geliefert.".to_string())
+}
+
+pub async fn generate_openai(
+    api_key: &str,
+    model: &str,
+    diff: &str,
+    notes: Option<&str>,
+) -> Result<String, String> {
+    if api_key.trim().is_empty() {
+        return Err("OpenAI-API-Key fehlt.".to_string());
+    }
+    openai_compatible_request(
+        "https://api.openai.com/v1/chat/completions".to_string(),
+        Some(api_key),
+        model,
+        diff,
+        notes,
+    )
+    .await
+}
+
+pub async fn generate_custom(
+    base_url: &str,
+    api_key: Option<&str>,
+    model: &str,
+    diff: &str,
+    notes: Option<&str>,
+) -> Result<String, String> {
+    if base_url.trim().is_empty() {
+        return Err("Endpoint-URL fehlt.".to_string());
+    }
+    let url = format!("{}/chat/completions", base_url.trim_end_matches('/'));
+    openai_compatible_request(url, api_key, model, diff, notes).await
+}
+
+#[derive(Serialize)]
+struct AnthropicMessage {
+    role: &'static str,
+    content: String,
+}
+
+#[derive(Serialize)]
+struct AnthropicRequest {
+    model: String,
+    max_tokens: u32,
+    system: String,
+    messages: Vec<AnthropicMessage>,
+}
+
+#[derive(Deserialize)]
+struct AnthropicContentBlock {
+    #[serde(default)]
+    text: Option<String>,
+}
+
+#[derive(Deserialize)]
+struct AnthropicResponse {
+    #[serde(default)]
+    content: Vec<AnthropicContentBlock>,
+}
+
+pub async fn generate_anthropic(
+    api_key: &str,
+    model: &str,
+    diff: &str,
+    notes: Option<&str>,
+) -> Result<String, String> {
+    if api_key.trim().is_empty() {
+        return Err("Anthropic-API-Key fehlt.".to_string());
+    }
+    let (system, user) = build_messages(diff, notes)?;
+    let body = AnthropicRequest {
+        model: model.to_string(),
+        max_tokens: DEFAULT_MAX_TOKENS,
+        system,
+        messages: vec![AnthropicMessage { role: "user", content: user }],
+    };
+
+    let client = http_client()?;
+    let response = client
+        .post("https://api.anthropic.com/v1/messages")
+        .header("x-api-key", api_key)
+        .header("anthropic-version", "2023-06-01")
+        .json(&body)
+        .send()
+        .await
+        .map_err(|err| format!("Anfrage an Anthropic fehlgeschlagen: {err}"))?;
+    let status = response.status();
+    let text = response
+        .text()
+        .await
+        .map_err(|err| format!("Antwort konnte nicht gelesen werden: {err}"))?;
+
+    if !status.is_success() {
+        return Err(format!("API-Fehler ({status}): {text}"));
+    }
+
+    let parsed: AnthropicResponse = serde_json::from_str(&text)
+        .map_err(|err| format!("Antwort konnte nicht verarbeitet werden: {err}"))?;
+
+    parsed
+        .content
+        .into_iter()
+        .find_map(|block| block.text)
+        .map(|text| sanitize_message(&text))
+        .filter(|text| !text.is_empty())
+        .ok_or_else(|| "Das Modell hat keine Antwort geliefert.".to_string())
+}
@@ -1,13 +1,56 @@
+mod cloud;
+
+pub use cloud::{generate_anthropic, generate_custom, generate_openai};
+
 use std::sync::Arc;

 use mistralrs::{GgufModelBuilder, Model, TextMessageRole, TextMessages};
 use tokio::sync::RwLock;

-// Small instruct model, quantized to keep the one-time download reasonable (~1 GB) while
-// still being fast enough for short, structured generations like a commit message on CPU.
-const HF_REPO: &str = "Qwen/Qwen2.5-1.5B-Instruct-GGUF";
-const GGUF_FILE: &str = "qwen2.5-1.5b-instruct-q4_k_m.gguf";
-const TOKENIZER_MODEL_ID: &str = "Qwen/Qwen2.5-1.5B-Instruct";
+/// One selectable local (on-device) model. Larger models produce better commit messages
+/// but take longer to download (first run only, then cached) and run slower on CPU.
+#[derive(Debug, Clone, serde::Serialize)]
+pub struct LocalModelOption {
+    pub id: &'static str,
+    pub label: &'static str,
+    pub approx_size_mb: u32,
+    repo: &'static str,
+    file: &'static str,
+    tokenizer_repo: &'static str,
+}
+
+pub const DEFAULT_LOCAL_MODEL_ID: &str = "qwen2.5-1.5b";
+
+pub const LOCAL_MODELS: &[LocalModelOption] = &[
+    LocalModelOption {
+        id: "qwen2.5-0.5b",
+        label: "Qwen2.5 0.5B Instruct — schnell, geringere Qualität",
+        approx_size_mb: 490,
+        repo: "Qwen/Qwen2.5-0.5B-Instruct-GGUF",
+        file: "qwen2.5-0.5b-instruct-q4_k_m.gguf",
+        tokenizer_repo: "Qwen/Qwen2.5-0.5B-Instruct",
+    },
+    LocalModelOption {
+        id: "qwen2.5-1.5b",
+        label: "Qwen2.5 1.5B Instruct — empfohlen",
+        approx_size_mb: 1050,
+        repo: "Qwen/Qwen2.5-1.5B-Instruct-GGUF",
+        file: "qwen2.5-1.5b-instruct-q4_k_m.gguf",
+        tokenizer_repo: "Qwen/Qwen2.5-1.5B-Instruct",
+    },
+    LocalModelOption {
+        id: "qwen2.5-3b",
+        label: "Qwen2.5 3B Instruct — beste Qualität, langsamer",
+        approx_size_mb: 2100,
+        repo: "Qwen/Qwen2.5-3B-Instruct-GGUF",
+        file: "qwen2.5-3b-instruct-q4_k_m.gguf",
+        tokenizer_repo: "Qwen/Qwen2.5-3B-Instruct",
+    },
+];
+
+fn find_local_model(model_id: &str) -> Option<&'static LocalModelOption> {
+    LOCAL_MODELS.iter().find(|option| option.id == model_id)
+}

