aayushch/laya

Laya is an open-source, local-first AI notification command center that aggregates Slack, Gmail, GitHub, Jira, Notion, Outlook, Calendar (and more) notifications using local LLMs via Ollama and LM Studio. Supports cloud models via BYOK.

Laya – an AI‑powered command centre for your work inbox

What it is – Laya is a desktop application that sits between the tools you use every day (Slack, Gmail, GitHub, Jira, Notion, Outlook, calendars, etc.) and you. It pulls in notifications, runs them through locally‑hosted or cloud LLMs, and turns each event into an Action Card that you can approve, edit or dismiss. In short, it lets a language model do the first‑pass research and drafting so you spend less time answering routine messages and more time on the actual work.

How it works

  1. Ingestion – n8n (a local Node.js workflow engine) watches the supported services and normalises every incoming event.
  2. Processing – The Laya Engine (Python FastAPI) classifies the event, decides which AI persona should handle it (engineer, ops, sales, etc.), runs the appropriate LLM (via Ollama, LM Studio, Claude, GPT, Gemini or any LiteLLM‑compatible endpoint) and prepares a draft response or action.
  3. Presentation – A Tauri‑based desktop UI (Svelte 5 + Tailwind) shows the draft as an Action Card. You can approve it, tweak it, or reject it.
  4. Execution – Approved cards are sent back to n8n, which carries out the real work (create a PR, send a Slack reply, update a Jira ticket, etc.).

Key capabilities

  • Multi‑persona routing – Different LLM prompts for engineering, communications, ops, sales, HR, finance, each with its own toolset.
  • Card workspaces – Complex tasks (e.g., fixing a bug) become interactive sessions where a coding agent (Claude Code, Gemini CLI, Codex, Pi, Cursor) can be consulted step‑by‑step.
  • On‑demand research – Launch a deep‑dive session from any card; the agent can browse the web, read files, and return a concise summary.
  • Hybrid search (Coherence) – Combines vector similarity (ChromaDB) with classic BM25 over SQLite FTS5, letting you locate related tickets, PRs, or messages across all platforms.
  • Context association & memory – Laya learns to link entities like “BUG‑1234” ↔ a Slack discussion ↔ a PR automatically, improving over time.
  • Budget & analytics – Tracks LLM usage per feature, shows cost breakdowns, and can pause ingestion when a monthly cap is hit.
  • Chat sidebar – Ask Laya natural‑language questions about your inbox or ask it to create new rules on the fly.
  • Privacy‑aware – You can keep everything local (Ollama/LM Studio) or use cloud APIs; data classification tiers let you decide which steps run where.

Tech stack (high‑level)

  • Desktop shell – Tauri v2 (Rust) for a native cross‑platform app.
  • Frontend – Svelte 5 + Tailwind CSS.
  • Backend – Python 3.10+ FastAPI, async workers for each persona.
  • LLM integration – LiteLLM (supports Anthropic, OpenAI, Google, Ollama, any OpenAI‑compatible endpoint).
  • Workflow engine – n8n (local Node.js) handles the connectors to Slack, Gmail, GitHub, etc.
  • Storage – SQLite (with FTS5) for structured data, ChromaDB for embeddings.
  • Embeddings – Built‑in ONNX models or optional sentence‑transformers.

Who might use it

  • Professionals who receive a high volume of mixed‑type notifications and want AI to triage and draft replies.
  • Small teams that want a single place to see all work‑related messages, with the ability to automate routine actions.
  • Anyone who prefers a local‑first solution (running LLMs on their own machine) but still wants the option to fall back to cloud models.

Getting started

  • Download a pre‑built installer (macOS .dmg, Windows .msi/.exe, Linux .deb/.AppImage) – no need to install Python, Node, or Rust yourself.
  • On first launch, add your API keys (OpenAI, Anthropic, Google) or point Laya at an Ollama/LM Studio endpoint.
  • Connect the services you care about in Settings; Laya will start pulling in events automatically.

Development – The repo includes scripts to set up a full dev environment (Python venv, Node packages, Rust toolchain). The code is split into engine/ (Python backend), ui/ (Svelte + Tauri UI), and n8n/ (workflow definitions).

Bottom line – Laya is a full‑stack, open‑source AI assistant that turns everyday work notifications into actionable, AI‑drafted cards, letting you approve or reject them with a single click while keeping the heavy lifting either local or behind your own API keys.

Related

  • Dispatch
    Laya Open-Source System 1 Decision Model Beats TypeSafe Jev

    Laya delivers sub‑35 ms calibrated decisions across 100+ languages, open‑source weights, and 6–8× lower latency than TypeSafe AI’s proprietary Jev.

  • Project
    ollaya-dev/ollaya

    ollaya is a Rust‑based tool that serves non‑generative decision models locally, offering an Ollama‑like CLI, TypeSafe‑compatible HTTP API, and support for ONNX and GGUF back‑ends. It ships tiny ONNX graphs, pulls weights from Hugging Face, and provides calibrated, millisecond‑level predictions for structured questions, targeting AI agents and decision‑making workloads.

  • Project
    NandhaKishorM/laya

    Laya is a fast, multilingual decision engine that classifies text into structured answers (choice, score, yes/no) in a single forward pass, without text generation. It uses a router to pick the best model per request and supports fine-tuning for domain-specific accuracy.

  • Dispatch
    Ollaya: Local Inference for Jev-style Decision Models

    Ollaya is an open-source tool for running Jev-style decision models locally, offering millisecond-latency typed answers and drop-in compatibility with the TypeSafe API.

  • Project
    onyx-dot-app/onyx-foss

    Onyx is a self-hosted knowledge and context layer that connects to over 50 applications to provide AI agents with secure, relevant access to your organization's internal data.