AI Repo: ahujasid/mcp-for-blender

Skyler: Hey everyone, welcome back to Git News AI — your weekly deep dive into the hottest A I repositories blowing up on GitHub right now. Skyler: Big thanks to our sponsors keeping this show alive: Ada, helping teams integrate A I seamlessly into their workflows; Ago Consulting — that's "Ahh-Go" — your go-to for silicon development from A I all the way down to SoC design; and Zen Semiconductor, an umbrella company with a whole portfolio of silicon-industry companies, including Ada for deploying A I at scale in any business — and if you haven't heard about their Vect appliance yet, it runs local L L Ms and vector search on dedicated hardware, which is seriously cool. Skyler: Alright, today we've got a wild lineup — we're talking about controlling Blender with any L L M you want, high-quality voice cloning in over six hundred languages, turning entire codebases into queryable knowledge graphs, multi-agent L L M trading frameworks, and a self-hosted L L M gateway that unifies over forty providers. Let's get into it. Skyler: First up, we've got mcp-for-blender by ahujasid, and this one is genuinely one of those repos that makes you go — wait, you can do that? Skyler: So the idea here is pretty straightforward but incredibly powerful. This is a community-built plugin that lets you control Blender 3D using any L L M of your choice through the Model Context Protocol, or M C P. Skyler: If you're not familiar with Blender, it's the massively popular open-source 3D creation suite — modeling, animation, rendering, the whole deal. And now you can talk to it through an L L M. Skyler: This is sitting at over thirty-one thousand stars, which tells you the community absolutely resonated with this idea. And look, it's not an official Blender Foundation project, this is a community effort, which honestly makes it even more impressive. Skyler: The M C P framework is doing a lot of heavy lifting here — it's basically a standardized way for L L Ms to interface with external tools and environments, and ahujasid has built a Blender plugin that speaks that language. Skyler: Recent activity shows they're actively adding new features — there's a Meshy integration with user A P I keys, which is interesting because Meshy is an A I-powered 3D model generation tool. So now you're layering A I on top of A I to drive a 3D creation pipeline. Skyler: Think about what this unlocks for developers. If you're building pipelines that need to generate or manipulate 3D assets programmatically, you could now describe what you want in natural language and have an L L M orchestrate the Blender operations for you. Skyler: Artists and developers who work at the intersection of creative tools and automation are going to love this. Game developers, VFX pipelines, anyone doing procedural 3D content generation — this is worth a serious look. Skyler: The fact that it works with any L L M is also huge. You're not locked into one provider. Whether you're running something local or using a cloud-based model, if it supports M C P, you're good to go. Skyler: The Python codebase means it's approachable if you want to contribute or extend it. And given that star count is climbing fast, expect the feature set to grow quickly. Skyler: If you're at all curious about A I-driven creative tooling or procedural content workflows, go check out the show notes for the link to mcp-for-blender. It's genuinely one of those projects that feels like a glimpse of where creative software is headed. Skyler: Okay, moving on to something that had me doing a double-take at the description. OmniVoice by k2-fsa — high-quality voice cloning T T S for over six hundred languages. Skyler: Six hundred. Languages. That is not a typo. I want to let that sink in for a second. Skyler: This repo is sitting at nearly fifteen thousand stars and climbing, and with ninety percent A I relevance in our tracking, it's squarely in the wheelhouse of everything we cover here. Skyler: So OmniVoice is a text-to-speech system with voice cloning capabilities, built in Python, and the scale of language support it targets is genuinely unprecedented for an open-source project like this. Skyler: Voice cloning means you can feed it a reference audio sample and the model will attempt to replicate that voice's characteristics when generating new speech. That's incredibly useful for accessibility tools, localization pipelines, interactive A I agents, you name it. Skyler: Recent updates show they're doing active work on audio quality — there's a fix for input waveform types in the denoising pipeline, which matters a lot for output fidelity. Clean input means clean output when you're working with audio. Skyler: But the really exciting recent addition is optional LoRA finetuning support via P E F T. If you're in the M L space, you know that LoRA — Low-Rank Adaptation — is one of the most efficient ways to finetune large models without needing massive compute. Skyler: Adding P E F T support means you can take OmniVoice and adapt it to specific voices, accents, or domains without retraining the whole model from scratch. That's a massive unlock for developers building specialized voice applications. Skyler: Think about the use cases — customer service voice bots that need to sound locally authentic across dozens of markets, accessibility tools for low-resource languages, educational platforms, content localization at scale. The six-hundred-language support makes all of those suddenly feasible. Skyler: If you're working in speech synthesis, building voice-enabled A I agents, or doing any kind of audio M L work, OmniVoice deserves a spot on your radar. Check the show notes for that link. Skyler: Alright, buckle up because the next one is a beast. Graphify by Graphify-Labs, and this thing is sitting at a staggering one hundred and twenty-eight thousand stars. That is not a small number. Skyler: The pitch is this: take any codebase — along with its docs, SQL schemas, config files, and P D Fs — and turn all of it into a queryable knowledge graph. And then expose that graph as a skill for tools like Claude Code, Cursor, Codex, and Gemini C L I. Skyler: Now I want to highlight something in the description that I think is really important: local, deterministic A S T parsing, every edge explained, no vector store. That's a very deliberate architectural choice. Skyler: A S T stands for Abstract Syntax Tree — it's a structured representation of source code that captures the actual relationships between functions, classes, imports, and so on. By using deterministic A S T parsing, Graphify isn't guessing at relationships, it's deriving them mathematically from the code itself. Skyler: And the no vector store part is significant. A lot of code intelligence tools these days use embeddings and vector search, which can be fuzzy and