Skip to main content
Glama
Sign In
deenesjakoruzh
yuntu-mcp-server

yuntu-mcp-server

by Drq112112Hjx
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Remote

yuntu-mcp-server

Yuntu Zhihui Compute Scheduling MCP Server (hosted version), implemented in Python FastMCP, with functionality aligned to the three tools of the Go version MCP Server:

  • chat_completion: calls the LLM for conversation, channel routing + failover, returns content and token usage (deducts actual tokens)

  • get_balance: queries the current tenant's balance (token_balance / balance)

  • list_models: lists currently available models

Note: this version does not yet include "price/discount" logic, only "balance > 0 check + deduct actual token count", to get the hosted pipeline working first. Pricing/discount will be integrated later.

How to Run

Local Development / Unit Testing

python -m venv .venv
.venv/Scripts/python -m pip install -e .
.venv/Scripts/python tests/test_smoke.py        # 逻辑冒烟测试（SQLite）

Directly Launched via uvx (after PyPI release, ModelScope hosting)

First publish to PyPI, then configure runtime parameters in environment variables, and uvx can launch it:

uvx yuntu-mcp-server

HTTP Mode (optional, aligned with Go version /mcp path)

MCP_TRANSPORT=http MCP_PORT=8081 uvx yuntu-mcp-server
# 监听 http://0.0.0.0:8081/mcp （Streamable HTTP）

Related MCP server: Synap MCP Server

Environment Variables

Variable

Required

Description

DB_DSN

Yes

Database connection string. MySQL: user:pass@tcp(host:3306)/db?charset=utf8mb4&parseTime=True&loc=Local; locally you can also use sqlite:///xx.db

API_KEY_ENCRYPT_SECRET

Yes

API Key encryption secret (≥16 characters), must match the Yuntu Zhihui main platform (used to decrypt channel API keys and look up tenants)

MCP_API_KEY

No

The current tenant's API Key (stdio has no HTTP Header, so the tenant is identified via runtime environment variable injection; when missing, get_balance/list_models return empty and chat_completion is rejected). Also compatible with reading YUNTU_API_KEY

MCP_TRANSPORT

No

stdio (default, uvx hosting uses stdio) or http

MCP_HOST / MCP_PORT

No

HTTP mode listen address/port, default 0.0.0.0:8081

Authentication Semantics

  • Consistent with the Go version: the api_keys table is looked up by key_hash (plaintext SHA-256), plaintext is never stored.

  • Validate enabled and expires_at, use tenant_id as the subject for billing/capability decisions.

  • The channel upstream_configs.api_key is decrypted with the same encryption secret before calling upstream (if decryption fails, it is treated as plaintext).

Publishing to PyPI

python -m pip install build
python -m build                              # 生成 dist/ 下的 wheel 与 sdist
python -m twine upload dist/*                # 需配置 PyPI 令牌

Verify after publishing:

DB_DSN="user:pass@tcp(host:3306)/db?charset=utf8mb4" \
API_KEY_ENCRYPT_SECRET="<与主平台一致>" \
MCP_API_KEY="<租户Key>" \
uvx yuntu-mcp-server

Available Tools

3 tools
chat_completionB

调用大模型进行对话，从渠道管理中选择可用渠道并自动扣除 Token。

Inputs: messages (required), model, temperature. Returns: result

get_balanceA

查询当前租户的 Token 余额（需配置 MCP_API_KEY 识别租户）。

Inputs: no named parameters. Returns: result

list_modelsA

列出当前可用的所有模型列表。

Inputs: no named parameters. Returns: result

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Deploy Server
    One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

    TDQS

    A3.9/5.0

    Scored across 3 tools

    Disambiguation5/5

    Each tool has a clearly distinct purpose: chat_completion handles LLM interactions, get_balance checks token quota, and list_models shows available models. No overlap or ambiguity between them.

    Naming Consistency5/5

    All three tool names follow a consistent verb_noun pattern in snake_case: chat_completion, get_balance, list_models. The naming is uniform and predictable.

    Tool Count4/5

    With only 3 tools, the surface is minimal but appropriate for a straightforward LLM gateway. It covers the essential actions (use model, check balance, list models) without unnecessary bloat, though a few more (e.g., channel management) could be expected.

    Completeness4/5

    The tool set covers the core user-facing operations for an LLM gateway: chatting, checking balance, and discovering models. It lacks administrative features like managing channels, but for the apparent consumer-focused scope, these are reasonable gaps.

    Maintenance

    ActivityStale
    ResponsivenessNo recent community issues

    Resources

    Related MCP Connectors

    Related MCP Servers