cbcoutinho
Joined 10 years ago
Karma
632
Submissions
282
Member since
Oct 8
Recent submissions
Comment · 3 months ago
I have developed an open-source memory system for agents accessible over MCP, which makes it possible to access it via any coding agent locally (claude-code) or via mobile (Claude AI, Mistral AI, etc).
The primary storage mechanism is .md files stored as Notes in Nextcloud; however, since Nextcloud supports a rich ecosystem of apps such as documents, rss feeds, calendar, etc, your knowledge base can grow with you. All content can optionally be indexed and available via semantic search - powered by a Qdrant vectordb.
The biggest cost drivers of a system like this is the memory required to host the vectordb - I'm really curious how others are optimizing their knowledge base. Thanks OP for doing to work in summarizing these tools!
If you're interested either the MCP server or Nextcloud App frontend, please check out:
Comment · 4 months ago
I'm working on a semantic layer for Nextcloud: https://astrolabecloud.com
The service is composed of two open-source services, namely a Nextcloud app (Astrolabe) and backend (nextcloud-mcp-server). I use the service as an MCP server across a number of apps, and others use it primarily for semantic search over large numbers of documents.
Both are open source, and I'm working on a managed offering, completely based in the EU, for individuals/teams that already use Nextcloud and want to be able to use semantic search across some or all of their documents.
Essentially your data stays in Nextcloud, and the MCP server backend keeps a vectordb in sync to enable semantic queries over your content. The number of supported apps is growing, including:
- notes
- deck cards
- files
- news items (RSS feeds)
- cookbook recipes
- contacts & calendar
And I'm adding support for other apps as I go.
Comment · 7 months ago
Comment · 8 months ago
MCP Servers are not simple wrappers around REST APIs, they can do much more and we will see more advanced use-cases surrounding MCP as MCP clients continually improve their conformance to the spec. Just wrapping REST APIs may be what MCP Servers are now, but that's just because MCP clients still need to catch up in supporting these more advanced features.
MCP Sampling (with tool calling) is one such example, where MCP servers will be able to generate responses using the MCP client's LLM - not requiring an API key themselves - including calling other MCP tools within the "sampling loop", and without corrupting the primary agent's context. This will generate an explosion in advanced MCP servers, although neither Claude Code, Gemini CLI, nor Opencode support this feature (yet).
Once MCP sampling becomes widely supported the pendulum will swing back and we'll see the boundary between MCP tools, async workflows, and generative responses begin to blur.
Comment · 9 months ago
The Nextcloud MCP Server [0] supports Qdrant as a vectordb to store embeddings and provide semantic search across your personal documents. This enables any LLM & MCP client (e.g. claude code) into a RAG system that you can use to chat with your files.
For local deployments, Qdrant supports storing embeddings in memory as well as in a local directory (similar to sqlite) - for larger deployments Qdrant supports running as a standalone service/sidecar and can be made available over the network.
Comment · 10 months ago
Great points. I had at least three of your example scenarios occur between my parents and I within the last couple days.
The more time I spend online the more I'm convinced I have never had a unique experience.
Comment · 10 months ago
Thanks for bringing up the point about power-of-attorney, I'll have to dive into that as well.
I dread the day I have to get more involved in their healthcare from afar, precisely because of the technology gap. The money grab from big-pharma is going to unrelenting
Comment · 10 months ago
I've been running openSUSE tumbleweed myself for years, and recommend Linux to like-minded power users. OP is preaching to the choir.
How do you all deal with (extended) family? This Christmas I spent time with my parents and the topic of Windows 11 came up again with all of its associated dark patterns.
What do you all do to help them out of this madness? Is Ubuntu/Fedora/etc the best option for seniors? My dad's entire career was in Silicon Valley 1.0 where Excel/Outlook was his bread and butter and feels married to Windows, but ever since leaving the workforce those skills are more of a hindrance than an asset.
Now that he's retired, he still uses Excel to plan vacations for example, but Windows is riddled with this BS and I am powerless to help him navigate this anti-consumer behavior. It's incredible that Microsoft is shooting their most loyal customers in the foot with this BS.
Do you all help your parents remotely? What kind of issues do you run into being your parents IT support?
Comment · 11 months ago
This is why I built the Nextcloud MCP server, so that you can talk with your own data. Obviously this is Nextcloud-specific, but if you're using it already then this is possible now.
https://github.com/cbcoutinho/nextcloud-mcp-server
The default MCP server deployment supports simple CRUD operations on your data, but if you enable vector search the MCP server will begin embedding docs/notes/etc. Currently ollama and openai are supporting embeddings providers.
The MCP server then exposes tools you can use to search your docs based on semantic search and/or bm25 (via qdrant fusion) as well as generate responses using MCP sampling.
Importantly, rather than generating responses itself, the server relies on MCP sampling so that you can use any LLM/MCP client. This MCP sampling/RAG pattern is extremely powerful and it wouldn't surprise me if there was something open source that generalizes this across other data sources.
Comment · 1 year ago
This is related to the removal of 'de minimis' rule that exempts parcels under $800 to ship duty free. This has caused some European postal services to stop/delay shipping some packages to the US [0]. The Dutch postal service for instance has stopped shipping to the US [1]
[0]: https://apnews.com/article/us-tariffs-goods-services-suspens...
Comment · 1 year ago
From the paper:
> Most language models face a fundamental tradeoff where powerful capabilities require substantial computational resources. We shatter this constraint with Jan-nano, a 4B parameter language model that redefines efficiency through radical specialization: instead of trying to know everything, it masters the art of finding anything instantly. Fine-tuned from Qwen3-4B using our novel multi-stage Reinforcement Learning with Verifiable Rewards (RLVR) system that completely eliminates reliance on next token prediction training (SFT), Jan-nano achieves 83.2% on SimpleQA benchmark with MCP integration while running on consumer hardware. With 128K context length, Jan-nano proves that intelligence isn't about scale, it's about strategy.
> For our MCP evaluation, we used mcp-server-serper which provides google search and scrape tools
[ my public key: https://keybase.io/cbcoutinho; my proof: https://keybase.io/cbcoutinho/sigs/9rNeyVIv5DF69DmgsU-p7NOhsLzg9aA3Azo6u-q_3vI ]