feat: chatbot model setting + tool-calling toggle, simplify admin AI config (v0.59.0)
Chatbot (general assistant + per-recipe Q&A) now resolves its default model from its own site setting (DEFAULT_CHAT_PROVIDER/MODEL) instead of sharing the generic "text" use case with recipe generation, so admins can point it at a different model. Added AI_TOOL_CALLING_ENABLED: turns off the chatbot's createRecipe/ addToShoppingList tools (and the forced-retry pass) for models that don't reliably support tool calling — it falls back to plain text answers. Simplified the admin AI Configuration page: it showed provider keys and routing settings twice (a read-only card, then an identical edit form). Merged into one editable section; the resolved fallback provider is now a one-line note instead of its own card. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
This commit is contained in:
@@ -1,25 +1,16 @@
|
||||
import type { Metadata } from "next";
|
||||
import { Badge } from "@/components/ui/badge";
|
||||
import { Card, CardContent, CardHeader, CardTitle } from "@/components/ui/card";
|
||||
import { CheckCircle, XCircle, Database, Cpu } from "lucide-react";
|
||||
import { getAllSiteSettings } from "@/lib/site-settings";
|
||||
import { AdminSettingsForm } from "@/components/admin/admin-settings-form";
|
||||
import { AdminDefaultModelForm } from "@/components/admin/admin-default-model-form";
|
||||
|
||||
export const metadata: Metadata = {};
|
||||
|
||||
const SETTING_GROUPS = [
|
||||
{
|
||||
title: "AI Provider Keys",
|
||||
description: "Override the environment variable API keys at runtime. Values are encrypted at rest.",
|
||||
keys: ["OPENAI_API_KEY", "ANTHROPIC_API_KEY", "OPENROUTER_API_KEY"] as const,
|
||||
},
|
||||
{
|
||||
title: "AI Configuration",
|
||||
description: "Provider routing and model defaults.",
|
||||
keys: ["OPENROUTER_DEFAULT_MODEL", "OLLAMA_BASE_URL"] as const,
|
||||
},
|
||||
];
|
||||
const SETTING_GROUP = {
|
||||
title: "Provider Keys & Routing",
|
||||
description: "Override the environment variable API keys at runtime (encrypted at rest), and set the OpenRouter/Ollama routing defaults. DB values here take precedence over environment variables — clear a value to fall back to the env var.",
|
||||
keys: ["OPENAI_API_KEY", "ANTHROPIC_API_KEY", "OPENROUTER_API_KEY", "OPENROUTER_DEFAULT_MODEL", "OLLAMA_BASE_URL"] as const,
|
||||
};
|
||||
|
||||
function resolveProvider(settings: Record<string, { value: string | null }>) {
|
||||
if (settings["OPENROUTER_API_KEY"]?.value) return { provider: "OpenRouter", description: "Routes to many models via openrouter.ai" };
|
||||
@@ -33,100 +24,34 @@ export default async function AdminAiConfigPage() {
|
||||
const settings = await getAllSiteSettings();
|
||||
const active = resolveProvider(settings);
|
||||
|
||||
const keyRows = [
|
||||
{ key: "OPENROUTER_API_KEY", label: "OPENROUTER_API_KEY" },
|
||||
{ key: "OPENAI_API_KEY", label: "OPENAI_API_KEY" },
|
||||
{ key: "ANTHROPIC_API_KEY", label: "ANTHROPIC_API_KEY" },
|
||||
];
|
||||
|
||||
return (
|
||||
<div className="space-y-8">
|
||||
<div>
|
||||
<h1 className="text-2xl font-bold tracking-tight">AI Configuration</h1>
|
||||
<p className="text-muted-foreground text-sm mt-1">
|
||||
Provider keys, routing, and default models for AI generation. DB values here take precedence over environment variables — clear a value to fall back to the env var.
|
||||
Provider keys, routing, and default models for AI generation.