 #[derive(Debug, Clone, Copy, PartialEq, Eq, serde::Serialize)]
 #[serde(rename_all = "lowercase")]
@@ -23,16 +66,19 @@ pub enum CommitAiPhase {
 #[derive(Debug, Clone, serde::Serialize)]
 pub struct CommitAiStatus {
    pub phase: CommitAiPhase,
+    pub model_id: Option<String>,
    pub error: Option<String>,
 }

 struct Inner {
    phase: CommitAiPhase,
+    model_id: Option<String>,
    error: Option<String>,
    model: Option<Arc<Model>>,
 }

-/// Cheap to clone: shares one model instance across the app via an inner `Arc`.
+/// Manages the local (on-device) model only. Cloud providers are stateless HTTP calls
+/// (see [`cloud`]) and don't need this — there's nothing to download or keep loaded.
 #[derive(Clone)]
 pub struct CommitAiEngine {
    inner: Arc<RwLock<Inner>>,
@@ -43,6 +89,7 @@ impl Default for CommitAiEngine {
        Self {
            inner: Arc::new(RwLock::new(Inner {
                phase: CommitAiPhase::Idle,
+                model_id: None,
                error: None,
                model: None,
            })),
@@ -59,30 +106,52 @@ impl CommitAiEngine {
        let guard = self.inner.read().await;
        CommitAiStatus {
            phase: guard.phase,
+            model_id: guard.model_id.clone(),
            error: guard.error.clone(),
        }
    }

-    /// Downloads (first run only; hf-hub caches the files afterwards) and loads the model.
-    /// Safe to call multiple times — only the first caller actually triggers a load, later
-    /// callers just return once the in-flight or previous attempt is done.
-    pub async fn ensure_loaded(&self) {
+    /// Downloads (first run only; hf-hub caches the files afterwards) and loads the given
+    /// local model. Safe to call repeatedly — a call for the model that's already
+    /// ready/loading is a no-op; a call for a *different* model switches to it (the
+    /// previous one is dropped once no generation is still using it).
+    pub async fn ensure_loaded(&self, model_id: &str) {
        {
-            let mut guard = self.inner.write().await;
-            if guard.phase != CommitAiPhase::Idle {
+            let guard = self.inner.read().await;
+            let same_model = guard.model_id.as_deref() == Some(model_id);
+            if same_model && matches!(guard.phase, CommitAiPhase::Ready | CommitAiPhase::Loading) {
                return;
            }
-            guard.phase = CommitAiPhase::Loading;
-            guard.error = None;
        }

-        let result = GgufModelBuilder::new(HF_REPO, vec![GGUF_FILE])
-            .with_tok_model_id(TOKENIZER_MODEL_ID)
+        let Some(option) = find_local_model(model_id) else {
+            let mut guard = self.inner.write().await;
+            guard.phase = CommitAiPhase::Error;
+            guard.model_id = Some(model_id.to_string());
+            guard.error = Some(format!("Unbekanntes lokales Modell: {model_id}"));
+            return;
+        };
+
+        {
+            let mut guard = self.inner.write().await;
+            guard.phase = CommitAiPhase::Loading;
+            guard.model_id = Some(model_id.to_string());
+            guard.error = None;
+            guard.model = None;
+        }
+
+        let result = GgufModelBuilder::new(option.repo, vec![option.file])
+            .with_tok_model_id(option.tokenizer_repo)
            .with_logging()
            .build()
            .await;

        let mut guard = self.inner.write().await;
+        // If the user switched to yet another model while this one was loading, drop this
+        // (now stale) result instead of overwriting the newer request's state.
+        if guard.model_id.as_deref() != Some(model_id) {
+            return;
+        }
        match result {
            Ok(model) => {
                guard.model = Some(Arc::new(model));
@@ -105,17 +174,11 @@ impl CommitAiEngine {
            let guard = self.inner.read().await;
            match (guard.phase, &guard.model) {
                (CommitAiPhase::Ready, Some(model)) => model.clone(),
-                _ => return Err("Das KI-Modell ist noch nicht bereit.".to_string()),
+                _ => return Err("Das lokale KI-Modell ist noch nicht bereit.".to_string()),
            }
        };

-        if diff.trim().is_empty() {
-            return Err(
-                "Keine gestagten Änderungen für eine Commit-Message vorhanden.".to_string(),
-            );
-        }
-
-        let (system, user) = build_messages(diff, notes);
+        let (system, user) = build_messages(diff, notes)?;
        let messages = TextMessages::new()
            .add_message(TextMessageRole::System, system)
            .add_message(TextMessageRole::User, user);
@@ -131,23 +194,60 @@ impl CommitAiEngine {
            .and_then(|choice| choice.message.content.clone())
            .ok_or_else(|| "Das Modell hat keine Antwort geliefert.".to_string())?;

-        Ok(content.trim().to_string())
+        let message = sanitize_message(&content);
+        if message.is_empty() {
+            return Err("Das Modell hat keine Antwort geliefert.".to_string());
+        }
+        Ok(message)
    }
 }

