feat(opencode): remplace le provider Ollama par llama.cpp

Le tool-calling local ne fonctionnait jamais via Ollama. Refonte du
support local d'OpenCode autour de llama.cpp: profil, catalogue,
matérialisation de la config OpenCode et surface first-run alignés sur
llama-server (backend + frontend).

QA vert (commandes réelles): domain 244, application 81+64, infra 263
(10 échecs = bind-port sandbox identiques sur develop, non-régression),
frontend 574, tsc propre. Réserve E2E live non bloquante faute de
llama-server joignable.

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

frontend/src/features/first-run/profile.ts

@@ -9,6 +9,7 @@ import type {
   ContextInjection,
   HttpChatConfig,
   InjectionStrategy,
+  OpenCodeConfig,
 } from "@/domain";
 
 /** A field-keyed validation error map (empty ⇒ valid). */
@@ -20,6 +21,7 @@ export type ProfileErrors = Partial<
     | "flag"
     | "var"
     | "endpoint"
+    | "baseURL"
     | "model"
     | "apiKeyEnv"
     | "requestTimeoutMs"
@@ -95,6 +97,23 @@ export function validateHttpChatConfig(c: HttpChatConfig): ProfileErrors {
   return errors;
 }
 
+/**
+ * Validates an {@link OpenCodeConfig} draft the way the backend
+ * `OpenCodeConfig::new` would (base URL http/https + non-empty, model non-empty;
+ * the API key is a free-form, optional secret). Returns an empty object when the
+ * draft is valid.
+ */
+export function validateOpenCodeConfig(c: OpenCodeConfig): ProfileErrors {
+  const errors: ProfileErrors = {};
+  if (!isValidHttpUrl(c.baseURL)) {
+    errors.baseURL = "Base URL must start with http:// or https://.";
+  }
+  if (c.model.trim().length === 0) {
+    errors.model = "Model is required.";
+  }
+  return errors;
+}
+
 /**
  * Validates a profile draft the way the backend would, so the wizard can surface
  * errors before any `invoke`. Returns an empty object when the draft is valid.
@@ -125,8 +144,15 @@ export function validateProfile(p: AgentProfile): ProfileErrors {
       Object.assign(errors, validateHttpChatConfig(p.chatHttp));
     }
   }
-  if (p.structuredAdapter === "openCode" && p.opencode?.model !== undefined) {
-    if (p.opencode.model.trim().length === 0) errors.model = "Model is required.";
+  // An OpenCode profile carries its own llama.cpp endpoint config; the backend
+  // refuses to persist it unless base URL + model are well-formed, so mirror that.
+  if (p.structuredAdapter === "openCode") {
+    if (!p.opencode) {
+      errors.baseURL = "Base URL must start with http:// or https://.";
+      errors.model = "Model is required.";
+    } else {
+      Object.assign(errors, validateOpenCodeConfig(p.opencode));
+    }
   }
   return errors;
 }
@@ -168,6 +194,19 @@ export function defaultHttpChatConfig(): HttpChatConfig {
   };
 }
 
+/**
+ * A sensible default {@link OpenCodeConfig} for a fresh OpenCode profile draft —
+ * pre-fills a local `llama-server` endpoint so the form starts valid. Mirrors the
+ * backend reference profile (`http://localhost:8080/v1`, `qwen3-coder-30b`).
+ */
+export function defaultOpenCodeConfig(): OpenCodeConfig {
+  return {
+    baseURL: "http://localhost:8080/v1",
+    apiKey: "sk-no-key",
+    model: "qwen3-coder-30b",
+  };
+}
+
 /** A fresh, empty custom-profile draft (id minted client-side). */
 export function emptyCustomProfile(): AgentProfile {
   return {