feat(ai): real local Ollama chat + BYOK public provider
Decisions 2026-08-05 evening (reprioritizing docs/AI-ASSISTANT-CONCEPT.md's original P1/P2 server-first sequencing to local-first, since a real Ollama instance already runs on this Mac with a full model set): - `local` ships free, no entitlement check — always available wherever supportsLocalLlm() is true. - `public` (BYOK) is available too, explicitly unmonitored for now — no seats/metering/consent backend. This reverses the concept doc's decision #1 (server-side-only key custody): the client holds its own key, matching vncmail-native's existing pattern. - `server` (VNC-hosted) stays unwired client-side; that infra is "this MacBook tonight, the dev k8s cluster tomorrow." New: - lib/ai/local-client.ts: listLocalModels/testLocalConnection/chatLocal/ chatPublic, ported near-verbatim from vncmail-native's proven src/api/ai.ts. Direct browser-side fetch, not proxied through this app's own server — a server-side proxy would reach the *server's* loopback, not the user's own laptop, which defeats the point of "local" once this app is hosted remotely. - lib/ai/key-store.ts: client-held BYOK storage (localStorage — this repo's existing convention for client state, no OS keychain reachable from a browser tab). - lib/ai/local-settings.ts: isolated persistence for provider/model/base-URL choices. Deliberately NOT folded into stores/settings-store.ts, which has a hand-maintained export/import enumeration this prototype-scope state doesn't belong in yet. - Retrieval reuses this app's own already-built app/api/offline/search (encrypted SQLite/FTS5 mail index) as context when available, and degrades to unaugmented chat — not an error — when it 404s/503s (no index in this session, e.g. plain browser rather than Electron). Rewrote components/settings/ai-assistant-settings.tsx: provider picker, local runtime config (base URL, model list/refresh, test connection with a CORS-aware diagnostic per the concept doc's own note on the browser row), public BYOK config (base URL, model, key, client-side consent toggle), and a working Ask box. Verified: typecheck clean, lint clean, translations pass, production build succeeds. Live-tested against the real Ollama on this machine (confirmed running: qwen2.5:32b, gemma4, deepseek-r1, llama3.2, hermes3, qwen3) via a local server + demo-mode session — admin flag round-trips correctly, the pane renders both provider options, and the CORS-diagnostic path fires correctly on a real (if here environment-sandboxed, not Ollama-side) connection failure. Full success end-to-end still wants a real, unsandboxed browser tab against this Mac's loopback to close out.
This commit is contained in:
Bernd Rodler
@@ -0,0 +1,26 @@
|
||||
// Client-held storage for the user's own public-provider API key (BYOK).
|
||||
//
|
||||
// Decision 2026-08-05 (reverses docs/AI-ASSISTANT-CONCEPT.md decision #1's
|
||||
// server-side-custody design): the user brings and holds their own key,
|
||||
// client-side, not VNC. This is the same custody model as
|
||||
// vncmail-native's lib/ai-key-store.ts (expo-secure-store there; this repo
|
||||
// has no OS keychain access from a browser tab, so localStorage is the
|
||||
// honest equivalent here — plain, not hidden behind a false sense of
|
||||
// "secure storage"). A fuller Paperclip-style key-management UI (multiple
|
||||
// providers, masking, rotation) is good follow-up work, not built tonight.
|
||||
const KEY_PREFIX = 'vncmail:ai:key:';
|
||||
|
||||
export function getAiApiKey(provider: 'public'): string | null {
|
||||
if (typeof window === 'undefined') return null;
|
||||
return window.localStorage.getItem(KEY_PREFIX + provider);
|
||||
}
|
||||
|
||||
export function setAiApiKey(provider: 'public', key: string): void {
|
||||
if (typeof window === 'undefined') return;
|
||||
window.localStorage.setItem(KEY_PREFIX + provider, key);
|
||||
}
|
||||
|
||||
export function clearAiApiKey(provider: 'public'): void {
|
||||
if (typeof window === 'undefined') return;
|
||||
window.localStorage.removeItem(KEY_PREFIX + provider);
|
||||
}
|
||||
@@ -0,0 +1,187 @@
|
||||
// The AI assistant's wire client — mirrors vncmail-native's src/api/ai.ts
|
||||
// (same prototype scope: local Ollama + BYOK public, no VNC-hosted `server`
|
||||
// class, no streaming) so the two clients stay in lockstep. Runs entirely
|
||||
// client-side (`'use client'` callers only) — a direct loopback/provider
|
||||
// fetch, matching docs/AI-ASSISTANT-CONCEPT.md §2's "local"/"public" rows,
|
||||
// not proxied through this app's own Next.js server. That distinction
|
||||
// matters once this app is hosted remotely: a server-side proxy would reach
|
||||
// the *server's* loopback, not the user's own laptop running Ollama.