-fn build_messages(diff: &str, notes: Option<&str>) -> (String, String) {
+/// Models occasionally ignore the "no code fences" instruction (small local models
+/// especially) — strip a wrapping ``` fence and wrapping quotes so the result can go
+/// straight into the commit-message box.
+pub(crate) fn sanitize_message(raw: &str) -> String {
+    let mut text = raw.trim().to_string();
+    if text.starts_with("```") {
+        text = match text.split_once('\n') {
+            // Drop the opening fence line (which may carry a language tag) and the closing fence.
+            Some((_fence, rest)) => rest.trim_end().trim_end_matches("```").trim().to_string(),
+            None => text.trim_matches('`').trim().to_string(),
+        };
+    }
+    let trimmed = text.trim();
+    if trimmed.len() >= 2 && trimmed.starts_with('"') && trimmed.ends_with('"') {
+        return trimmed[1..trimmed.len() - 1].trim().to_string();
+    }
+    trimmed.to_string()
+}
+
+pub(crate) fn build_messages(diff: &str, notes: Option<&str>) -> Result<(String, String), String> {
+    if diff.trim().is_empty() {
+        return Err("Keine gestagten Änderungen für eine Commit-Message vorhanden.".to_string());
+    }
+
    // grobe Token-Schätzung, kleine Modelle haben oft 8–32k Kontext
    const MAX_CHARS: usize = 24_000;
    let diff = if diff.len() > MAX_CHARS {
-        format!("{}\n\n[... Diff gekürzt ...]", &diff[..MAX_CHARS])
+        // Byte-Index auf eine gültige UTF-8-Zeichengrenze zurückziehen, sonst
+        // paniken Slices mitten in einem Umlaut o. Ä.
+        let mut cut = MAX_CHARS;
+        while !diff.is_char_boundary(cut) {
+            cut -= 1;
+        }
+        format!("{}\n\n[... Diff gekürzt ...]", &diff[..cut])
    } else {
        diff.to_string()
    };

    let system = "Du bist ein Werkzeug, das Git-Commit-Messages erzeugt. \
 Antworte ausschließlich mit der Commit-Message im Conventional-Commits-Format \
-(<type>(<scope>): <subject>), optional gefolgt von einem Body nach einer Leerzeile. \
-Subject imperativ, max. 72 Zeichen. Kein Vorspann, keine Erklärung, keine Code-Fences, in Englisch antworten"
+(<type>(<scope>): <subject>), gefolgt von einem Body nach einer Leerzeile. \
+Subject imperativ, max. 72 Zeichen. \
+Der Body ist Pflicht: Fasse in einem kurzen Absatz zusammen, was und warum geändert wurde, \
+und liste danach die wesentlichen Änderungen als Stichpunkte (- ...) auf, \
+gruppiert nach betroffenem Bereich/Datei. Zeilen im Body max. 72 Zeichen. \
+Kein Vorspann, keine Erklärung, keine Code-Fences, in Englisch antworten"
        .to_string();

    let mut user = String::new();
@@ -155,5 +255,5 @@ Subject imperativ, max. 72 Zeichen. Kein Vorspann, keine Erklärung, keine Code-
        user.push_str(&format!("Anmerkungen des Entwicklers:\n{n}\n\n"));
    }
    user.push_str(&format!("Staged diff:\n{diff}"));
-    (system, user)
+    Ok((system, user))
 }
@@ -479,6 +479,11 @@ pub fn get_file_patch(path: String, file: String, staged: bool) -> Result<String
    Ok(String::from_utf8_lossy(&output).to_string())
 }

+#[tauri::command]
+pub fn commit_ai_local_models() -> Vec<commit_ai::LocalModelOption> {
+    commit_ai::LOCAL_MODELS.to_vec()
+}
+
 #[tauri::command]
 pub async fn commit_ai_status(
    engine: tauri::State<'_, commit_ai::CommitAiEngine>,
@@ -486,28 +491,91 @@ pub async fn commit_ai_status(
    Ok(engine.status().await)
 }

+/// Kicks off the (first-run-only) download and model load in the background and returns
+/// immediately; the frontend polls `commit_ai_status` to know when it's ready.
 #[tauri::command]
-pub async fn commit_ai_generate(
-    path: String,
-    notes: Option<String>,
-    engine: tauri::State<'_, commit_ai::CommitAiEngine>,
-) -> Result<String, String> {
-    let repo = resolve_repo(&path)?;
+pub fn commit_ai_load(model_id: String, engine: tauri::State<'_, commit_ai::CommitAiEngine>) {
+    let engine = engine.inner().clone();
+    tauri::async_runtime::spawn(async move {
+        engine.ensure_loaded(&model_id).await;
+    });
+}
+
+// The prompt is built from `git diff --cached` only, i.e. exactly the staged changes —
+// unstaged edits and untracked files never influence the generated message.
+fn staged_diff(repo: &Path) -> Result<String, String> {
+    // Full staged file list (nothing excluded) so the model knows the complete scope
+    // even when the detailed diff below is filtered or truncated for context size.
+    let name_status = run_git(repo, ["diff", "--cached", "--name-status", "-M"])?;
+    let file_list = String::from_utf8_lossy(&name_status).trim().to_string();
+
+    // Generated lockfiles say nothing useful about intent but easily blow the small
+    // context window of local models, so keep them out of the detailed diff.
    let diff = run_git(
-        &repo,
+        repo,
        [
            "diff",
            "--cached",
            "--no-ext-diff",
            "--no-textconv",
            "--unified=3",
+            "--",
+            ".",
+            ":(exclude)*package-lock.json",
+            ":(exclude)*pnpm-lock.yaml",
+            ":(exclude)*yarn.lock",
+            ":(exclude)*bun.lockb",
+            ":(exclude)*Cargo.lock",
+            ":(exclude)*composer.lock",
+            ":(exclude)*Gemfile.lock",
+            ":(exclude)*poetry.lock",
+            ":(exclude)*go.sum",
        ],
    )?;
-    let diff = String::from_utf8_lossy(&diff).to_string();
+    let diff = String::from_utf8_lossy(&diff);