|
||||
</p>
|
||||
<div className="flex items-center gap-2 mt-3">
|
||||
<span className="text-xs text-muted-foreground">Fallback provider (no default set below):</span>
|
||||
<Badge variant="secondary" className="text-xs">{active.provider}</Badge>
|
||||
<span className="text-xs text-muted-foreground">— {active.description}</span>
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<Card>
|
||||
<CardHeader>
|
||||
<CardTitle className="text-base">Active Provider (fallback, no default set below)</CardTitle>
|
||||
</CardHeader>
|
||||
<CardContent>
|
||||
<div className="flex items-center gap-3">
|
||||
<Badge className="text-sm px-3 py-1">{active.provider}</Badge>
|
||||
<span className="text-muted-foreground text-sm">{active.description}</span>
|
||||
</div>
|
||||
<p className="text-xs text-muted-foreground mt-3">
|
||||
Priority: OpenRouter → OpenAI → Anthropic → Ollama (first configured key wins) — only when a user has no personal preference, no BYOK key, and no default is set for that use case below.
|
||||
</p>
|
||||
</CardContent>
|
||||
</Card>
|
||||
|
||||
<Card>
|
||||
<CardHeader>
|
||||
<CardTitle className="text-base">API Keys</CardTitle>
|
||||
</CardHeader>
|
||||
<CardContent>
|
||||
{keyRows.map(({ key, label }) => {
|
||||
const meta = settings[key];
|
||||
const present = !!meta?.value;
|
||||
return (
|
||||
<div key={key} className="flex items-center justify-between py-2 border-b last:border-0">
|
||||
<div className="flex items-center gap-2">
|
||||
{present ? (
|
||||
<CheckCircle className="h-4 w-4 text-green-500" />
|
||||
) : (
|
||||
<XCircle className="h-4 w-4 text-muted-foreground" />
|
||||
)}
|
||||
<span className="font-mono text-sm">{label}</span>
|
||||
</div>
|
||||
<div className="flex items-center gap-2">
|
||||
{present && meta?.fromDb && (
|
||||
<span className="flex items-center gap-1 text-xs text-muted-foreground">
|
||||
<Database className="h-3 w-3" /> DB override
|
||||
</span>
|
||||
)}
|
||||
{present && !meta?.fromDb && (
|
||||
<span className="flex items-center gap-1 text-xs text-muted-foreground">
|
||||
<Cpu className="h-3 w-3" /> from .env
|
||||
</span>
|
||||
)}
|
||||
<Badge variant={present ? "default" : "secondary"}>
|
||||
{present ? "Configured" : "Not set"}
|
||||
</Badge>
|
||||
</div>
|
||||
</div>
|
||||
);
|
||||
})}
|
||||
<div className="flex items-center justify-between py-2 border-b">
|
||||
<span className="font-mono text-sm text-muted-foreground">OLLAMA_BASE_URL</span>
|
||||
<span className="text-sm text-muted-foreground font-mono">
|
||||
{settings["OLLAMA_BASE_URL"]?.value || "http://localhost:11434 (default)"}
|
||||
</span>
|
||||
</div>
|
||||
<div className="flex items-center justify-between py-2">
|
||||
<span className="font-mono text-sm text-muted-foreground">OPENROUTER_DEFAULT_MODEL</span>
|
||||
<span className="text-sm text-muted-foreground font-mono">
|
||||
{settings["OPENROUTER_DEFAULT_MODEL"]?.value || "google/gemini-flash-1.5 (default)"}
|
||||
</span>
|
||||
</div>
|
||||
</CardContent>
|
||||
</Card>
|
||||
|
||||
{SETTING_GROUPS.map((group) => (
|
||||
<AdminSettingsForm key={group.title} group={group} settings={settings} />
|
||||
))}
|
||||
<AdminSettingsForm group={SETTING_GROUP} settings={settings} />
|
||||
|
||||
<AdminDefaultModelForm
|
||||
initialValues={{
|
||||
DEFAULT_TEXT_PROVIDER: settings["DEFAULT_TEXT_PROVIDER"]?.value ?? null,
|
||||
DEFAULT_TEXT_MODEL: settings["DEFAULT_TEXT_MODEL"]?.value ?? null,
|
||||
DEFAULT_CHAT_PROVIDER: settings["DEFAULT_CHAT_PROVIDER"]?.value ?? null,
|
||||
DEFAULT_CHAT_MODEL: settings["DEFAULT_CHAT_MODEL"]?.value ?? null,
|
||||
DEFAULT_VISION_PROVIDER: settings["DEFAULT_VISION_PROVIDER"]?.value ?? null,
|
||||