|
||||
|
||||
export interface ChatMessage {
|
||||
role: 'system' | 'user' | 'assistant';
|
||||
content: string;
|
||||
}
|
||||
|
||||
// ── Local: Ollama's native API, not the OpenAI-compat shim — one fewer path
|
||||
// assumption (no "/v1" prefix to guess at) for a runtime this code talks to directly. ──
|
||||
|
||||
interface OllamaTagsResponse {
|
||||
models?: Array<{ name: string }>;
|
||||
}
|
||||
|
||||
interface OllamaChatResponse {
|
||||
message?: { content?: string };
|
||||
}
|
||||
|
||||
export async function listLocalModels(baseUrl: string): Promise<string[]> {
|
||||
const res = await fetch(`${baseUrl.replace(/\/+$/, '')}/api/tags`);
|
||||
if (!res.ok) throw new Error(`Ollama returned ${res.status}`);
|
||||
const body = (await res.json()) as OllamaTagsResponse;
|
||||
return (body.models ?? []).map((m) => m.name).filter(Boolean);
|
||||
}
|
||||
|
||||
/**
|
||||
* Diagnoses the specific failure rather than a generic "connection failed" —
|
||||
* docs/AI-ASSISTANT-CONCEPT.md §3 calls this out explicitly for the browser
|
||||
* row: a CORS rejection (the runtime is up but refused this page's origin)
|
||||
* looks identical to "nothing is listening" unless told apart. `fetch`
|
||||
* itself can't distinguish them (a CORS failure and a connection refusal
|
||||
* both surface as `TypeError: Failed to fetch`), so this only upgrades the
|
||||
* message when the caller can tell us there's a live page origin to name.
|
||||
*/
|
||||
export async function testLocalConnection(
|
||||
baseUrl: string,
|
||||
): Promise<{ ok: boolean; error?: string }> {
|
||||
try {
|
||||
await listLocalModels(baseUrl);
|
||||
return { ok: true };
|
||||
} catch (err) {
|
||||
const origin = typeof window !== 'undefined' ? window.location.origin : null;
|
||||
const hint = origin
|
||||
? ` Reachable in principle, but if Ollama is actually running, it likely refused this page's origin (${origin}) — start it with OLLAMA_ORIGINS=${origin}.`
|
||||
: '';
|
||||
return {
|
||||
ok: false,
|
||||
error: (err instanceof Error ? err.message : String(err)) + hint,
|
||||
};
|
||||
}
|
||||
}
|
||||
|
||||
export async function chatLocal(
|
||||
baseUrl: string,
|
||||
model: string,
|
||||
messages: ChatMessage[],
|
||||
): Promise<string> {
|
||||
const res = await fetch(`${baseUrl.replace(/\/+$/, '')}/api/chat`, {
|
||||
method: 'POST',
|
||||
headers: { 'Content-Type': 'application/json' },
|
||||
body: JSON.stringify({ model, messages, stream: false }),
|
||||
});
|
||||
if (!res.ok) throw new Error(`Ollama returned ${res.status}`);
|
||||
const body = (await res.json()) as OllamaChatResponse;
|
||||
const content = body.message?.content;
|
||||
if (!content) throw new Error('Ollama returned no message content');
|
||||
return content;
|
||||
}
|
||||
|
||||
// ── Public: OpenAI-compatible chat-completions. OpenRouter by default, but any
|
||||
// endpoint speaking this shape works unmodified (self-hosted vLLM, LiteLLM, etc). ──
|
||||
|
||||
interface OpenAiChatResponse {
|
||||
choices?: Array<{ message?: { content?: string } }>;
|
||||
}
|
||||
|
||||
export async function chatPublic(
|
||||
baseUrl: string,
|
||||
apiKey: string,
|
||||
model: string,
|
||||
messages: ChatMessage[],
|
||||
): Promise<string> {
|
||||
const res = await fetch(`${baseUrl.replace(/\/+$/, '')}/chat/completions`, {
|
||||
method: 'POST',
|
||||
headers: {
|
||||
'Content-Type': 'application/json',
|
||||
Authorization: `Bearer ${apiKey}`,
|
||||
},
|
||||
body: JSON.stringify({ model, messages }),
|
||||
});
|
||||
if (!res.ok) throw new Error(`Provider returned ${res.status}`);
|
||||
const body = (await res.json()) as OpenAiChatResponse;
|
||||
const content = body.choices?.[0]?.message?.content;
|
||||