-    engine
-        .generate_commit_message(&diff, notes.as_deref())
-        .await
+    if file_list.is_empty() {
+        return Ok(diff.to_string());
+    }
+    Ok(format!("Staged files:\n{file_list}\n\n{diff}"))
+}
+
+#[tauri::command]
+pub async fn commit_ai_generate(
+    path: String,
+    notes: Option<String>,
+    provider: String,
+    model: Option<String>,
+    api_key: Option<String>,
+    base_url: Option<String>,
+    engine: tauri::State<'_, commit_ai::CommitAiEngine>,
+) -> Result<String, String> {
+    let repo = resolve_repo(&path)?;
+    let diff = staged_diff(&repo)?;
+    let notes = notes.as_deref();
+    let model = model.filter(|value| !value.trim().is_empty());
+    let api_key = api_key.filter(|value| !value.trim().is_empty());
+    let base_url = base_url.filter(|value| !value.trim().is_empty());
+
+    match provider.as_str() {
+        "local" => engine.generate_commit_message(&diff, notes).await,
+        "openai" => {
+            let api_key = api_key.ok_or_else(|| "OpenAI-API-Key fehlt.".to_string())?;
+            let model = model.unwrap_or_else(|| "gpt-4o-mini".to_string());
+            commit_ai::generate_openai(&api_key, &model, &diff, notes).await
+        }
+        "anthropic" => {
+            let api_key = api_key.ok_or_else(|| "Anthropic-API-Key fehlt.".to_string())?;
+            let model = model.unwrap_or_else(|| "claude-3-5-haiku-latest".to_string());
+            commit_ai::generate_anthropic(&api_key, &model, &diff, notes).await
+        }
+        "custom" => {
+            let base_url = base_url.ok_or_else(|| "Endpoint-URL fehlt.".to_string())?;
+            let model = model.ok_or_else(|| "Modellname fehlt.".to_string())?;
+            commit_ai::generate_custom(&base_url, api_key.as_deref(), &model, &diff, notes).await
+        }
+        other => Err(format!("Unbekannter KI-Provider: {other}")),
+    }
 }

 #[tauri::command]
@@ -4,33 +4,22 @@ mod git;

 use git::{
    SearchCancellationState, apply_file_patch, cancel_code_search, cancel_file_history,
-    checkout_branch, commit, commit_ai_generate, commit_ai_status, compare_commits,
-    compare_file_to_head, compare_file_to_parent, create_branch, cred_delete, cred_load, cred_save,
-    delete_branch, diff_file_against_working_tree, get_file_patch, get_remote_url, get_status,
-    list_branches, list_commits, list_file_history, list_repository_files, merge_branch,
-    open_repo_in_explorer, open_repository, open_repository_bundle, open_repository_file, pull,
-    push, read_conflict, rename_branch, resolve_conflict, resolve_conflict_side,
-    restore_file_from_commit, restore_files, restore_to_commit, search_code_introductions,
-    stage_files, unstage_files,
+    checkout_branch, commit, commit_ai_generate, commit_ai_load, commit_ai_local_models,
+    commit_ai_status, compare_commits, compare_file_to_head, compare_file_to_parent,
+    create_branch, cred_delete, cred_load, cred_save, delete_branch,
+    diff_file_against_working_tree, get_file_patch, get_remote_url, get_status, list_branches,
+    list_commits, list_file_history, list_repository_files, merge_branch, open_repo_in_explorer,
+    open_repository, open_repository_bundle, open_repository_file, pull, push, read_conflict,
+    rename_branch, resolve_conflict, resolve_conflict_side, restore_file_from_commit,
+    restore_files, restore_to_commit, search_code_introductions, stage_files, unstage_files,
 };

 fn main() {
-    let commit_ai_engine = commit_ai::CommitAiEngine::new();
-
    tauri::Builder::default()
        .plugin(tauri_plugin_updater::Builder::new().build())
        .manage(SearchCancellationState::default())
-        .manage(commit_ai_engine.clone())
+        .manage(commit_ai::CommitAiEngine::new())
        .plugin(tauri_plugin_dialog::init())
-        .setup(move |_app| {
-            // Kick off the (first-run-only) download and model load in the background so the
-            // "Generate with AI" button becomes enabled once it's ready, without blocking startup.
-            let engine = commit_ai_engine.clone();
-            tauri::async_runtime::spawn(async move {
-                engine.ensure_loaded().await;
-            });
-            Ok(())
-        })
        .invoke_handler(tauri::generate_handler![
            open_repository,
            open_repo_in_explorer,
@@ -48,6 +37,8 @@ fn main() {
            apply_file_patch,
            commit,
            commit_ai_status,
+            commit_ai_load,
+            commit_ai_local_models,
            commit_ai_generate,
            pull,
            push,
@@ -5,6 +5,7 @@
  import { AlertCircle, BookOpen, FolderOpen, GitBranch, GitMerge, LoaderCircle, Plus, Search, X } from "@lucide/svelte";

  import TitleBar from "./lib/TitleBar.svelte";
+  import AiSettingsDialog from "./lib/components/AiSettingsDialog.svelte";
  import BranchPanel from "./lib/components/BranchPanel.svelte";
  import CommitPanel from "./lib/components/CommitPanel.svelte";
  import CompareDialog from "./lib/components/CompareDialog.svelte";
@@ -26,6 +27,8 @@
    checkoutBranch,
    commit,
    commitAiGenerate,
+    commitAiLoad,
+    commitAiLocalModels,
    commitAiStatus,
    compareCommits,
    cancelCodeSearch,
@@ -64,6 +67,7 @@
  } from "./lib/git";

  import type {
+    AiSettings,
    CommitAiPhase,
    ConflictFile,
    ExplorerNode,
@@ -77,6 +81,7 @@
    GitRepositoryFile,
    GitSearchHit,
    GitStatus,
+    LocalModelOption,
    PatchApplyAction,
    PreparedResolution,
    StoredCredential,
@@ -104,6 +109,7 @@

  const OPEN_REPOS_KEY = "gitlite.openRepos.v1";
  const RECENT_REPOS_KEY = "gitlite.recentRepos.v1";
+  const AI_SETTINGS_KEY = "gitlite.aiSettings.v1";

