Nymph

正體中文 | English

Your personal AI technical consultant
From Web-Tech-TW
Meticulously crafted — a multi-platform intelligent technical consultant powered by open-source AI

Nymph is a community-driven senior technical consultant agent service, operated and maintained by the Taiwan Web Technology Promotion Organization (Web Tech TW), providing developers with professional, in-depth, and contextual system architecture consulting, debugging guidance, and engineering experience retrieval.

---

Core Capabilities

• Senior technical architecture consulting: Professional advice and debugging diagnosis across frontend, backend, cloud architecture, networking, and security literacy.
• Community engineering knowledge retrieval: Instantly search years of accumulated technical discussions, architecture decision records, and real debugging notes from the community.
• Long-context multi-turn conversation: Context understanding and conversational memory, continuously tracking the problem thread across multiple rounds of discussion.
• Cross-platform, multi-channel support: Available through Discord and LINE, or integrated directly into your personal development environment via the Model Context Protocol (MCP).

---

How to Use Nymph

1. On Discord

• Invite the bot: Click Invite Nymph to your Discord server.
• Server channels: @Nymph in any channel of a server Nymph has joined, then ask your question.
• Direct messages: Send Nymph a DM for one-on-one technical consulting.

2. On LINE

• One-on-one consulting: Add the Nymph official account as a friend and chat directly.
• Group consulting: Invite Nymph into a LINE group and ask questions right in the group.

3. In your personal dev tools (via remote MCP)

Through the Model Context Protocol (MCP) standard, you can integrate Nymph as a remote technical consultant directly into your editor or AI client (such as Antigravity, Claude Code/Desktop, Codex, Cursor, etc.).

Step 1: Request an MCP connection token

1. Open the token management page of the Nymph service website (https://web-tech.tw/nymph/mcp/tokens).
2. Click "Sign in with Sara" and log in through the Taiwan Web Technology Promotion Organization's unified Sara identity system.
3. Enter a token label (e.g. Cursor IDE or My MacBook) and click "Issue new token".
4. Copy the generated token.

Step 2: Mount Nymph in your tool

Taking Antigravity as an example, add the Nymph server to your MCP configuration file (such as mcp_config.json or mcp.json):

{
  "mcpServers": {
    "nymph": {
      "type": "http",
      "url": "https://web-tech.tw/nymph/mcp",
      "headers": {
        "Authorization": "Bearer <YOUR_MCP_TOKEN>"
      }
    }
  }
}

Step 3: Invoke Nymph tools in your editor

Once connected, your AI assistant can directly invoke the following Nymph capabilities in conversation:

• consult_nymph_wisdom: Start an in-depth technical consultation, architecture analysis, or debugging diagnosis with Nymph.
• absorb_nymph_wisdom: Search and absorb the community's accumulated engineering knowledge base, architecture decisions, and solutions.
• my_nymph_impression: Retrieve the profile of this MCP connection token holder.

---

Example Consultations

• Architecture and technology selection

  "We are planning a large frontend-backend separated project. Please analyze the trade-offs between a micro-frontend architecture and a Monorepo approach in terms of maintenance cost and performance."

• Troubleshooting and error diagnosis

  "Compiling native Node.js modules inside a Docker container fails with a node-gyp error. What are the common causes and troubleshooting steps?"

• Community knowledge base lookup

  "Please search the community knowledge base for past best practices and decision records on OAuth2 and JWT token refresh mechanisms."

• Networking and security concepts

  "Please explain the difference between HTTP/2 multiplexing and HTTP/3's QUIC-based connections in packet-loss scenarios."

---

Self-hosting Nymph

Nymph is open-sourced under the MIT License — you are welcome to deploy your own instance.

1. Prerequisites

• Bun (>= 1.2)
• MongoDB
• An OpenAI-compatible API key
• As needed: a Discord bot token, LINE official account credentials

2. Install and Configure

bun install
cp .env.sample .env     # Fill in your keys and connection settings
cp mcp.sample.toml mcp.toml

Main environment variables:

Variable	Description
BASE_URL / HTTP_PORT	HTTP service URL and port (default: 3000)
MONGODB_URI	MongoDB connection string
OPENAI_API_KEY	OpenAI-compatible API key
OPENAI_BASE_URL	OpenAI-compatible base URL
OPENAI_MODEL	Model name
OPENAI_REASONING_EFFORT	Optional reasoning effort (e.g. low / medium / high; supported values depend on the provider and model)
DISCORD_BOT_TOKEN	Discord bot token
LINE_CHANNEL_ACCESS_TOKEN / LINE_CHANNEL_SECRET	LINE official account credentials
SARA_INTE_HOST / SARA_RECV_HOST	Sara unified identity system endpoints

3. Don't Have an LLM API Key?

No budget for an LLM API? No problem!

NVIDIA Build hosts 100+ models behind a free, OpenAI-compatible API. Sign up with an email, no credit card required, open any model page, and click Get API Key to receive a free nvapi- key.

The recommended model on NVIDIA NIM is openai/gpt-oss-20b, an open-weight model with strong multilingual reasoning and low latency — it is also the default in .env.sample:

OPENAI_API_KEY="nvapi-..."
OPENAI_BASE_URL="https://integrate.api.nvidia.com/v1"
OPENAI_MODEL="openai/gpt-oss-20b"

4. Run

# Development (hot reload)
bun run dev

# Production
bun run start

Or run it with Docker:

docker build -t nymph .
docker run --env-file .env -p 3000:3000 nymph

---

About Us

Nymph is maintained and promoted by the Taiwan Web Technology Promotion Organization, aiming to help community members exchange technical knowledge, pass on engineering experience, and grow together.

• Official website: Taiwan Web Technology Promotion Organization
• Install the bot: Discord bot | LINE official account

---

License

This project is licensed under the MIT License.