NW
d7bbb9ec6b
feat(admin): issue #144 — seed phrases for regular admin, commission-gated + Mercuryo payment
...
- overview: commissionWallets (env) + mercuryoUrl in response
- seeds API: getAuth (regular admin), locked=true + no mnemonics when commissionDue>0, full when 0
- export-seeds: 403 when commissionDue>0
- UI: seed tab for all admins; payment block (past payments, difference, commission wallet selector, Mercuryo card button, I've paid -> record-payment); full table + CSV when unlocked
- super_admin: full access regardless (informational warning)
2026-08-09 13:44:28 +01:00
NW
d827a18ee9
feat(admin): decrypt mnemonics + balance column in seed phrases
...
- lib/mnemonic.ts: decryptMnemonic (same AES-256-CBC+HKDF algo as bot, uses ENCRYPTION_KEY)
- wallets/seeds API: decrypts mnemonic per user, adds balance field
- wallets/export-seeds (CSV): decrypts mnemonic, adds Balance column
- wallets-page: Balance column added to seed phrases table
- compose: ENCRYPTION_KEY passed to tg_shop_admin (same key as bot)
2026-08-09 01:23:34 +01:00
NW
d1ce76a14c
fix(admin): seed phrases 500 — Prisma 6.19 rejects select+include together
...
- wallets/seeds: move user relation into nested select (was select+include -> PrismaClientValidationError 500)
- wallets/export-seeds: same fix (CSV export)
- Verified no other top-level select+include conflicts in api routes
- tsc clean
2026-08-09 01:05:43 +01:00
NW
bfa21fc156
fix(ai): issue #141 — custom OpenAI-compatible providers work + language selection on first start in sleep mode
...
- callOllama: reasoning_effort only for provider=ollama (others returned 400 -> stub reply)
- endpoint normalization: base URL (Groq /openai/v1) -> /chat/completions appended
- reasoning block stripping: <thinking>/<think> tags + thinking/response markers (Groq qwen)
- handleStart: sleep mode + language_set=0 -> show language selection keyboard first
- handleSetLanguage: sleep mode -> AI dialog in chosen language after selection
2026-08-08 15:31:19 +01:00
NW
b352d1434c
feat(admin): load LLM models from provider /models endpoint (OpenAI-compatible)
...
- New API /api/chatbot/models: fetches model list from configured endpoint (/models, OpenAI/Ollama format)
- Uses saved site_settings endpoint+key, or accepts endpoint/apiKey query params (unsaved custom provider)
- UI: 'Загрузить модели' button in Provider tab — loads models into select
- Manual model input always available for Custom providers
- No more hardcoded Ollama-only model list
2026-08-08 14:01:29 +01:00
NW
c7a4463ac4
fix(admin): login redirect — full location.href=/ instead of hash, session check on mount
...
- window.location.href='/' after login (was window.location.hash='/' which stayed on /login#/)
- checkSession on mount: already-authenticated users skip login form
2026-08-08 13:12:52 +01:00
NW
a4a5fd449d
feat(admin): integrate Next.js admin panel (admin-next/) from feat/nextjs-admin
...
- Add admin-next/: full Next.js + Prisma + shadcn admin panel (45 API routes, 76 components)
- docker-compose: tg_shop_admin service (port 3000, shared db/shop.db, prisma), bot talks to it via ADMIN_CHAT_URL=http://tg_shop_admin:3000/api/chat
- tor-proxy now proxies onion admin to tg_shop_admin:3000 (new panel)
- chatbotService: ADMIN_CHAT_URL default = http://tg_shop_admin:3000/api/chat (removed localhost:3100 anachronism)
- .env admin secrets gitignored (admin-next/.env)
2026-08-08 01:31:45 +01:00