  // ── State ──────────────────────────────────────────────────────────────────

@@ -130,6 +136,9 @@
  let commitAiPhase: CommitAiPhase = "idle";
  let commitAiGenerating = false;
  let commitAiPollTimer: ReturnType<typeof setInterval> | undefined;
+  let aiSettings: AiSettings = defaultAiSettings();
+  let aiSettingsOpen = false;
+  let localModelOptions: LocalModelOption[] = [];
  let errorMessage = "";
  let operation = "";
  let compareFrom = "";
@@ -213,7 +222,7 @@
    loadRepoLists();
    autoRefreshTimer = setInterval(() => { void autoRefreshTick(); }, AUTO_REFRESH_INTERVAL);
    void checkForUpdates();
-    startCommitAiPolling();
+    void initCommitAi();
  });

  onDestroy(() => {
@@ -256,29 +265,85 @@

  // ── Commit AI ──────────────────────────────────────────────────────────────

+  function stopCommitAiPolling() {
+    if (commitAiPollTimer) { clearInterval(commitAiPollTimer); commitAiPollTimer = undefined; }
+  }
+
  async function pollCommitAiStatus() {
    try {
      const result = await commitAiStatus();
      commitAiPhase = result.phase;
    } catch { /* ignore transient errors */ }
-    if (commitAiPhase === "ready" || commitAiPhase === "error") {
-      if (commitAiPollTimer) { clearInterval(commitAiPollTimer); commitAiPollTimer = undefined; }
-    }
+    if (commitAiPhase === "ready" || commitAiPhase === "error") stopCommitAiPolling();
  }

  function startCommitAiPolling() {
-    // The model downloads (first run only) and loads in the background on app start;
-    // poll until it's ready (or failed) so the "AI" button can enable itself.
+    // Only the local model has a download/load phase worth polling — cloud providers are
+    // plain API calls with nothing to wait for.
+    stopCommitAiPolling();
+    if (aiSettings.provider !== "local") return;
    void pollCommitAiStatus();
    commitAiPollTimer = setInterval(() => { void pollCommitAiStatus(); }, 2000);
  }

+  async function initCommitAi() {
+    aiSettings = loadAiSettings();
+    try {
+      localModelOptions = await commitAiLocalModels();
+    } catch { /* AI features stay disabled if this fails; not fatal to the app */ }
+    if (aiSettings.provider === "local") {
+      try { await commitAiLoad(aiSettings.localModelId); } catch { /* surfaced via status polling */ }
+    }
+    startCommitAiPolling();
+  }
+
+  function saveAiSettings(next: AiSettings) {
+    const modelChanged = next.provider === "local" && next.localModelId !== aiSettings.localModelId;
+    aiSettings = next;
+    persistAiSettings(next);
+    aiSettingsOpen = false;
+    if (next.provider === "local" && (modelChanged || commitAiPhase === "idle")) {
+      commitAiPhase = "idle";
+      void commitAiLoad(next.localModelId);
+    }
+    startCommitAiPolling();
+  }
+
  async function generateCommitMessageWithAi() {
-    if (!activeRepoPath || commitAiPhase !== "ready" || commitAiGenerating) return;
+    if (!activeRepoPath || commitAiGenerating) return;
+    if (aiSettings.provider === "local" && commitAiPhase !== "ready") return;
    commitAiGenerating = true;
    errorMessage = "";
    try {
-      commitMessage = await commitAiGenerate(activeRepoPath, commitMessage.trim() || undefined);
+      const notes = commitMessage.trim() || undefined;
+      if (aiSettings.provider === "local") {
+        commitMessage = await commitAiGenerate(activeRepoPath, { provider: "local", notes });
+      } else if (aiSettings.provider === "openai") {
+        const cred = await credLoad("ai:openai");
+        commitMessage = await commitAiGenerate(activeRepoPath, {
+          provider: "openai",
+          notes,
+          model: aiSettings.openaiModel,
+          apiKey: cred?.password,
+        });
+      } else if (aiSettings.provider === "anthropic") {
+        const cred = await credLoad("ai:anthropic");
+        commitMessage = await commitAiGenerate(activeRepoPath, {
+          provider: "anthropic",
+          notes,
+          model: aiSettings.anthropicModel,
+          apiKey: cred?.password,
+        });
+      } else {
+        const cred = await credLoad("ai:custom");
+        commitMessage = await commitAiGenerate(activeRepoPath, {
+          provider: "custom",
+          notes,
+          model: aiSettings.customModel,
+          baseUrl: aiSettings.customBaseUrl,
+          apiKey: cred?.password,
+        });
+      }
    } catch (error) {
      errorMessage = errorToMessage(error);
    } finally {
@@ -446,6 +511,37 @@
    }
  }

+  function defaultAiSettings(): AiSettings {
+    return {
+      provider: "local",
+      localModelId: "qwen2.5-1.5b",
+      openaiModel: "gpt-4o-mini",
+      anthropicModel: "claude-3-5-haiku-latest",
+      customBaseUrl: "",
+      customModel: "",
+    };
+  }
+
+  function loadAiSettings(): AiSettings {
+    try {
+      const stored = JSON.parse(localStorage.getItem(AI_SETTINGS_KEY) ?? "null") as unknown;
+      if (stored && typeof stored === "object") {
+        return { ...defaultAiSettings(), ...(stored as Partial<AiSettings>) };
+      }
+    } catch {
+      // Fall through to defaults below.
+    }
+    return defaultAiSettings();
+  }
+
+  function persistAiSettings(next: AiSettings) {
+    try {
+      localStorage.setItem(AI_SETTINGS_KEY, JSON.stringify(next));
+    } catch {
+      // Local storage is best-effort only; AI generation must keep working without it.
+    }
+  }
+
  function rememberRecentRepo(path: string) {
    recentRepoPaths = uniqueRepoPaths([path, ...recentRepoPaths]).slice(0, 40);
    persistRepoLists();
@@ -1742,11 +1838,13 @@
            {isBusy}
            {operation}
            {stagedCount}
+            commitAiProvider={aiSettings.provider}
            {commitAiPhase}
            {commitAiGenerating}
            onCommit={commitChanges}
            onCommitMessageChange={(msg) => { commitMessage = msg; }}
            onGenerateCommitMessage={generateCommitMessageWithAi}
+            onOpenAiSettings={() => { aiSettingsOpen = true; }}
          />
        </div>
      </section>
@@ -1850,6 +1948,16 @@
  />
 {/if}

