Phase 7 : outils web_search et run_shell avec validation

Backend :
- tools.py : web_search (DuckDuckGo HTML sans clé, ou SearxNG via SEARX_URL) et
  run_shell (confiné au workspace, sortie bornée, timeout)
- agent.py : run_shell sensible -> émet tool_confirm et ne s'exécute pas tant que
  l'utilisateur n'a pas validé (confirm_shell)
- agent_config.py : web_search/run_shell désactivés par défaut, flag confirm_shell
- routes/shell.py : POST /api/shell/run exécute une commande validée
- routes/chat.py : relaie l'événement tool_confirm, passe confirm_shell

Frontend :
- streamChat : événement tool_confirm
- store : pendingShell + approveShell/rejectShell (réinjecte le résultat à l'agent)
- ChatPanel : carte de validation de commande (Approuver / Refuser)
- SettingsView : toggles web_search/run_shell, marqueur sensible, switch confirm_shell
- ToolCard : glyphes web_search/run_shell, statut 'à valider', aperçu d'argument

Tests : run_shell confiné, gate de confirmation (commande dangereuse non exécutée),
route shell, parser DuckDuckGo.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SVay7z3y7q2gEe54ByAE6N
This commit is contained in:
MichaelandClaude Opus 4.8 committed 2026-06-29 21:09:27 +00:00
1 parent 2db6beea5c
commit de2befe31e
12 files changed
+398 -26

