Skip to content

This release adds 3 notable features for engineering teams evaluating rollout.

Published 25d MCP Developer Tools
✓ No known CVEs patched
Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Topics

ai claude-code deepseek deepseek-v4 gemini-cli llm
+3 more
mcp mcp-server typescript

Summary

AI summary

Updates Quality, prefix, and deepseek-mcp.tahirl.com across a mixed release.

Full changelog

Added

  • New tool: deepseek_fim, Fill-in-the-Middle completion. Provide a prompt (prefix) and an optional suffix; the model completes the text in between. Built for code completion and content infilling. Runs against DeepSeek's Beta completions endpoint in non-thinking mode, with output capped at 4096 tokens. Supports model, max_tokens, temperature, and stop, with the same cache-aware cost tracking as deepseek_chat. On the npm/stdio server it also reuses the circuit breaker and model fallback (v4-flash <-> v4-pro). The deepseek-chat / deepseek-reasoner aliases resolve to v4-flash since FIM has no thinking mode.
  • deepseek_fim ships on both the npm/stdio server and the hosted worker endpoint (deepseek-mcp.tahirl.com).
  • fim capability listed on both models in the deepseek://models resource.

Quality

  • 296 tests (+16). Both surfaces verified live against the real DeepSeek V4 Beta API before release.

Full Changelog: https://github.com/arikusi/deepseek-mcp-server/compare/v2.0.0...v2.1.0

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Track arikusi/deepseek-mcp-server

Get notified when new releases ship.

Sign up free

About arikusi/deepseek-mcp-server

MCP server for DeepSeek AI with chat, reasoning, multi-turn sessions, function calling, thinking mode, and cost tracking.

All releases →

Related context

Earlier breaking changes

  • v1.8.0 Requires Node.js 20 or newer; Node 18 support dropped.

Beta — feedback welcome: [email protected]