+<!-- Choose the AI provider/model used to generate commit messages -->
+{#if aiSettingsOpen}
+  <AiSettingsDialog
+    settings={aiSettings}
+    localModels={localModelOptions}
+    onSave={saveAiSettings}
+    onClose={() => { aiSettingsOpen = false; }}
+  />
+{/if}
+
 <!-- Compare: pick the two commits to diff -->
 {#if compareSelectOpen}
  <CompareSelectDialog
@@ -1334,6 +1334,7 @@
  .commit-form textarea { flex: 1 1 0; min-height: 80px; resize: none; }
  .commit-actions-row { display: flex; gap: 8px; }
  .commit-ai-button { flex: 0 0 auto; min-width: 64px; justify-content: center; }
+  .commit-ai-settings-button { flex: 0 0 auto; width: 38px; min-width: 38px; padding: 0; justify-content: center; }
  .commit-block-reason {
    margin: 0;
    padding: 8px 10px;
@@ -1605,6 +1606,48 @@
    max-height: calc(100vh - 32px);
    overflow: auto;
  }
+  .ai-settings-dialog {
+    display: block;
+    width: min(560px, calc(100vw - 32px));
+    height: auto;
+    max-height: calc(100vh - 32px);
+    overflow: auto;
+  }
+  .ai-settings-form {
+    display: flex;
+    flex-direction: column;
+    gap: 14px;
+    padding: 16px;
+  }
+  .ai-provider-options {
+    display: grid;
+    grid-template-columns: repeat(2, minmax(0, 1fr));
+    gap: 8px;
+  }
+  .ai-provider-option {
+    display: flex;
+    align-items: center;
+    justify-content: center;
+    gap: 7px;
+    min-height: 38px;
+    padding: 0 10px;
+    border: 1px solid var(--color-border-subtle);
+    border-radius: 8px;
+    background: rgba(255,255,255,0.03);
+    color: var(--color-ink-dim);
+    font-size: 12.5px;
+    font-weight: 700;
+  }
+  .ai-provider-option:hover {
+    border-color: var(--color-border);
+    color: var(--color-ink);
+    background: var(--color-surface-hover);
+  }
+  .ai-provider-option.active {
+    border-color: rgba(100,108,255,0.5);
+    color: #f5f7ff;
+    background: linear-gradient(180deg, rgba(100,108,255,0.22), rgba(65,209,255,0.1));
+  }
  .new-branch-form {
    display: flex;
    flex-direction: column;
@@ -0,0 +1,256 @@
+<script lang="ts">
+  import { onMount } from "svelte";
+  import { AlertCircle, Bot, Check, Cpu, Eye, EyeOff, Globe, Key, LoaderCircle, X } from "@lucide/svelte";
+  import { credDelete, credLoad, credSave } from "../git";
+  import type { AiSettings, CommitAiProvider, LocalModelOption } from "../types";
+
+  interface Props {
+    settings: AiSettings;
+    localModels: LocalModelOption[];
+    onSave: (settings: AiSettings) => void;
+    onClose: () => void;
+  }
+
+  let { settings, localModels = [], onSave, onClose }: Props = $props();
+
+  type CloudProvider = Exclude<CommitAiProvider, "local">;
+
+  const CRED_KEYS: Record<CloudProvider, string> = {
+    openai: "ai:openai",
+    anthropic: "ai:anthropic",
+    custom: "ai:custom",
+  };
+
+  let provider = $state<CommitAiProvider>("local");
+  let localModelId = $state("");
+  let openaiModel = $state("");
+  let anthropicModel = $state("");
+  let customBaseUrl = $state("");
+  let customModel = $state("");
+
+  let openaiApiKey = $state("");
+  let anthropicApiKey = $state("");
+  let customApiKey = $state("");
+  let showKey = $state(false);
+  let loadingKeys = $state(true);
+  let saving = $state(false);
+  let error = $state("");
+
+  $effect(() => {
+    provider = settings.provider;
+    localModelId = settings.localModelId;
+    openaiModel = settings.openaiModel;
+    anthropicModel = settings.anthropicModel;
+    customBaseUrl = settings.customBaseUrl;
+    customModel = settings.customModel;
+  });
+
+  onMount(() => {
+    (async () => {
+      try {
+        const [openai, anthropic, custom] = await Promise.all([
+          credLoad(CRED_KEYS.openai),
+          credLoad(CRED_KEYS.anthropic),
+          credLoad(CRED_KEYS.custom),
+        ]);
+        openaiApiKey = openai?.password ?? "";
+        anthropicApiKey = anthropic?.password ?? "";
+        customApiKey = custom?.password ?? "";
+      } catch (err) {
+        error = err instanceof Error ? err.message : String(err);
+      } finally {
+        loadingKeys = false;
+      }
+    })();
+  });
+
+  async function persistKey(target: CloudProvider, value: string) {
+    const key = CRED_KEYS[target];
+    const trimmed = value.trim();
+    if (trimmed) {
+      await credSave(key, "api-key", trimmed, null);
+    } else {
+      await credDelete(key);
+    }
+  }
+
+  async function handleSave() {
+    saving = true;
+    error = "";
+    try {
+      await Promise.all([
+        persistKey("openai", openaiApiKey),
+        persistKey("anthropic", anthropicApiKey),
+        persistKey("custom", customApiKey),
+      ]);
+      onSave({
+        provider,
+        localModelId,
+        openaiModel: openaiModel.trim() || "gpt-4o-mini",
+        anthropicModel: anthropicModel.trim() || "claude-3-5-haiku-latest",
+        customBaseUrl: customBaseUrl.trim(),
+        customModel: customModel.trim(),
+      });
+    } catch (err) {
+      error = err instanceof Error ? err.message : String(err);
+    } finally {
+      saving = false;
+    }
+  }
+
+  function formatSize(mb: number): string {
+    return mb >= 1000 ? `${(mb / 1000).toFixed(1)} GB` : `${mb} MB`;
+  }
+
+  let selectedLocalModel = $derived(localModels.find((option) => option.id === localModelId));
+</script>
+
+<div
+  class="dialog-backdrop"
+  role="presentation"
+  onclick={(e) => { if (e.target === e.currentTarget) onClose(); }}
+>
+  <div class="dialog ai-settings-dialog" role="dialog" aria-modal="true" aria-label="AI settings" tabindex="-1">
+    <header class="dialog-header">
+      <div>
+        <span class="eyebrow">Commit AI</span>
+        <h2 class="mt-0.5 text-ink text-base font-bold leading-tight">AI settings</h2>
+      </div>
+      <button class="dialog-close" type="button" onclick={onClose} title="Close">
+        <X size={18} aria-hidden="true" />
+      </button>
+    </header>
+
+    <form class="ai-settings-form" onsubmit={(e) => { e.preventDefault(); void handleSave(); }}>
+      <div class="ai-provider-options" role="radiogroup" aria-label="AI provider">
+        <button type="button" class="ai-provider-option" class:active={provider === "local"} onclick={() => { provider = "local"; }}>
