app.provider-ranking-mcp

@lssm-tech/app.provider-ranking-mcp

MCP server that exposes provider-ranking operations and resources for benchmark ingestion, ranking refreshes, and leaderboard queries.

What It Does

• Hosts a ContractSpec MCP server over provider-ranking contracts and runtime registries.
• Exposes benchmark ingest, custom benchmark run, and ranking refresh operations.
• Publishes leaderboard and model-profile resources backed by the ranking store.
• Enriches incomplete ranking data with metadata from @lssm-tech/lib.ai-providers.

Running Locally

From packages/apps/provider-ranking-mcp:

• bun run dev
• bun run start
• bun run build

Usage

bun run dev

Architecture

• src/server.ts boots the MCP server and transport wiring.
• src/index.ts exposes the package entrypoints for reuse or embedding.
• The server composes provider-ranking contracts, module orchestration, and MCP resource/tool exposure.

Public Entry Points

• Exports . and ./server for embedding or direct server startup.
• Primary MCP surfaces are benchmark-ingest tools, ranking refresh tools, leaderboard resources, and model-profile resources.
• Export . resolves through ./dist/index.mjs.
• Export ./server resolves through ./dist/server.mjs.
• Export ./* resolves through ./*.

Local Commands

• bun run dev — tsdown --watch
• bun run start — node dist/server.js
• bun run build — tsdown
• bun run lint — bun run lint:fix
• bun run lint:check — biome check .
• bun run lint:fix — biome check --write --unsafe --only=nursery/useSortedClasses . && biome check --write .
• bun run clean — rm -rf dist

Recent Updates

• Replace eslint+prettier by biomejs to optimize speed.
• Add latest models and align defaults.
• Resolve lint, build, and type errors across nine packages.
• Add AI provider ranking system with ranking-driven model selection.

Notes

• The current implementation uses an in-memory store; restart the process to clear rankings and benchmark results.