PrismNote

A fast, local-first, Jupyter-compatible data-science notebook with built-in AI.

What it is

PrismNote is a data-science notebook served by a single executable: a Rust (Axum) backend, a React + TypeScript (Vite) frontend embedded in the release binary, and a thin Python launcher on PyPI that downloads the matching prebuilt server binary on first run. It executes code in 15+ languages, runs SQL against local engines and 9+ databases, sandboxes untrusted/AI-generated code in disposable Docker containers, and provides integrated code assistance via Claude, OpenAI, or a local Ollama model. The core notebook semantics (cross-cell state, real error handling, per-notebook kernel isolation, .ipynb import/export) are verified to match real Jupyter behavior.

Features / Highlights

• Jupyter-compatible kernel semantics — cross-cell state, real tracebacks, and per-notebook kernel isolation (a regression that used to share one global Python namespace across notebooks is fixed and re-verified live).
• SQL cells against real backends — SQLite and DuckDB (embedded, no server), PostgreSQL and MySQL via real servers; all covered by integration tests with genuine CREATE TABLE/INSERT/SELECT round trips.
• Cloud warehouse connectors — real connection + query execution for Snowflake, BigQuery, Redshift, Azure Synapse, Databricks, Athena (real SigV4), Presto, and Trino.
• Sandboxed code execution — each run spawns a brand-new disposable Docker container: no network by default, memory/CPU/process limits, a wall-clock timeout, and real stdout/stderr/exit-code capture.
• Integrated AI, privacy-friendly — code explain/fix/complete, NL-to-SQL, and RunPod fine-tuning, all disabled until you configure a provider (including local Ollama).
• Key integrations — dbt (real CLI + manifest/run-result parsing), GitHub (Contents API backup/sync), Airflow REST, Kubernetes (kubectl), RunPod GraphQL, and real cloud-storage clients for S3/GCS/Azure Blob/Google Drive.
• Single-binary distribution — the release build embeds the frontend; pip install prismnote gives you one command to run.

Who it's for / Use cases

• Data scientists and analysts who want the Jupyter experience plus real SQL and warehouse access from the same notebook — no separate BI tool.
• People running untrusted or AI-generated code who want real container sandboxing (no network, enforced limits, timeout) rather than in-process execution.
• Local-first users who don't want a heavyweight hosted notebook platform; everything runs on your machine by default.

What it is not

• No prebuilt binary for Windows/Linux yet. The published release binary is macOS Apple Silicon only; other platforms must build from source.
• MongoDB is not implemented (an explicit error, not fake success).
• Data-quality scoring is not wired to real execution — the two scoring code paths are the one integration still without a working implementation.
• Idle kernels are never evicted (resource management improvement, not a correctness gap).
• Not without known gaps. Backend tests pass (170, plus 114 frontend/backend vitest), but frontend lint (435 errors) and Playwright e2e (57/85) currently fail in CI without failing the build — tracked in ROADMAP_HONEST.md.
• The PyPI metadata says Development Status :: 5 - Production/Stable; the repo's own docs are more honest — this is actively developed single-maintainer software, and ROADMAP_HONEST.md is the source of truth for what's broken or unverified.

Project status

v1.12.0 (PyPI). Active and CI-green for build/tests; the honest, current set of bugs and gaps is maintained in ROADMAP_HONEST.md. Release history: CHANGELOG.md and GitHub Releases.

Installation

pip install prismnote
prismnote

On first run, the thin Python launcher (python/prismnote/_cli.py) downloads the matching prebuilt server binary from GitHub Releases and execs it. As of this version the prebuilt binary is published for macOS (Apple Silicon) only; other platforms build from source:

git clone https://github.com/Mullassery/PrismNote.git
cd PrismNote
make build          # builds the frontend first and embeds it into the binary

make build is the only path that produces a binary that serves the UI (target/release/prismnote). A bare cargo build --release builds a backend with no frontend assets.

Requirements: Rust (stable), Node 20.19+ (Vite 8), and Docker for sandboxed code execution.

Quick start

pip install prismnote
prismnote             # serves the app; printed URL, default http://localhost:8000

Create a notebook, add a SQL cell, and query, e.g., a local DuckDB file — or a Python cell and run it. Enable AI/config by copying .env.example to .env (Google Sign-In and AI features are disabled until configured).

Usage

• SQL backends: SQLite and DuckDB run embedded with no server required; PostgreSQL/MySQL need a reachable server. MongoDB returns an explicit not-implemented error.
• Sandboxed languages: Python, Bash/shell, JavaScript (Node), Ruby — each run in a fresh, limited, disposable Docker container (requires Docker).
• Warehouse queries: Snowflake, BigQuery, Redshift, Azure Synapse, Databricks, Athena, Presto, Trino via crates/server/src/cloud_warehouse/.
• Cloud storage: S3, GCS, Azure Blob, Google Drive clients wired to the /cloud-storage API (unit-tested HTTP clients; automatic notebook-backup scheduling on top is not yet built).
• AI/MCP: see docs/guides/AI_MCP_INTEGRATION.md and the MCP examples in examples/.
• Development: backend on http://localhost:8000 (cargo run); frontend dev server on http://localhost:5173 (cd frontend && npm install && npm run dev).

Documentation

• docs/README.md — documentation index
• Getting Started · Multi-Language Support · Multi-Terminal · Execution Backends
• API Reference · Settings · Security
• ROADMAP_HONEST.md — verified list of what's broken, untested, or not built

Contributing

See CONTRIBUTING.md (dev setup, real test/lint commands) and CODE_OF_CONDUCT.md. Security issues: report via SECURITY.md, not public issues.

Related projects

• MudFish — Rust-native web crawling and intelligence engine; feed crawl results into notebooks and RAG pipelines
• ClusterAudienceKit — RFM analysis, K-Means clustering, CLV, churn detection, and lookalikes for data science work
• PyReverseETL — quality-validated data activation with audit trail and lineage, a natural step after analysis
• PyInferenceManager — multi-provider LLM inference executor, complements PrismNote's AI integration
• MessageBirds — open-source, event-driven Customer Data & Engagement Platform (Rust + TypeScript)

Browse all projects →

License

Apache License 2.0. See LICENSE for the full terms.