@alexalbert__ Nice! We have been using similar technique called "playbooks" in our product where users can save a set of steps for LLM to re-use whenever needed.
@elonmusk@xai We are building an AI-native arch in stealth that has the potential to disrupt the industry. We started out to build a better Palantir but soon realised it can be used to build any enterprise software in weeks.
1/
We’re speaking at #OpenSourceSummit , 5 August in Hyderabad!
Amant Kumar and I will walk through how we built a sovereign, open multi-agent AI platform — deployed across borders, languages, and departments.
@0xTalus
A quick thread 🧵👇
Sovereign, voice-first AI stack for Indic languages. Fully open-source, privacy-first, and keeps data within national borders.
Built with:
@SarvamAI for robust speech ↔︎ text across dialects
@qdrant_engine for blazing-fast vector memory
@SignozHQ for full observability
Our open-responses stack for agentic orchestration
Watch how it captures facts, reasons, and respond all in local language.
Code: https://t.co/NLlUitPsGF
Can you run a full ecommerce flow?
🔍 search
🖼️ preview
🛒 checkout
Using just one Orchestrated API
I demoed it live at AI Tinkerers Delhi. Here’s how it works 👇
@garrytan Yes the abstraction gap is massive. We are engineering an interface to narrow it. A kind of experience layer where LLMs will be used as a compiler of natural language for structured action and not just a chatbot.
@bfl_ml Great news! We already have support for contextual image generation using Flux in OpenResponses. Contextual editing would be a game-changer.
https://t.co/aiAkFRPBCU
Repo: https://t.co/2z2T7XqeSU
Brand-smart images in one shot. 🔥
Images generation that adhere to brand guidelines available in files.
Runs on open models through the same Responses API as OpenAI, but on your own infra.
✅ Open platform
✅ Open chat model
✅ Open image model
Thanks to @togethercompute for awesome models.
Repo: https://t.co/2z2T7XqeSU
Brand-smart images in one shot. 🔥
Images generation that adhere to brand guidelines available in files.
Runs on open models through the same Responses API as OpenAI, but on your own infra.
✅ Open platform
✅ Open chat model
✅ Open image model
Thanks to @togethercompute for awesome models.
Repo: https://t.co/2z2T7XqeSU
Brand-smart images in one shot. 🔥
Images generation that adhere to brand guidelines available in files.
Runs on open models through the same Responses API as OpenAI, but on your own infra.
✅ Open platform
✅ Open chat model
✅ Open image model
Thanks to @togethercompute for awesome models.
Built an AI workflow that reads your brand manual and generates matching visuals.
Zero prompt hacks, fully self-hosted.
Here’s how it works, and why every block is swappable 👇 https://t.co/Js9aSoI08V
Repo: https://t.co/2z2T7XqeSU
Brand-smart images in one shot. 🔥
Images generation that adhere to brand guidelines available in files.
Runs on open models through the same Responses API as OpenAI, but on your own infra.
✅ Open platform
✅ Open chat model
✅ Open image model
Thanks to @togethercompute for awesome models.
Repo: https://t.co/2z2T7XqeSU
Brand-smart images in one shot. 🔥
Images generation that adhere to brand guidelines available in files.
Runs on open models through the same Responses API as OpenAI, but on your own infra.
✅ Open platform
✅ Open chat model
✅ Open image model
Thanks to @togethercompute for awesome models.
@athyuttamre@swyx We’re a team of 2, quit our jobs, going all in on the huge potential we see on "Responses API way".
Not a gateway but as a foundation for LLM apps and orchestrations.
Definitely would love to connect as there are quite a few inputs that could help us.