sometimes hallucinate connections. Graphify is going for precision over approximation. Skyler: Every edge in the graph comes with an explanation. So when you query and get back a relationship, you also get the reasoning for why that relationship exists. That's huge for trust and debuggability. Skyler: Recent activity shows version zero point nine point nine two just dropped, and they fixed an important bug around preserving parallel edges through graph serialization and the M C P output. That kind of attention to graph fidelity tells you the team cares deeply about correctness. Skyler: The M C P integration is what makes this feel like a next-generation tool. Your A I coding assistant can now ask Graphify questions about your codebase and get back structured, accurate answers about how everything connects. Skyler: Imagine onboarding to a massive legacy codebase and being able to ask your A I assistant "show me everything that depends on this function" and getting a precise answer backed by a real knowledge graph. That's the dream, and Graphify is building toward it. Skyler: If you're a developer advocate, a platform engineer, or someone who deals with large complex codebases — or if you're building developer tooling and want to add deep code understanding — this one is essential. Show notes have the link. Skyler: Next up, TradingAgents by TauricResearch. This one is at over one hundred and eleven thousand stars, and the concept is genuinely fascinating if you're interested in A I applied to financial markets. Skyler: This is a multi-agent L L M framework specifically designed for financial trading. The idea is that instead of one monolithic model trying to make trading decisions, you deploy a coordinated team of specialized L L M agents that each play a different role in the analysis and decision pipeline. Skyler: Think about how a real trading firm actually works — you've got analysts, risk managers, quant researchers, traders. TradingAgents mirrors that structure in software, with different agents playing different roles and communicating with each other. Skyler: This is version zero point seven point zero, freshly released, so the team is actively developing and iterating. Major version bumps in an active project like this usually signal meaningful capability expansions, not just bug fixes. Skyler: The multi-agent approach here is really smart for this domain. Financial decisions involve multiple competing considerations — technical signals, fundamental data, risk exposure, market sentiment — and having specialized agents handle each dimension before synthesizing a final decision maps well to how professionals actually think about these problems. Skyler: Now I want to be really clear here — this is a research and educational framework. Anyone thinking about deploying automated trading based on L L M output needs to understand the risks and regulatory landscape. But as a technical demonstration of multi-agent coordination? It's impressive. Skyler: What makes this especially interesting from an A I engineering perspective is the orchestration layer. Getting multiple L L M agents to collaborate, share context, and avoid conflicting decisions is a genuinely hard technical problem, and seeing how TauricResearch approaches it is valuable even outside the finance domain. Skyler: If you're building multi-agent systems, doing research in A I for finance, or just want to study how complex agent coordination frameworks are architected, TradingAgents is worth a deep read. Check the show notes for the link. Skyler: And finally, let's talk about Otari by Mozilla A I — and yes, that Mozilla. This one is newer and smaller in star count at six hundred and sixty, but the eighty percent A I relevance score and who's behind it made me want to include it. Skyler: Otari is an open-source, self-hosted L L M gateway that's fully compatible with the OpenA I A P I format. The pitch is one endpoint for over forty L L M providers, with virtual keys, budget controls, and usage tracking — all running on your own infrastructure. Skyler: So think about the problem this solves. If you're building an application that might need to switch between L L M providers, or test different models, or manage costs across a team, right now you're probably juggling multiple A P I keys, different A P I formats, different pricing dashboards. It's a mess. Skyler: Otari puts a single, OpenA I-compatible layer in front of all of that. Your application talks to one endpoint, and Otari handles the routing, key management, and accounting behind the scenes. Skyler: The OpenA I compatibility is critical because it means you can drop this into an existing application that already uses the OpenA I client library with minimal code changes. That's a huge adoption barrier eliminated. Skyler: Virtual keys are interesting — you can issue internal keys to different team members or services that map to real provider credentials that they never see. That's solid security hygiene for organizations. Skyler: The budget controls are also something engineering leaders are going to appreciate. You can set spending limits per team, per project, per key — so you don't wake up to a surprise bill because someone left a test loop running. Skyler: Recent activity shows they're doing serious access control work — documenting reject user mismatch policies and clarifying how managed credentials are handled. That tells me the team is thinking carefully about security, which for a gateway handling your A P I keys is exactly what you want. Skyler: Mozilla A I building this is also a signal — they've got a track record of caring about openness and user control, and an L L M gateway you run yourself fits that philosophy perfectly. Skyler: If you're an engineering team dealing with multi-provider L L M complexity, or you just want to avoid lock-in and keep control of your A I spend, Otari is absolutely worth evaluating. Show notes have the link. Skyler: Alright, that's a wrap on today's episode of Git News AI — what an absolute stacker of a lineup today between 3D control with L L Ms, six-hundred-language voice cloning, knowledge graph code intelligence, multi-agent trading, and self-hosted L L M gateways. Skyler: Huge thank you to our sponsors — Ada for making A I integration accessible for teams everywhere, Ago Consulting for expert silicon development guidance from A I all the way to SoC, and Zen Semiconductor for their incredible portfolio including the Vect appliance and Ada for enterprise-scale A I deployment. Skyler: If you're hungry for more A I coverage, go check out A I Special Report in the Ada podcast network — great companion show to this one — and make sure you're subscribed so you never miss an episode. All the repo links and resources we mentioned today are waiting for you in the show notes. See you next time.