prismnote 0.4.2

Modern, open-source Jupyter-compatible data science notebook with Rust performance

◆ PrismNote

A fast, modern, open-source data-science notebook. Rust engine · React UI · local-first · AI-native.

License: MIT · Status: Beta (v0.4) · PyPI: prismnote

---

Overview

PrismNote is a local-first alternative to Jupyter, Zeppelin, and Databricks notebooks. A Rust + Axum backend drives a persistent Python kernel, and a modern, VS Code-style React UI gives you Python + SQL, rich output and live charts, local AI assistance, scheduled jobs, git, and generated cloud-deploy files — all on your machine.

• Local-first & private — runs locally; local AI via Ollama, no account required.
• Jupyter-compatible — native .ipynb import/export.
• Batteries included — kernel, SQL, charts, AI, jobs, git, deploy, search, terminal.

Status note: PrismNote is beta. The notebook, kernel, AI, jobs, git, and charts are functional. Cloud-warehouse connectors are scaffolding (query results are currently stubbed), and deployment files are generated templates you review before shipping. See the roadmap.

---

Install & run

From source (recommended today)

Requires Rust (stable), Node 18+, and Python 3.8+ on PATH:

# Python runtime deps for the kernel (ipykernel is required; the rest enable
# rich output, %sql, and charts)
pip install ipykernel pandas matplotlib rich duckdb

# 1) backend — http://localhost:8000
cargo run

# 2) frontend — http://localhost:5173 (proxies /api and /ws to the backend)
cd frontend && npm install && npm run dev

Open http://localhost:5173.

From PyPI

pip install prismnote
prismnote

The PyPI package is a thin launcher: on first run it downloads the prebuilt server binary for your platform from the matching GitHub Release (vX.Y.Z). If a release binary isn't published for your platform/version yet, use the from-source steps above. (pip install prismnote succeeds regardless; only the prebuilt launch depends on a release.)

Enable AI (optional)

• Local, free, private — Ollama: install from https://ollama.com, then ollama pull qwen2.5-coder. The UI talks to Ollama from the browser, so allow the dev origin once:

  OLLAMA_ORIGINS=http://localhost:5173 ollama serve
  

• Claude / OpenAI: set it in AI settings, or via env (persisted to ~/.prismnote/ai_config.json):

  export PRISMNOTE_AI_PROVIDER=claude   # or openai / ollama
  export ANTHROPIC_API_KEY=...          # or OPENAI_API_KEY
  

---

Features

Notebook & execution

• Persistent shared kernel — variables/imports/functions from one cell are available in later cells.
• Per-cell interpreters (magics): %python (default), %sql (in-process DuckDB; can query DataFrames from other cells — needs pip install duckdb), %sh / !cmd (shell), %md (rendered markdown).
• Rich output — text, HTML, matplotlib figures, and pandas DataFrames.
• Chart switcher — view any DataFrame as Table / Bar / Line.
• Live streamed output over WebSocket while a cell runs.
• Interrupt a running cell and restart the kernel.
• Dynamic input widgets — prism.input / slider / select / checkbox re-run the cell when changed.

AI assistance

Local models via Ollama, or Claude / OpenAI.

• In-cell AI edit (diff accept/reject), Fix with AI on errors, Explain.
• Inline autocomplete (ghost text) when Ollama is connected.
• Teacher persona — explanations of the "why" plus a contextual tip.
• Agent panel — Plan/Act assistant that proposes and applies cell edits.

Errors that help

• Natural-language explanations for common Python errors.
• In-editor markers on the offending line/column, with a collapsible traceback.

Workflow

• Jobs — save a whole notebook and run it as a unit: manual / interval / daily, with status + run history.
• Airflow — a stable run-by-name trigger and a generated DAG.
• Source control — init, clone, commit, push, pull, status from the UI.
• Cloud deploy — generates Dockerfile, docker-compose.yml, k8s.yaml, and fly.toml (review before deploying).

Workspace & UX

• VS Code-style layout; all panels collapsible, each with independent font +/-.
• Responsive — side panels auto-collapse on narrow windows.
• Global search, command palette, integrated terminal, a Python console (shares the kernel), and a file browser that works in any browser.
• Dark/light themes; .ipynb import/export.

---

Architecture

┌──────────────────────────────┐        ┌─────────────────────────────┐
│  React + TypeScript (Vite)   │  HTTP  │      Rust backend (Axum)    │
│  Monaco · Tailwind · zustand │ ─────▶ │  REST API + WebSocket       │
│  cells · panels · AI · jobs  │ ◀───── │  routing, jobs, git, deploy │
└──────────────────────────────┘   WS   └──────────────┬──────────────┘
                                                        │ stdin/stdout (JSON)
                                                ┌───────▼────────┐
                                                │ Persistent      │
                                                │ Python kernel   │
                                                │ (shared globals)│
                                                └─────────────────┘

• Backend spawns one long-lived python process and talks to it over a line-framed JSON protocol; outputs are Jupyter-style MIME bundles, streamed live.
• Kernel has a shared namespace, is SIGINT-interruptible and restartable, with matplotlib (Agg) and rich preloaded.

---

Keyboard shortcuts

All of these are wired:

Shortcut	Action
⌘N / ⌘O / ⌘S	New / Open / Save notebook
⌘K	Global search (in a focused cell, ⌘K is AI edit instead)
⇧⌘P	Command palette
⌘,	Settings
⌘⇧↵	Run all cells
⇧↵	Run the focused cell

(⌘/Ctrl depending on platform.)

---

Deploy

Open Deploy to Cloud to copy/download the generated artifacts, then:

docker compose up -d            # Docker
kubectl apply -f k8s.yaml       # Kubernetes
fly launch --copy-config --now  # Fly.io

---

Repository layout

crates/server/   Rust backend (api, kernel, jobs, db, ws, deploy, git…)
frontend/        React app (components, hooks, api clients)
python/          PyPI launcher package (prismnote)
*.md             Architecture & comparison docs

Comparisons: ZEPPELIN_COMPARISON.md · DATABRICKS_COMPARISON.md · NOTEBOOK_COMPARISON_MATRIX.md

---

Roadmap

• Functional cloud-warehouse query execution (currently scaffolded/stubbed).
• Prebuilt release binaries for all platforms (so pip install runs out of the box).
• Distributed compute (Spark) and a catalog/data browser.
• Real-time collaboration (live cursors / co-editing).
• Notebook parameters and multi-notebook job composition.

---

License

MIT — see LICENSE.