Release 1.11.0: Events Tracking & Specialized Code Embeddings

We’re pleased to announce Remembrances MCP 1.11.0, bringing two powerful new capabilities: a comprehensive Events & Logs system for temporal tracking and Dual Code Embeddings for optimized code search.

📅 Events & Logs System

Track activities, conversations, logs, and milestones with the new Events system. Unlike regular memories, events are designed for time-ordered data with powerful temporal queries.

What Makes Events Special?

Hybrid Search: Events combine vector similarity with BM25 text search. Search by meaning (“authentication issues”) and get results ranked by both semantic relevance and keyword matching.

Time-Based Queries: Find events from the last 24 hours, last week, or within specific date ranges. Perfect for tracking what happened when.

Subject Organization: Categorize events by topic using subject patterns like conversation:topic, error:type, or milestone:name.

Use Cases

📝 Conversation Memory: Track important discussions across multiple sessions:

save_event({
  "user_id": "project-alpha",
  "subject": "conversation:sprint-planning",
  "content": "Team agreed to prioritize auth improvements. MVP deadline set for March 15."
})

⚠️ Error Tracking: Log and search through incidents:

save_event({
  "user_id": "api-service",
  "subject": "error:database",
  "content": "Connection pool exhausted. Increased pool size from 10 to 25."
})

🚀 Milestone Tracking: Mark and find achievements:

save_event({
  "user_id": "product",
  "subject": "milestone:release",
  "content": "Version 2.0 launched with dark mode and multi-language support."
})

Powerful Search

Find events with combined filters:

search_events({
  "user_id": "api-service",
  "subject": "error:database",
  "query": "connection timeout",
  "last_days": 7,
  "limit": 20
})

This finds database errors from the last week that are semantically related to connection timeouts.

🧠 Dual Code Embeddings

Generic text embedding models work well for natural language, but code has different patterns and semantics. Version 1.11.0 introduces support for specialized code embedding models.

Why Specialized Models?

Code-specific embedding models like CodeRankEmbed or Jina Code Embeddings are trained on source code and understand:

• Programming language syntax and patterns
• Code semantics and relationships
• Natural language to code mapping

This translates to better results when searching code semantically.

How to Configure

Use a dedicated model for code indexing while keeping your general model for text:

GGUF (Local):

# Main model for text
gguf-model-path: "./nomic-embed-text-v1.5.Q4_K_M.gguf"

# Code-specific model
code-gguf-model-path: "./coderankembed.Q4_K_M.gguf"

Ollama:

ollama-model: "nomic-embed-text"
code-ollama-model: "jina/jina-embeddings-v3"

OpenAI:

openai-model: "text-embedding-3-small"
code-openai-model: "text-embedding-3-large"

Automatic Fallback

If you don’t configure a code-specific model, Remembrances uses your default embedding model for everything. Upgrade at your own pace!

Recommended Models

Getting Started

Upgrade

Download from GitHub Releases and replace your existing binary.

Try Events

Start tracking with:

save_event({
  "user_id": "my-project",
  "subject": "log:session",
  "content": "Started working on the new dashboard feature."
})

Search later:

search_events({
  "user_id": "my-project",
  "query": "dashboard feature",
  "last_days": 30
})

Configure Code Embeddings

Add to your config.yaml:

code-gguf-model-path: "./coderankembed.Q4_K_M.gguf"

Then re-index your projects to benefit from code-optimized embeddings.

What’s Next

We continue to enhance Remembrances MCP with:

• More event search capabilities
• Additional code embedding model support
• Performance improvements for high-volume event logging

Thank you for your continued support and feedback!

Questions or feedback? Open an issue on GitHub!

• ←Previous
• Next→