if (!content) throw new Error('Provider returned no message content');
|
||||
return content;
|
||||
}
|
||||
|
||||
// ── Retrieval: this app's own already-built offline search surface
|
||||
// (app/api/offline/search/route.ts), not a client-side index — the
|
||||
// encrypted SQLite/FTS5 store it reads only exists in Electron's main
|
||||
// process. A 404/503 there means "no index in this session", not an error:
|
||||
// degrade to an unaugmented chat rather than fail the question. ──
|
||||
|
||||
export interface AskSource {
|
||||
id: string;
|
||||
subject: string;
|
||||
}
|
||||
|
||||
export interface AskResult {
|
||||
answer: string;
|
||||
sources: AskSource[];
|
||||
/** True when the question was answered without any retrieved context. */
|
||||
unaugmented: boolean;
|
||||
}
|
||||
|
||||
interface OfflineSearchHit {
|
||||
id: string;
|
||||
title: string;
|
||||
snippet?: string;
|
||||
}
|
||||
|
||||
interface OfflineSearchResponse {
|
||||
ok: true;
|
||||
hits: OfflineSearchHit[];
|
||||
contextBlock: string;
|
||||
}
|
||||
|
||||
async function retrieveContext(question: string): Promise<OfflineSearchResponse | null> {
|
||||
const res = await fetch(`/api/offline/search?q=${encodeURIComponent(question)}&limit=6`);
|
||||
if (!res.ok) return null; // 404 (no index configured) or 503 (unavailable this session) — both mean "no retrieval", not an error
|
||||
const body = (await res.json()) as OfflineSearchResponse;
|
||||
return body.ok ? body : null;
|
||||
}
|
||||
|
||||
export function buildPrompt(question: string, contextBlock: string): ChatMessage[] {
|
||||
return [
|
||||
{
|
||||
role: 'system',
|
||||
content:
|
||||
"You answer questions about the user's email using only the numbered excerpts " +
|
||||
'below as context. Cite sources by their number in brackets, e.g. [1]. If the ' +
|
||||
"excerpts don't contain the answer, say so plainly rather than guessing.",
|
||||
},
|
||||
{ role: 'user', content: `${contextBlock}\n\nQuestion: ${question}` },
|
||||
];
|
||||
}
|
||||
|
||||
export interface AskConfig {
|
||||
provider: 'local' | 'public';
|
||||
localBaseUrl: string;
|
||||
localModel: string | null;
|
||||
publicBaseUrl: string;
|
||||
publicModel: string;
|
||||
publicApiKey: string | null;
|
||||
}
|
||||
|
||||
export async function askMail(question: string, config: AskConfig): Promise<AskResult> {
|
||||
if (config.provider === 'local' && !config.localModel) {
|
||||
throw new Error('No local model selected');
|
||||
}
|
||||
if (config.provider === 'public' && !config.publicApiKey) {
|
||||
throw new Error('No public API key saved');
|
||||
}
|
||||
|
||||
const retrieved = await retrieveContext(question);
|
||||
const messages = retrieved
|
||||
? buildPrompt(question, retrieved.contextBlock)
|
||||
: [{ role: 'user' as const, content: question }];
|
||||
|
||||
const answer =
|
||||
config.provider === 'public'
|
||||
? await chatPublic(config.publicBaseUrl, config.publicApiKey as string, config.publicModel, messages)
|
||||
: await chatLocal(config.localBaseUrl, config.localModel as string, messages);
|
||||
|
||||
return {
|
||||
answer,
|
||||
sources: (retrieved?.hits ?? []).map((h) => ({ id: h.id, subject: h.title })),
|
||||
unaugmented: !retrieved,
|
||||
};
|
||||
}
|
||||
@@ -0,0 +1,43 @@
|
||||
// Small, isolated persistence for AI Assistant settings — deliberately NOT
|
||||
// folded into stores/settings-store.ts tonight. That store's export/import
|
||||
// feature enumerates every field by hand; this is prototype-scope UI state
|
||||
// (docs/AI-ASSISTANT-CONCEPT.md §12's P0/P5), and migrating it into the
|
||||
// shared store belongs with whichever phase makes these settings real
|
||||
// product config rather than a local-AI test harness.
