feat: initialize agent and claude skill libraries with comprehensive knowledge bases, workflow templates, and implementation artifacts.

.serena/memories/project_overview.md

@@ -0,0 +1,21 @@
+# SurfSense - Project Overview
+
+## Purpose
+Open-source alternative to NotebookLM — personal knowledge base with AI chat, 25+ external connectors (Google Drive, Notion, Jira, Slack...), real-time multiplayer, desktop app, podcast/video generation.
+
+## Tech Stack
+- **Backend**: Python 3.12, FastAPI, Celery (Redis broker), PostgreSQL (pgvector), Alembic, LiteLLM, LangGraph, uv package manager
+- **Frontend**: Next.js 16 (Turbopack), React 19, TypeScript, Tailwind v4, Jotai, @rocicorp/zero (real-time sync), pnpm
+- **Real-time**: zero-cache (rocicorp/zero:0.26.2) → Postgres logical replication
+- **Services (Docker)**: PostgreSQL pgvector, Redis, SearXNG, pgAdmin, zero-cache
+- **Desktop**: Electron (surfsense_desktop/)
+- **Browser Extension**: surfsense_browser_extension/
+
+## Architecture
+```
+surfsense_backend/   - FastAPI + Celery workers
+surfsense_web/       - Next.js frontend
+surfsense_desktop/   - Electron desktop app
+surfsense_browser_extension/ - Browser extension
+docker/              - docker-compose.dev.yml & .env
+```

.serena/memories/suggested_commands.md

@@ -0,0 +1,73 @@
+# SurfSense - Suggested Commands
+
+## Docker Services (Infrastructure)
+```bash
+# Start all infra services (db + pgadmin)
+cd /Users/luisphan/Documents/GitHub/SurfSense
+docker compose -f docker/docker-compose.dev.yml --env-file docker/.env up -d db pgadmin
+
+# Start zero-cache standalone (BE & FE run local)
+docker run -d \
+  --name surfsense-zero-cache \
+  --network surfsense-dev_default \
+  --add-host "host.docker.internal:host-gateway" \
+  -p 4848:4848 \
+  -v surfsense-dev-zero-cache:/data \
+  -e ZERO_UPSTREAM_DB="postgresql://postgres:postgres@surfsense-dev-db-1:5432/surfsense?sslmode=disable" \
+  -e ZERO_CVR_DB="postgresql://postgres:postgres@surfsense-dev-db-1:5432/surfsense?sslmode=disable" \
+  -e ZERO_CHANGE_DB="postgresql://postgres:postgres@surfsense-dev-db-1:5432/surfsense?sslmode=disable" \
+  -e ZERO_REPLICA_FILE="/data/zero.db" \
+  -e ZERO_ADMIN_PASSWORD="surfsense-zero-admin" \
+  -e ZERO_APP_PUBLICATIONS="zero_publication" \
+  -e ZERO_NUM_SYNC_WORKERS="4" \
+  -e ZERO_UPSTREAM_MAX_CONNS="20" \
+  -e ZERO_CVR_MAX_CONNS="30" \
+  -e ZERO_QUERY_URL="http://host.docker.internal:3000/api/zero/query" \
+  -e ZERO_MUTATE_URL="http://host.docker.internal:3000/api/zero/mutate" \
+  rocicorp/zero:0.26.2
+
+# SearXNG: reuse mrholmes-searxng trên port 8888
+# Redis: reuse redis-server local trên localhost:6379/1
+```
+
+## Backend (FastAPI)
+```bash
+cd /Users/luisphan/Documents/GitHub/SurfSense/surfsense_backend
+uv sync                        # install deps
+uv run alembic upgrade head    # run migrations
+uv run python main.py --reload # start dev server on port 8001
+# OR Celery worker:
+uv run celery -A app.celery_app worker -Q surfsense --loglevel=info
+```
+
+## Frontend (Next.js)
+```bash
+cd /Users/luisphan/Documents/GitHub/SurfSense/surfsense_web
+pnpm install
+pnpm dev                       # http://localhost:3000
+
+# DB commands (drizzle)
+pnpm db:generate
+pnpm db:migrate
+pnpm db:studio                 # Drizzle Studio UI
+```
+
+## Ports Summary
+| Service       | Port  | Notes |
+|---------------|-------|-------|
+| PostgreSQL    | 5432  | docker |
+| pgAdmin       | 5050  | docker, http://localhost:5050 |
+| Redis (local) | 6379  | db=1 for SurfSense |
+| SearXNG       | 8888  | shared mrholmes-searxng |
+| zero-cache    | 4848  | docker standalone |
+| Backend       | 8001  | port 8000 used by chainlens |
+| Frontend      | 3000  | Next.js |
+
+## Health Checks
+```bash
+curl http://localhost:8001/health    # {"status":"ok"}
+curl http://localhost:3000           # HTML
+nc -z localhost 4848 && echo OK     # zero-cache
+redis-cli ping                       # PONG
+docker ps --filter "name=surfsense" # docker services
+```