+          <Cpu size={16} aria-hidden="true" />
+          Local AI
+        </button>
+        <button type="button" class="ai-provider-option" class:active={provider === "openai"} onclick={() => { provider = "openai"; }}>
+          <Bot size={16} aria-hidden="true" />
+          OpenAI
+        </button>
+        <button type="button" class="ai-provider-option" class:active={provider === "anthropic"} onclick={() => { provider = "anthropic"; }}>
+          <Bot size={16} aria-hidden="true" />
+          Anthropic (Claude)
+        </button>
+        <button type="button" class="ai-provider-option" class:active={provider === "custom"} onclick={() => { provider = "custom"; }}>
+          <Globe size={16} aria-hidden="true" />
+          Custom endpoint
+        </button>
+      </div>
+
+      {#if provider === "local"}
+        <label class="cred-field">
+          <span class="cred-field-label">Model</span>
+          <select bind:value={localModelId}>
+            {#each localModels as option (option.id)}
+              <option value={option.id}>{option.label} — {formatSize(option.approx_size_mb)}</option>
+            {/each}
+          </select>
+        </label>
+        <div class="cred-token-hint">
+          <AlertCircle size={13} aria-hidden="true" />
+          <span>
+            Beim Wechsel wird das Modell{selectedLocalModel ? ` (${formatSize(selectedLocalModel.approx_size_mb)})` : ""}
+            im Hintergrund heruntergeladen — je nach Internetverbindung kann das mehrere Minuten dauern.
+            Danach bleibt es lokal zwischengespeichert und lädt beim nächsten Start sofort.
+          </span>
+        </div>
+      {:else if provider === "openai"}
+        <label class="cred-field">
+          <span class="cred-field-label">Model</span>
+          <input type="text" bind:value={openaiModel} placeholder="gpt-4o-mini" autocomplete="off" spellcheck="false" />
+        </label>
+        <div class="cred-field">
+          <span class="cred-field-label">API key</span>
+          <div class="cred-input">
+            <Key size={15} class="cred-field-icon" aria-hidden="true" />
+            <input
+              type={showKey ? "text" : "password"}
+              bind:value={openaiApiKey}
+              placeholder="sk-..."
+              autocomplete="off"
+              spellcheck="false"
+              disabled={loadingKeys}
+            />
+            <button type="button" class="cred-reveal" tabindex="-1" onclick={() => { showKey = !showKey; }} aria-label={showKey ? "Hide" : "Show"}>
+              {#if showKey}<EyeOff size={14} aria-hidden="true" />{:else}<Eye size={14} aria-hidden="true" />{/if}
+            </button>
+          </div>
+        </div>
+      {:else if provider === "anthropic"}
+        <label class="cred-field">
+          <span class="cred-field-label">Model</span>
+          <input type="text" bind:value={anthropicModel} placeholder="claude-3-5-haiku-latest" autocomplete="off" spellcheck="false" />
+        </label>
+        <div class="cred-field">
+          <span class="cred-field-label">API key</span>
+          <div class="cred-input">
+            <Key size={15} class="cred-field-icon" aria-hidden="true" />
+            <input
+              type={showKey ? "text" : "password"}
+              bind:value={anthropicApiKey}
+              placeholder="sk-ant-..."
+              autocomplete="off"
+              spellcheck="false"
+              disabled={loadingKeys}
+            />
+            <button type="button" class="cred-reveal" tabindex="-1" onclick={() => { showKey = !showKey; }} aria-label={showKey ? "Hide" : "Show"}>
+              {#if showKey}<EyeOff size={14} aria-hidden="true" />{:else}<Eye size={14} aria-hidden="true" />{/if}
+            </button>
+          </div>
+        </div>
+      {:else}
+        <label class="cred-field">
+          <span class="cred-field-label">Endpoint URL</span>
+          <input type="text" bind:value={customBaseUrl} placeholder="http://localhost:11434/v1" autocomplete="off" spellcheck="false" />
+        </label>
+        <label class="cred-field">
+          <span class="cred-field-label">Model</span>
+          <input type="text" bind:value={customModel} placeholder="llama3.1" autocomplete="off" spellcheck="false" />
+        </label>
+        <div class="cred-field">
+          <span class="cred-field-label">API key (optional)</span>
+          <div class="cred-input">
+            <Key size={15} class="cred-field-icon" aria-hidden="true" />
+            <input
+              type={showKey ? "text" : "password"}
+              bind:value={customApiKey}
+              placeholder="Optional"
+              autocomplete="off"
+              spellcheck="false"
+              disabled={loadingKeys}
+            />
+            <button type="button" class="cred-reveal" tabindex="-1" onclick={() => { showKey = !showKey; }} aria-label={showKey ? "Hide" : "Show"}>
+              {#if showKey}<EyeOff size={14} aria-hidden="true" />{:else}<Eye size={14} aria-hidden="true" />{/if}
+            </button>
+          </div>
+        </div>
+        <div class="cred-token-hint">
+          <Globe size={13} aria-hidden="true" />
+          <span>Für lokale OpenAI-kompatible Server wie Ollama oder LM Studio. Die Basis-URL sollte auf /v1 enden.</span>
+        </div>
+      {/if}
+
+      {#if error}
+        <p class="commit-block-reason">{error}</p>
+      {/if}
+
+      <div class="new-branch-actions">
+        <button class="btn-secondary" type="button" onclick={onClose} disabled={saving}>
+          Cancel
+        </button>
+        <button class="btn-primary" type="submit" disabled={saving || loadingKeys}>
+          {#if saving}
+            <LoaderCircle class="spin" size={16} aria-hidden="true" />
+          {:else}
+            <Check size={16} aria-hidden="true" />
+          {/if}
+          Save
+        </button>
+      </div>
+    </form>
+  </div>
+</div>
@@ -1,6 +1,6 @@
 <script lang="ts">
-  import { Check, LoaderCircle, Sparkles } from "@lucide/svelte";
-  import type { CommitAiPhase } from "../types";
+  import { Check, LoaderCircle, Settings, Sparkles } from "@lucide/svelte";
+  import type { CommitAiPhase, CommitAiProvider } from "../types";