DEFAULT_VISION_MODEL: settings["DEFAULT_VISION_MODEL"]?.value ?? null,
|
||||
DEFAULT_MEAL_PLAN_PROVIDER: settings["DEFAULT_MEAL_PLAN_PROVIDER"]?.value ?? null,
|
||||
DEFAULT_MEAL_PLAN_MODEL: settings["DEFAULT_MEAL_PLAN_MODEL"]?.value ?? null,
|
||||
}}
|
||||
initialToolCallingEnabled={settings["AI_TOOL_CALLING_ENABLED"]?.value !== "false"}
|
||||
/>
|
||||
</div>
|
||||
);
|
||||
|
||||
@@ -20,6 +20,9 @@ const ALLOWED_KEYS: SiteSettingKey[] = [
|
||||
"DEFAULT_VISION_MODEL",
|
||||
"DEFAULT_MEAL_PLAN_PROVIDER",
|
||||
"DEFAULT_MEAL_PLAN_MODEL",
|
||||
"DEFAULT_CHAT_PROVIDER",
|
||||
"DEFAULT_CHAT_MODEL",
|
||||
"AI_TOOL_CALLING_ENABLED",
|
||||
"GITEA_URL",
|
||||
"GITEA_TOKEN",
|
||||
"GITEA_REPO",
|
||||
|
||||
@@ -7,6 +7,7 @@ import { applyRateLimit } from "@/lib/rate-limit";
|
||||
import { withAiQuota, resolveAiConfigOrError } from "@/lib/ai/ai-error";
|
||||
import { getModelConfigForUseCase } from "@/lib/ai/resolve-user-key";
|
||||
import { resolveModel } from "@/lib/ai/factory";
|
||||
import { isAiToolCallingEnabled } from "@/lib/site-settings";
|
||||
import { getUserPrivateBio, buildUserBioContext } from "@/lib/ai/user-bio";
|
||||
import { createRecipeTool } from "@/lib/ai/tools/create-recipe-tool";
|
||||
import { addToShoppingListTool } from "@/lib/ai/tools/add-to-shopping-list-tool";
|
||||
@@ -68,9 +69,10 @@ export async function POST(req: NextRequest) {
|
||||
const limited = await applyRateLimit(`rl:ai:${session!.user.id}`, 30, 60);
|
||||
if (limited) return limited;
|
||||
|
||||
const [configResult, privateBio] = await Promise.all([
|
||||
resolveAiConfigOrError(() => getModelConfigForUseCase(session!.user.id, "text")),
|
||||
const [configResult, privateBio, toolCallingEnabled] = await Promise.all([
|
||||
resolveAiConfigOrError(() => getModelConfigForUseCase(session!.user.id, "chat")),
|
||||
getUserPrivateBio(session!.user.id),
|
||||
isAiToolCallingEnabled(),
|
||||
]);
|
||||
if (!configResult.ok) return configResult.response;
|
||||
const aiConfig = configResult.data;
|
||||
@@ -79,16 +81,23 @@ export async function POST(req: NextRequest) {
|
||||
const locale = (session!.user as { locale?: string }).locale ?? "en";
|
||||
const lang = LANG[locale] ?? "English";
|
||||
|
||||
const system = `You are Epicure, a helpful culinary assistant answering general cooking questions — not tied to any specific recipe (techniques, substitutions, timing, equipment, food safety, etc). If asked who you are or what model/AI you're built on, say you're Epicure — never name the underlying model or provider. If a question has nothing to do with cooking or food, politely redirect. Keep answers under 200 words. Respond in ${lang}.\n\nYou have two tools. Using one only drafts something for the user to review — it never saves by itself.\n- createRecipe: the user is asking you to create, save, or write down a recipe (e.g. "make me a recipe for X", "give me a recipe for Y", "write that down"). This includes any request for a full recipe, not only ones that say the word "create" or "save".\n- addToShoppingList: the user is asking to add ingredients/items to a shopping list.\n\nIMPORTANT: when the user's request matches createRecipe, you MUST call that tool instead of writing the recipe's ingredients or steps directly in your text reply. Never output a full ingredient list or numbered steps as plain text — that content belongs in the tool call, not the message. Your text reply in that case should just be a short line like "Here's a draft — check it below and confirm if it looks right." Only skip the tool if the user is asking a general question (no specific recipe requested) or explicitly wants prose, not a structured recipe.${bioContext}`;
|
||||
const toolsInstructions = toolCallingEnabled
|
||||
