MARE
Adaptive Retrieval Engine for Agentic Stack
What it does
MARE started with a simple question: do we really need to embed everything? Instead of chunking and embedding the full dataset, MARE uses a small semantic navigation index and lets an agent investigate data through MongoDB tools. In a 10K-incident benchmark, it used just 604 vectors vs 60,000 for RAG, while scoring 19/20 vs 18/20 on held-out questions. RAG remains faster for simple lookup. MARE explores where agentic retrieval helps on multi-hop, investigative questions.
Mongo Adaptive Retrieval Engine: navigate MongoDB instead of Top-K RAG - shekhartata/mare
MARE is an agentic retrieval layer on MongoDB: a small navigation index to find the neighborhood, then live find / filter / count against the source of record. It is not a cheaper chunk index. RAG embeds every chunk and answers from Top-K. MARE embeds neighborhoods, lets the agent hop, and cites database.collection:document_id . This repo is a working demo of that pattern , not a drop-in for an arbitrary Mongo cluster. Query time is schema-blind; the index and hops assume this SaaS-ops corpus ( mare_demo : customers, tickets, deployments, migrations, incidents, logs, joined on customer_id ). Pointing MONGODB_URI at your database is not enough. Documents still need tenant_id (injected on…from github.com
Does the same job
all alternatives →
Ragie: Agent-Ready RAG-as-a-ServiceSep 2025 · ▲21Smarter RAG with Agentic Retrieval & Context-Aware MCP
- CVClient Vector Search – Embeddings and Semantic Search with 5 Lines2023 · clientvectorsearch.com · ▲11
Hey HN, we're excited to show you client-vector-search, a client-side library that helps you embed, store, search, and cache vectors in your browser or node env. We needed it at https://searchbase.app and that's why we've built it. with it you get: 1. easy setup: you only need to add 5 lines of code to build a semantic search 2. no embedding api needed: you don't need an api and have to pay for it unless ure scaling up millions 3. faster search: modern hardware is better than cheap cloud computers (0.5vCPUs) 4. zero latency: no back-and-forth with server-side 5. easy integration…
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 16d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 26d ago · cactuscompute.com
Launched alongside, August 2026
the whole month →- TL
Life & fun · 10d ago · louisabraham.github.io
- SA
Hello HN! I found that picking out plausible but diverse skin tones for my digital art and game development projects was kind of difficult, and I got curious about if there was a way to define a color space that made it easy. I've built a color picker and procedural generation algorithm based on the space as well as a bunch of other fun js features and demos throughout the page that use the equations. If you find it interesting, I have lots of explanations of how I built it and what properties the space has. The methodology might be a bit shaky, but hopefully the result is as helpful for…
Life & fun · Aug 2026 · toneyalexander.github.io
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 16d ago · simedw.com