|
||||
export type AiProvider = 'local' | 'public';
|
||||
|
||||
export interface AiLocalSettings {
|
||||
provider: AiProvider | null;
|
||||
localBaseUrl: string;
|
||||
localModel: string | null;
|
||||
publicBaseUrl: string;
|
||||
publicModel: string;
|
||||
publicConsentAccepted: boolean;
|
||||
}
|
||||
|
||||
const STORAGE_KEY = 'vncmail:ai:settings';
|
||||
|
||||
export const DEFAULT_AI_SETTINGS: AiLocalSettings = {
|
||||
provider: null,
|
||||
localBaseUrl: 'http://127.0.0.1:11434',
|
||||
localModel: null,
|
||||
publicBaseUrl: 'https://openrouter.ai/api/v1',
|
||||
publicModel: '',
|
||||
publicConsentAccepted: false,
|
||||
};
|
||||
|
||||
export function loadAiSettings(): AiLocalSettings {
|
||||
if (typeof window === 'undefined') return { ...DEFAULT_AI_SETTINGS };
|
||||
try {
|
||||
const raw = window.localStorage.getItem(STORAGE_KEY);
|
||||
if (!raw) return { ...DEFAULT_AI_SETTINGS };
|
||||
return { ...DEFAULT_AI_SETTINGS, ...JSON.parse(raw) };
|
||||
} catch {
|
||||
return { ...DEFAULT_AI_SETTINGS };
|
||||
}
|
||||
}
|
||||
|
||||
export function saveAiSettings(settings: AiLocalSettings): void {
|
||||
if (typeof window === 'undefined') return;
|
||||
window.localStorage.setItem(STORAGE_KEY, JSON.stringify(settings));
|
||||
}
|
||||
+16
-5
@@ -1,9 +1,20 @@
|
||||
// Shared client/server contract for the AI Assistant feature.
|
||||
// docs/AI-ASSISTANT-CONCEPT.md §9 (entitlement), §11 (client shape), §12 (P0).
|
||||
//
|
||||
// P0 scope only: this file defines the schema so it never needs a breaking
|
||||
// migration later (decision #4 — entitlement from day one, cheap now). No
|
||||
// provider class is implemented behind it yet; see the doc's phase table.
|
||||
// This file defines the schema so it never needs a breaking migration later
|
||||
// (decision #4 — entitlement from day one, cheap now).
|
||||
//
|
||||
// Decisions 2026-08-05 evening simplify the doc's original P1/P2 sequencing
|
||||
// for now — local-first, nothing metered yet:
|
||||
// - `local` ships free, always available, no entitlement check at all.
|
||||
// - `public` is available too, but explicitly UNMONITORED for the moment
|
||||
// (no seats, no metering, no consent-record backend — §7.3/§9/§10 are
|
||||
// not built yet). The client-side "this leaves the organisation"
|
||||
// acknowledgement still shows (cheap, honest), it just isn't
|
||||
// server-enforced yet.
|
||||
// - `server` (VNC-hosted, EU/CH) isn't wired up client-side yet — infra is
|
||||
// "this MacBook tonight, the dev k8s cluster tomorrow" per that
|
||||
// decision, sequenced after `local` rather than before it.
|
||||
|
||||
export type AiClass = 'local' | 'server' | 'public';
|
||||
|
||||
@@ -25,10 +36,10 @@ export interface AiPolicy {
|
||||
}
|
||||
|
||||
export const DEFAULT_AI_ENTITLEMENT: AiEntitlement = {
|
||||
licensed: false,
|
||||
licensed: true,
|
||||
subject: 'tenant',
|
||||
tier: 'base',
|
||||
classes: [],
|
||||
classes: ['local', 'public'],
|
||||
expiresAt: null,
|
||||
graceUntil: null,
|
||||
};
|
||||
|
||||
Reference in New Issue
Block a user