.serena/memories/trollllm-integration.md

@@ -0,0 +1,45 @@
+# TrollLLM — Hướng dẫn tích hợp vào SurfSense
+
+## Base URL
+- OpenAI-compatible endpoint: `https://chat.trollllm.xyz/v1`
+- Anthropic-compatible endpoint: `https://chat.trollllm.xyz` (không có /v1)
+
+## Danh sách model chính xác (tên phải đúng 100%)
+| Model ID | Provider | Ghi chú |
+|---|---|---|
+| `claude-haiku-4.5` | Anthropic | Speed |
+| `claude-sonnet-4` | Anthropic | Balanced |
+| `claude-sonnet-4.5` | Anthropic | Balanced |
+| `claude-sonnet-4.6` | Anthropic | Balanced |
+| `claude-opus-4.5` | Anthropic | Reasoning |
+| `claude-opus-4.6` | Anthropic | Reasoning |
+| `gemini-3-flash-preview` | Google | Speed (**KHÔNG phải** gemini-3-flash) |
+| `gemini-3.1-pro-preview` | Google | Multimodal |
+| `gpt-5.2` | OpenAI | Reasoning |
+| `gpt-5.4` | OpenAI | Reasoning |
+| `gpt-5.2-codex` | OpenAI | Code |
+| `gpt-5.3-codex` | OpenAI | Code |
+
+## Cách add model vào SurfSense (đúng cách)
+
+### Cách 1 — Dùng Provider = OPENAI (khuyến nghị, dùng cho mọi model)
+- **LLM Provider**: `OPENAI`
+- **Model Name**: tên chính xác từ bảng trên (ví dụ `claude-sonnet-4.6`)
+- **API Key**: TrollLLM API key
+- **API Base URL**: `https://chat.trollllm.xyz/v1`
+
+### Cách 2 — Dùng Custom Provider (LiteLLM format)
+- **LLM Provider**: `Custom Provider`
+- **Custom Provider Name**: tùy ý (ví dụ `trollllm`)
+- **Model Name**: phải có prefix `openai/` → `openai/claude-sonnet-4.6`
+- **API Key**: TrollLLM API key
+- **API Base URL**: `https://chat.trollllm.xyz/v1`
+
+## Lỗi phổ biến
+1. **Tên model sai**: `gemini-3-flash` ❌ → phải là `gemini-3-flash-preview` ✅
+2. **Custom Provider thiếu prefix**: `gemini-3-flash-preview` ❌ → `openai/gemini-3-flash-preview` ✅
+3. **Base URL sai**: `https://trollllm.xyz/v1` ❌ → `https://chat.trollllm.xyz/v1` ✅
+
+## Lưu ý đặc biệt
+- TrollLLM yêu cầu `User-Agent` header để bypass Cloudflare, nhưng SurfSense/LiteLLM thường tự set header này.
+- Nếu dùng Anthropic SDK format: dùng `x-api-key` header thay vì `Authorization: Bearer`.