Check out OpenResponses - A drop-in replacement of OpenAI platform APIs that works with open models, runs locally, and gives you tracing, RAG, tools, and privacy out of the box. Docs: openresponses.masaic.ai/introduction Github: github.com/masaic-ai-plat…

Introducing OpenAI Codex cli with your own model. Groq Inc OpenAI Google AI Anthropic DeepSeek openresponses.masaic.ai/use-cases/code…

Gemini 2.5 Flash ⚡️with Codex CLI using OpenResponses. openresponses.masaic.ai/use-cases/code…

Introducing Iterative Agentic Search: we call it Chunk Walking. A smarter way to explore documents with LLM-guided refinement at every step. - Multi-iteration query refinement - Dynamic filters from doc attributes - Memory across steps - Auto-stop when enough is found

You just need to follow three simple steps: 1. Upload files 2. Add them to vector store 3. Call open-responses API with following tool { "type": "agentic_search", "vector_store_ids": ["vs_product_docs"] } Get started here: openresponses.masaic.ai/quickstart#set… Compatible with: OpenAI

Every time I ask the agent to think step by step, I wonder if I should be doing the same.

Every startup becomes a logging company. Some just admit it sooner.

The hardest part of building agents isn’t planning what they’ll do. It’s predicting what they’ll believe.

AI evaluation metrics like coherence, fluency, and similarity are standard, but do they reflect real business impact? Most businesses don’t speak in terms of BLEU scores or perplexity. They care about metrics like Net Promoter Score (NPS), customer satisfaction, and retention.

OpenResponses recreated the OpenAI MCP demo with same code, but powered by groq + llama. No rewrites. No lock-in. Just clean abstractions. Github: github.com/masaic-ai-plat… OpenAI Developers Groq Inc Meta

Repo: github.com/masaic-ai-plat… 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

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 👇 github.com/masaic-ai-plat…

Energy startups are the new shovels in the AI gold rush.

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 👇

Sovereign, voice-first AI stack for Indic languages. Fully open-source, privacy-first, and keeps data within national borders. Built with: Sarvam AI for robust speech ↔︎ text across dialects Qdrant for blazing-fast vector memory SigNoz for full observability

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. xmxnt A quick thread 🧵👇