MCP and Duplicate Code Gen Issue Fix

🐛 Critical Bug Fix

Fixed Duplicate Code Generation

Natural language requests were generating code twice, causing:

• ⚠️ Unnecessary LLM API calls (doubled costs)
• ⚠️ High CPU usage
• ⚠️ Duplicate code blocks in output

Fixed: Removed duplicate code block in natural language processing path. Code generation now runs once per request as intended.

Impact

✅ Single code generation per request
✅ Reduced API costs
✅ Better performance
✅ Cleaner output

Note: Slash commands (/save, /run, /validate) were not affected.

🎛️ MCP Docs and Examples

Release 0.1.2

🎉 Highlights

MCP Integration - Full Model Context Protocol support with commands for connecting to external tools, APIs, and data sources.

Performance Improvements - Fast mode enabled by default for faster responses. RAG disabled by default (enable with /enable-rag for better code quality).

✨ New Features

• MCP Commands: /mcp-servers, /mcp-connect, /mcp-disconnect, /mcp-tools, /mcp-call, /mcp-resources, /mcp-read, /mcp-prompts, /mcp-prompt
• MCP Support: stdio, SSE, and WebSocket transports
• Welcome Screen: Now displays RAG and Fast Mode status with tips
• Code Generation Tips: Automatic suggestions to enable RAG for better quality

📚 Documentation

• Complete MCP integration guides
• Filesystem MCP assistant tutorial
• GitHub triage copilot tutorial
• Working examples and configuration templates

🐛 Bug Fixes

• Fixed /mcp-tools auto-connect
• Fixed color rendering in performance display
• Improved MCP error handling

🚀 Upgrade

pip install --upgrade dspy-code

Note: For better code quality, use /enable-rag and /fast-mode off (slower but more accurate).

🚄 v0.1.1: uv Support, Fast Mode and RAG Toggle

🎉 v0.1.1 — Faster, Smarter, Friendlier
This release has following features

🎥 Watch This Release in Action on YouTube

🧰 First-class venv & uv support, including project-root venv detection and uv-based install flows.
💻 Smarter DSPy code generation that suggests dspy.LM config matching your connected model.
🧠 RAG & speed controls via /fast-mode, /disable-rag, /enable-rag, and /status so you can trade quality vs speed.

🎉 DSPy Code 0.1.0 – First Public Beta Release

🧪 DSPy Code 0.1.0 – First Public Release

🎥 Watch This Release in Action on YouTube

DSPy Code is an AI-powered, interactive CLI for developing, optimizing, and understanding DSPy programs – essentially “Claude Code for DSPy”.

Key Features

• Interactive Development Workflow

  • Natural language → DSPy Signatures, Modules, and complete Programs.

• Model-Aware & Provider-Agnostic

  • Local models via Ollama (for example gpt-oss:120b).
  • Cloud models: OpenAI, Anthropic (Claude), Google Gemini.
  • New /model command for interactive selection; /connect for direct power-user control.
  • Generated programs include provider-aware dspy.LM(...) examples, and the CLI prints a copy-paste dspy.LM snippet that matches your connected model.

• GEPA Optimization & Evaluation

  • End-to-end GEPA workflows: /optimize, /optimize-start, /optimize-status, /optimize-resume, /optimize-cancel.
  • Metric support (Accuracy, F1, ROUGE, BLEU, etc.).
  • Clear guidance on cloud costs and hardware requirements (32 GB RAM recommended for heavy local runs).

• Project, RAG, and MCP Integration

  • /init, /project, /refresh-index, /index-status for project and codebase management.
  • Codebase RAG over your own DSPy project for intelligent Q&A.
  • Session management (/sessions, /session, /history, /clear) and export/import (/export, /import) for deployment.
  • Built-in MCP client to connect DSPy Code to external tools and data sources.

• Docs & Developer Experience

  • Improved docs but still needs further improvement for better developer experience