? `\n\nYou have two tools. Using one only drafts something for the user to review — it never saves by itself.\n- createRecipe: the user is asking you to create, save, or write down a recipe (e.g. "make me a recipe for X", "give me a recipe for Y", "write that down"). This includes any request for a full recipe, not only ones that say the word "create" or "save".\n- addToShoppingList: the user is asking to add ingredients/items to a shopping list.\n\nIMPORTANT: when the user's request matches createRecipe, you MUST call that tool instead of writing the recipe's ingredients or steps directly in your text reply. Never output a full ingredient list or numbered steps as plain text — that content belongs in the tool call, not the message. Your text reply in that case should just be a short line like "Here's a draft — check it below and confirm if it looks right." Only skip the tool if the user is asking a general question (no specific recipe requested) or explicitly wants prose, not a structured recipe.`
|
||||
: "";
|
||||
const system = `You are Epicure, a helpful culinary assistant answering general cooking questions — not tied to any specific recipe (techniques, substitutions, timing, equipment, food safety, etc). If asked who you are or what model/AI you're built on, say you're Epicure — never name the underlying model or provider. If a question has nothing to do with cooking or food, politely redirect. Keep answers under 200 words. Respond in ${lang}.${toolsInstructions}${bioContext}`;
|
||||
|
||||
const result = await withAiQuota(session!.user.id, session!.user.tier as "free" | "pro" | "family", () =>
|
||||
generateText({
|
||||
model,
|
||||
system,
|
||||
prompt: parsed.data.question,
|
||||
tools: { createRecipe: createRecipeTool, addToShoppingList: addToShoppingListTool },
|
||||
stopWhen: stepCountIs(3),
|
||||
}), { skipQuota: aiConfig.isByok }
|
||||
generateText(
|
||||
toolCallingEnabled
|
||||
? {
|
||||
model,
|
||||
system,
|
||||
prompt: parsed.data.question,
|
||||
tools: { createRecipe: createRecipeTool, addToShoppingList: addToShoppingListTool },
|
||||
stopWhen: stepCountIs(3),
|
||||
}
|
||||
: { model, system, prompt: parsed.data.question }
|
||||
), { skipQuota: aiConfig.isByok }
|
||||
);
|
||||
if (!result.ok) return result.response;
|
||||
|
||||
@@ -101,8 +110,10 @@ export async function POST(req: NextRequest) {
|
||||
// Catch both: the prose-list shape, or the user's own message clearly
|
||||
// asking for a recipe in the first place. Rather than surface either
|
||||
// failure, force the tool call on one extra pass — same system+prompt,
|
||||
// just no longer letting the model opt out.
|
||||
if (final.toolCalls.length === 0 && (looksLikeUnstructuredRecipe(final.text) || looksLikeRecipeCreationRequest(parsed.data.question, locale))) {
|
||||
// just no longer letting the model opt out. Skipped entirely when tool
|
||||
// calling is turned off admin-side (e.g. a local model that can't
|
||||
// reliably call tools) — there's no tool to force in that case.
|
||||
if (toolCallingEnabled && final.toolCalls.length === 0 && (looksLikeUnstructuredRecipe(final.text) || looksLikeRecipeCreationRequest(parsed.data.question, locale))) {
|
||||
try {
|
||||
const forced = await generateText({
|
||||
model,
|
||||
|
||||
@@ -64,7 +64,7 @@ ${stepList || "None listed"}
|
||||
`.trim();
|
||||
|
||||
const [configResult, privateBio] = await Promise.all([
|
||||
resolveAiConfigOrError(() => getModelConfigForUseCase(session!.user.id, "text")),
|
||||
resolveAiConfigOrError(() => getModelConfigForUseCase(session!.user.id, "chat")),
|
||||
getUserPrivateBio(session!.user.id),
|
||||
]);
|
||||
if (!configResult.ok) return configResult.response;
|
||||
|
||||
Reference in New Issue
Block a user