I'm trying to find a way to translate audio in a FOSS, ideally offline way. I currently use Jan.AI for everything but I realized that I've never tried to upload files to it before, and my current configuration doesn't seem to allow uploading.

after making this post a while ago, i tried out these three techniques for providing tool-result data to the LLM

• append to assistant message
• send as user response
• send model-specific tool-response type

Findings

turns out - the assistant-message appending works great for larger LLMs, but not so well for smol ones.

meanwhile the user-side method works better than expected!

i didnt spend too much time with the model-specific tool role stuff, since i want my tooling to remain model-agnostic.

i will probably switch to the user-side method now for gopilot, leaving behind the assistant-only approach

Tool call formatting improvements

Turns out - my initial tool calling formatting was SUPER token-inefficient - who knew...

So I went from this formatting

okay lemme look that up online
{"tool_name": "web_search", "args": {"query": "how to make milk rice"}}
just put milk and rice in a bowl and mix em

to this, MUCH simpler format

okay lemme look that up online
Tool: web_search("how to make milk rice")
Result: just put milk and rice in a bowl and mix em

which is like - just.... WAY better!!!!

• tokens reduced from 43 down to 24 (cost savings)
• way easier to read
• relies on models code-writing ability
• allows for specific assignment like in json: Tool: web_search(query="my query here")

i hope this is useful to someone out there.

if so, maybe share where you are applying it and tell us about your experience! <3

Hugging Face: https://huggingface.co/PlayHT/PlayDiffusion

the goal is to have an agent that can:

• Understand a complex problem description.
• Generate initial algorithmic solutions.
• Rigorously test its own code.
• Learn from failures and successes.
• Evolve increasingly sophisticated and efficient algorithms over time.

https://storage.googleapis.com/deepmind-media/DeepMind.com/Blog/alphaevolve-a-gemini-powered-coding-agent-for-designing-advanced-algorithms/AlphaEvolve.pdf

Model: https://huggingface.co/fancyfeast/llama-joycaption-beta-one-hf-llava

Demo: https://huggingface.co/spaces/fancyfeast/joy-caption-beta-one

Release Post: https://civitai.com/articles/14672

Github: https://github.com/ace-step/ACE-Step

Checkpoints: https://huggingface.co/ACE-Step/ACE-Step-v1-3.5B

Demo: https://huggingface.co/spaces/ACE-Step/ACE-Step

https://huggingface.co/nvidia/parakeet-tdt-0.6b-v2

im building som dum lil foss llm thingy for godot and now im interested in letting users implement their own MCP servers.

so like - okay, the model context protocol page says, that most servers use stdio for every interaction. So now - the request format can be seen here, its apparently a JSONrpc thing.

so - first thing i want to do is retrieving all the capabilities the server has.

i looked through all the tabs in the latest docs, but could not find the command for listing all the capabilities. so i installed some filesystem mcp server which runs well and tried this:

PS C:\Users\praktikant> npx -y @modelcontextprotocol/server-filesystem "C:\Users\praktikant\Desktop"
Secure MCP Filesystem Server running on stdio
Allowed directories: [ 'C:\\Users\\praktikant\\Desktop' ]
{\
"jsonrpc": "2.0",\
"id": 1,\
"method": "capabilities",\
"params": {}\
}

- aaaaaand nothing was returned. no string, no nothing.

so maybe its not a string which is sent via stdio but some other byte-based thing?

if anyone has experience with this, or is gud at guessing, pls tell me what u think i might be missing here <3

There are two main approaches in total:

1. Step to step
2. Begin to steps to end
   Currently, these are the two mainstream methods of instantiation.

It is widely recognized that if AI is not aligned with human values, it could cause harm to society.
Yet, this does not mean such systems lack intelligence.

So, what truly defines intelligence?
Why do so many researchers focus solely on intelligence aligned with human values?
Is it because their own understanding is limited, or because machines are not yet truly intelligent?

I believe intelligence should not be confined to narrow, human-centric definitions.
What we call "intelligence" today might be an illusion.
True intelligence cannot be defined—
the moment we define it, we lose its essence.

please tell me, thanks

Today we announce Mistral Small 3.1: the best model in its weight class.

Building on Mistral Small 3, this new model comes with improved text performance, multimodal understanding, and an expanded context window of up to 128k tokens. The model outperforms comparable models like Gemma 3 and GPT-4o Mini, while delivering inference speeds of 150 tokens per second.

Mistral Small 3.1 is released under an Apache 2.0 license.