  interface Props {
    commitMessage: string;
@@ -10,11 +10,13 @@
    isBusy: boolean;
    operation: string;
    stagedCount: number;
+    commitAiProvider: CommitAiProvider;
    commitAiPhase: CommitAiPhase;
    commitAiGenerating: boolean;
    onCommit: () => void;
    onCommitMessageChange: (msg: string) => void;
    onGenerateCommitMessage: () => void;
+    onOpenAiSettings: () => void;
  }

  let {
@@ -25,11 +27,13 @@
    isBusy = false,
    operation = "",
    stagedCount = 0,
+    commitAiProvider = "local",
    commitAiPhase = "idle",
    commitAiGenerating = false,
    onCommit = () => {},
    onCommitMessageChange = () => {},
    onGenerateCommitMessage = () => {},
+    onOpenAiSettings = () => {},
  }: Props = $props();

  function handleSubmit(event: SubmitEvent) {
@@ -37,15 +41,20 @@
    onCommit();
  }

-  function aiButtonTitle(phase: CommitAiPhase, staged: number): string {
-    if (phase === "loading") return "AI model is downloading/loading — this happens once";
-    if (phase === "error") return "AI model failed to load";
+  function aiButtonTitle(provider: CommitAiProvider, phase: CommitAiPhase, staged: number): string {
    if (staged === 0) return "Stage changes first";
+    if (provider === "local" && phase === "loading") return "AI model is downloading/loading — this happens once";
+    if (provider === "local" && phase === "error") return "AI model failed to load — check AI settings";
    return "Generate commit message with AI from the staged diff";
  }

+  let localModelLoading = $derived(commitAiProvider === "local" && commitAiPhase === "loading");
  let canGenerate = $derived(
-    hasRepository && !isBusy && !commitAiGenerating && commitAiPhase === "ready" && stagedCount > 0,
+    hasRepository &&
+    !isBusy &&
+    !commitAiGenerating &&
+    stagedCount > 0 &&
+    (commitAiProvider !== "local" || commitAiPhase === "ready"),
  );
 </script>

@@ -82,15 +91,25 @@
        type="button"
        onclick={onGenerateCommitMessage}
        disabled={!canGenerate}
-        title={aiButtonTitle(commitAiPhase, stagedCount)}
+        title={aiButtonTitle(commitAiProvider, commitAiPhase, stagedCount)}
      >
-        {#if commitAiGenerating || commitAiPhase === "loading"}
+        {#if commitAiGenerating || localModelLoading}
          <LoaderCircle class="spin" size={16} aria-hidden="true" />
        {:else}
          <Sparkles size={16} aria-hidden="true" />
        {/if}
        AI
      </button>
+      <button
+        class="btn-secondary commit-ai-settings-button"
+        type="button"
+        onclick={onOpenAiSettings}
+        disabled={isBusy}
+        title="AI settings"
+        aria-label="AI settings"
+      >
+        <Settings size={16} aria-hidden="true" />
+      </button>
    </div>
  </form>
 </section>
@@ -1,6 +1,7 @@
 import { invoke } from "@tauri-apps/api/core";

 import type {
+  CommitAiProvider,
  CommitAiStatus,
  ConflictFile,
  GitBranch,
@@ -9,6 +10,7 @@ import type {
  GitRepositoryFile,
  GitSearchHit,
  GitStatus,
+  LocalModelOption,
  PatchApplyAction,
  RepositoryBundle,
  StoredCredential,
@@ -99,8 +101,31 @@ export function commitAiStatus(): Promise<CommitAiStatus> {
  return invoke<CommitAiStatus>("commit_ai_status");
 }

-export function commitAiGenerate(path: string, notes?: string): Promise<string> {
-  return invoke<string>("commit_ai_generate", { path, notes });
+export function commitAiLoad(modelId: string): Promise<void> {
+  return invoke<void>("commit_ai_load", { modelId });
+}
+
+export function commitAiLocalModels(): Promise<LocalModelOption[]> {
+  return invoke<LocalModelOption[]>("commit_ai_local_models");
+}
+
+export interface CommitAiGenerateOptions {
+  notes?: string;
+  provider: CommitAiProvider;
+  model?: string;
+  apiKey?: string;
+  baseUrl?: string;
+}
+
+export function commitAiGenerate(path: string, options: CommitAiGenerateOptions): Promise<string> {
+  return invoke<string>("commit_ai_generate", {
+    path,
+    notes: options.notes,
+    provider: options.provider,
+    model: options.model,
+    apiKey: options.apiKey,
+    baseUrl: options.baseUrl,
+  });
 }

 export function pull(path: string, username?: string, password?: string): Promise<GitStatus> {
@@ -8,12 +8,29 @@ export type FileStatusKind =
  | "unknown";

 export type CommitAiPhase = "idle" | "loading" | "ready" | "error";
+export type CommitAiProvider = "local" | "openai" | "anthropic" | "custom";

 export interface CommitAiStatus {
  phase: CommitAiPhase;
+  model_id: string | null;
  error: string | null;
 }

+export interface LocalModelOption {
+  id: string;
+  label: string;
+  approx_size_mb: number;
+}
+
+export interface AiSettings {
+  provider: CommitAiProvider;
+  localModelId: string;
+  openaiModel: string;
+  anthropicModel: string;
+  customBaseUrl: string;
+  customModel: string;
+}
+
 export interface GitStatus {
  repo_path: string;
  current_branch: string | null;