feat(memory): config embedders (LOT C2) + suggestion contextuelle (LOT C3) + contexte projet partagé

- LOT C2 (§14.5.3) : use cases de configuration des embedders déclaratifs
  (List/Save/Delete + DescribeEmbedderEngines : modèles ONNX recommandés,
  environnement local détecté, stratégies compilées). UI EmbedderSettings.
- LOT C3 (§14.5.5) : suggestion contextuelle best-effort à l'activation quand la
  mémoire dépasse le budget de recall sans embedder configuré (event
  EmbedderSuggested, anti-spam 1×/session, « ne plus demander »).
- Contexte projet partagé .ideai/CONTEXT.md (model-agnostic) injecté à tous les
  agents/profils au lancement, avant la persona. UI ProjectContextPanel.

Tests : backend workspace vert (0 échec) ; frontend 306/306.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

crates/domain/src/ports.rs

@@ -32,7 +32,7 @@ use crate::events::DomainEvent;
 use crate::ids::{AgentId, SessionId};
 use crate::markdown::MarkdownDoc;
 use crate::memory::{Memory, MemoryIndexEntry, MemoryLink, MemorySlug};
-use crate::profile::AgentProfile;
+use crate::profile::{AgentProfile, EmbedderProfile};
 use crate::project::{Project, ProjectPath};
 use crate::remote::RemoteKind;
 use crate::skill::{Skill, SkillScope};
@@ -798,6 +798,94 @@ pub trait ProfileStore: Send + Sync {
     async fn mark_configured(&self) -> Result<(), StoreError>;
 }
 
+/// CRUD for the declarative [`EmbedderProfile`]s in the global IDE store
+/// (`embedder.json`, LOT C, §14.5.3). Mirrors [`ProfileStore`] for the embedding
+/// engines: adding an engine is *data* the user configures, not code.
+///
+/// There is **no** `is_configured`/`mark_configured`: the default embedder posture
+/// is `none` **by absence of the file** (an empty/missing `embedder.json` ⇒ recall
+/// stays the dependency-free naïve étage 1). A configured embedder takes effect at
+/// the **next IDE start** (the composition root freezes the recall for the session).
+#[async_trait]
+pub trait EmbedderProfileStore: Send + Sync {
+    /// Lists all configured embedder profiles (empty when none configured).
+    ///
+    /// # Errors
+    /// [`StoreError`] on failure.
+    async fn list(&self) -> Result<Vec<EmbedderProfile>, StoreError>;
+
+    /// Saves (creates or replaces by id) an embedder profile.
+    ///
+    /// # Errors
+    /// [`StoreError`] on failure.
+    async fn save(&self, profile: &EmbedderProfile) -> Result<(), StoreError>;
+
+    /// Deletes an embedder profile by id.
+    ///
+    /// # Errors
+    /// [`StoreError::NotFound`] if no profile carries that id.
+    async fn delete(&self, id: &str) -> Result<(), StoreError>;
+}
+
+/// Best-effort snapshot of the embedding *environment*: which local engines are
+/// already installed/cached on this machine. Drives the "configure an embedder?"
+/// UI (C2/C3) so it can detect an existing engine before suggesting a download.
+#[derive(Debug, Clone, Default, PartialEq, Eq)]
+pub struct EmbedderEnvReport {
+    /// Whether an Ollama-style local embedding server was detected (best-effort;
+    /// always `false` in a build without the HTTP capability).
+    pub ollama_detected: bool,
+    /// Ids of the recommended ONNX models already present in the local cache.
+    pub onnx_cached_models: Vec<String>,
+}
+
+/// Probes the local embedding environment (installed/cached engines). **Best-effort
+/// by contract**: [`inspect`](Self::inspect) never errors — an unreachable host or
+/// an unreadable cache simply yields a report with the corresponding fields unset.
+#[async_trait]
+pub trait EmbedderEnvInspector: Send + Sync {
+    /// Returns a best-effort snapshot of the local embedding environment.
+    async fn inspect(&self) -> EmbedderEnvReport;
+}
+
+/// The persisted user response to the "configure an embedder?" suggestion
+/// (ARCHITECTURE §14.5.5, LOT C3), stored per project under
+/// `.ideai/memory/.embedder-prompt.json`. Absence of the file ⇒ never answered.
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum EmbedderPromptDismissal {
+    /// "Plus tard" — re-proposable on a future session (not this one).
+    Later,
+    /// "Ne plus demander" — never publish the suggestion again (persistent).
+    Never,
+}
+
+/// Persistence of the per-project embedder-suggestion state
+/// (`.ideai/memory/.embedder-prompt.json`, LOT C3). A fine-grained port (Interface
+/// Segregation) so the suggestion use cases depend on this alone, not on a wider
+/// store. **Best-effort reads**: a missing/malformed file is *not* an error — it
+/// reads as "never answered" ([`None`]).
+#[async_trait]
+pub trait EmbedderPromptStore: Send + Sync {
+    /// Reads the recorded dismissal for `root`'s project, or `None` when the user
+    /// has never answered (no file / unreadable / malformed).
+    ///
+    /// # Errors
+    /// [`StoreError`] only on an unexpected I/O failure the adapter chooses to
+    /// surface; a plain missing file degrades to `Ok(None)`.
+    async fn read(&self, root: &ProjectPath)
+        -> Result<Option<EmbedderPromptDismissal>, StoreError>;
+
+    /// Records the user's dismissal choice for `root`'s project.
+    ///
+    /// # Errors
+    /// [`StoreError`] on a persistence failure.
+    async fn write(
+        &self,
+        root: &ProjectPath,
+        dismissal: EmbedderPromptDismissal,
+    ) -> Result<(), StoreError>;
+}
+
 /// Reads/writes agent `.md` contexts and the project manifest, within a project.
 #[async_trait]
 pub trait AgentContextStore: Send + Sync {
@@ -895,8 +983,7 @@ pub trait GitPort: Send + Sync {
     ///
     /// # Errors
     /// [`GitError`] on failure.
-    async fn log(&self, root: &ProjectPath, limit: usize)
-        -> Result<Vec<GitCommitInfo>, GitError>;
+    async fn log(&self, root: &ProjectPath, limit: usize) -> Result<Vec<GitCommitInfo>, GitError>;
 
     /// Returns the commit graph (all local branches, topological + time sort).
     ///