No files matched your search

+15 -1
View File
@@ -42,7 +42,7 @@ export interface ToolCall {
name: string;
args: Record<string, unknown>;
summary?: string;
status?: "ok" | "error" | "running";
status?: "ok" | "error" | "running" | "pending";
}
export interface Message {
@@ -62,6 +62,18 @@ export interface AgentConfig {
top_k: number;
max_tokens: number;
tools: Record<string, boolean>;
confirm_shell: boolean;
}
export async function runShell(
command: string
): Promise<{ command: string; exit_code: number; output: string }> {
const res = await fetch("/api/shell/run", {
method: "POST",
headers: { "Content-Type": "application/json" },
body: JSON.stringify({ command }),
});
return res.json();
}
export async function getConfig(): Promise<{
@@ -134,6 +146,7 @@ export async function streamChat(
onToken: (t: string) => void;
onToolCall: (call: ToolCall) => void;
onToolResult: (call: ToolCall) => void;
onToolConfirm: (command: string) => void;
onDone: (full: string) => void;
onError: (msg: string) => void;
}
@@ -173,6 +186,7 @@ export async function streamChat(
else if (event === "tool_call")
handlers.onToolCall({ ...payload, status: "running" });
else if (event === "tool_result") handlers.onToolResult(payload);
else if (event === "tool_confirm") handlers.onToolConfirm(payload.command);
else if (event === "done") handlers.onDone(payload.content);
else if (event === "error") handlers.onError(payload.message);
} catch {
+24 -3
View File
@@ -4,11 +4,16 @@ import type { ToolCall } from "../api/client";
export function ToolCard({ call }: { call: ToolCall }) {
const running = call.status === "running";
const error = call.status === "error";
const pending = call.status === "pending";
// Argument principal affiché entre parenthèses (path le plus souvent).
// Argument principal affiché entre parenthèses selon l'outil.
const mainArg =
(call.args?.path as string) ??
(call.args?.query as string) ??
(call.args?.command as string);
const argPreview =
typeof call.args?.path === "string"
? `("${call.args.path}")`
typeof mainArg === "string"
? `("${mainArg.length > 48 ? mainArg.slice(0, 48) + "…" : mainArg}")`
: Object.keys(call.args ?? {}).length
? "(…)"
: "()";
@@ -32,6 +37,8 @@ export function ToolCard({ call }: { call: ToolCall }) {
<span className="h-2.5 w-2.5 animate-spin rounded-full border-2 border-muted-3 border-t-accent" />
en cours
</span>
) : pending ? (
<span className="font-mono text-[11px] text-warn">⏸ à valider</span>
) : (
<span
className={`flex items-center gap-1.5 font-mono text-[11px] ${
@@ -74,6 +81,20 @@ function ToolGlyph({ name }: { name: string }) {
<path d="M4 7h6l2 2h8v9H4z" />
</svg>
);
if (name === "web_search")
return (
<svg {...common}>
<circle cx="11" cy="11" r="7" />
<path d="m21 21-4.3-4.3" />
</svg>
);
if (name === "run_shell")
return (
<svg {...common}>
<path d="m6 9 3 3-3 3M13 15h5" />
<rect x="2" y="4" width="20" height="16" rx="2" />
</svg>
);
// read_file (défaut)
return (
<svg {...common}>
+50
View File
@@ -15,6 +15,9 @@ export function ChatPanel() {
sendMessage,
currentSessionId,
config,
pendingShell,
approveShell,
rejectShell,
} = useStore();
const activeTools = config
@@ -100,6 +103,13 @@ export function ChatPanel() {
pending
/>
)}
{pendingShell && (
<ShellConfirm
command={pendingShell}
onApprove={approveShell}
onReject={rejectShell}
/>
)}
</div>
)}
</div>
@@ -201,6 +211,46 @@ function Bubble({ msg, pending }: { msg: Message; pending?: boolean }) {
);
}
function ShellConfirm({
command,
onApprove,
onReject,
}: {
command: string;
onApprove: () => void;
onReject: () => void;
}) {
return (
<div className="ml-[43px] overflow-hidden rounded-[11px] border border-[#4a3a2a] bg-card-deep">
<div className="flex items-center gap-2 border-b border-line-soft px-[13px] py-2.5">
<span className="font-mono text-[12.5px] font-semibold text-warn">
run_shell
</span>
<span className="text-xs text-muted-2">· commande sensible à valider</span>
</div>
<div className="px-[13px] py-3">
<pre className="m-0 mb-3 overflow-auto whitespace-pre-wrap rounded-lg bg-sunken px-3 py-2.5 font-mono text-[12px] text-ink-2">
$ {command}
</pre>
<div className="flex items-center gap-2">
<button
onClick={onApprove}
className="flex h-[30px] items-center gap-1.5 rounded-lg bg-accent px-3.5 text-[12.5px] font-bold text-[#1f1b16]"
>
Approuver &amp; exécuter
</button>
<button
onClick={onReject}
className="flex h-[30px] items-center rounded-lg border border-line-strong px-3.5 text-[12.5px] font-semibold text-ink-3"
>
Refuser
</button>
</div>
</div>
</div>
);
}
function Chip({ children }: { children: React.ReactNode }) {
return (
<div className="flex h-[25px] items-center gap-1.5 rounded-[7px] border border-line-strong bg-card-soft px-[9px] text-[11.5px] text-ink-3">
+28 -3
View File
@@ -8,7 +8,10 @@ const TOOL_DESC: Record<string, string> = {
read_file: "Lire un fichier du projet",
write_file: "Créer / modifier un fichier",
list_dir: "Lister un répertoire",
web_search: "Recherche web",
run_shell: "Exécuter une commande",
};
const SENSITIVE = new Set(["run_shell"]);
/** Vue Configuration (Frame 2) — Modèle & génération, outils, invite système. */
export function SettingsView() {
@@ -279,6 +282,9 @@ export function SettingsView() {
</span>
<span className="flex-1 text-xs text-muted-2">
{TOOL_DESC[name] ?? ""}
{SENSITIVE.has(name) && (
<span className="text-warn"> · sensible</span>
)}
</span>
<button
onClick={() =>
@@ -298,9 +304,28 @@ export function SettingsView() {
</div>
);
})}
<div className="mt-1 border-t border-line-soft pt-2.5 font-mono text-[11px] text-muted-3">
web_search · run_shell — arrivent bientôt (sensibles)
</div>
{draft.tools.run_shell && (
<div className="mt-2 flex items-center gap-3 rounded-[10px] border border-line-soft bg-base px-3 py-2.5">
<span className="flex-1 text-xs text-ink-3">
Demander une validation avant chaque commande shell
</span>
<button
onClick={() => set("confirm_shell", !draft.confirm_shell)}
className="relative h-5 w-[34px] rounded-full transition-colors"
style={{
background: draft.confirm_shell ? "#f0a15c" : "#3a342c",
}}
>
<span
className="absolute top-0.5 h-4 w-4 rounded-full transition-all"
style={{
background: draft.confirm_shell ? "#1f1b16" : "#6f675c",
left: draft.confirm_shell ? 16 : 2,
}}
/>
</button>
</div>
)}
</div>
</div>
</div>
+33
View File
@@ -9,6 +9,7 @@ import {
fileContent,
listFiles,
listSessions,
runShell,
saveConfig,
streamChat,
type AgentConfig,
@@ -42,6 +43,10 @@ interface LokiState {
refreshConfig: () => Promise<void>;
updateConfig: (patch: Partial<AgentConfig>) => Promise<void>;
pendingShell: string | null; // commande shell en attente de validation
approveShell: () => Promise<void>;
rejectShell: () => Promise<void>;
openPreview: (path: string) => Promise<void>;
setSelectedModel: (name: string) => void;
refreshStatus: () => Promise<void>;
@@ -72,6 +77,32 @@ export const useStore = create<LokiState>((set, get) => ({
previewContent: "",
config: null,
availableTools: [],
pendingShell: null,
approveShell: async () => {
const cmd = get().pendingShell;
if (!cmd) return;
set({ pendingShell: null });
let report: string;
try {
const r = await runShell(cmd);
report =
`J'ai validé la commande \`${cmd}\` (code ${r.exit_code}).\n` +
`Sortie :\n\`\`\`\n${r.output || "(vide)"}\n\`\`\``;
} catch {
report = `Échec de l'exécution de \`${cmd}\`.`;
}
await get().refreshFiles();
// On renvoie le résultat à l'agent pour qu'il poursuive.
await get().sendMessage(report);
},
rejectShell: async () => {
const cmd = get().pendingShell;
if (!cmd) return;
set({ pendingShell: null });
await get().sendMessage(`J'ai refusé la commande \`${cmd}\`. N'exécute pas cette commande.`);
},
refreshConfig: async () => {
const { config, available_tools } = await getConfig();
@@ -171,6 +202,7 @@ export const useStore = create<LokiState>((set, get) => ({
streaming: true,
streamContent: "",
streamTools: [],
pendingShell: null,
});
await streamChat(
@@ -190,6 +222,7 @@ export const useStore = create<LokiState>((set, get) => ({
}
set({ streamTools: tools });
},
onToolConfirm: (command) => set({ pendingShell: command }),
onDone: async () => {
// Repère un fichier HTML écrit pour l'afficher automatiquement.
const writtenHtml = [...get().streamTools]