HiveMind WebChat

[!WARNING] HiveMind is pre-release software under active development. Expect bugs and breaking changes between releases.

A browser-based WebChat terminal for HiveMind. It serves a small chat page from a local web server. The page connects to a hivemind-core instance as a satellite using the HiveMind-js client, so you can talk to your voice assistant from any browser on the network.

Where it sits

HiveMind is a mesh: satellite devices connect to a central hivemind-core instance over an authenticated, encrypted protocol. WebChat is one such satellite, a text front end. The Python package here is only a static web server (Tornado). The actual HiveMind connection runs in the browser via HiveMind-js. You point the page at a running hivemind-core instance and enter the access key it issued you.

browser (this page + HiveMind-js)  ──websocket──►  hivemind-core  ──►  OVOS / agent

Documentation

Full docs live under docs/: getting started, configuration, architecture (Python backend + JS frontend + the headless bridge), deployment, dependencies, testing, and troubleshooting.

Install

pip install hivemind-webchat

Or from source:

git clone https://github.com/JarbasHiveMind/HiveMind-webchat
cd HiveMind-webchat
pip install .

The HTTP server only needs tornado. hivemind-bus-client (2.x) and ovos-utils are pulled in for the optional headless bridge. Dependency policy lives entirely in pyproject.toml (no requirements.txt / setup.py / MANIFEST.in). The bus-client 2.x stack resolves from prerelease min-version pins with no --pre. See docs/dependencies.md.

Quickstart

1. Run hivemind-core and issue an access key

On the machine that will host the assistant, install and run hivemind-core, then add a client for the webchat:

hivemind-core add-client
# note the printed access key and password/encryption key
hivemind-core listen --port 5678

2. Start the WebChat server

hivemind-webchat --port 9090

This serves the chat page at http://localhost:9090. The port here is the web server's HTTP port, not the hivemind-core port.

3. Connect from the browser

Open http://localhost:9090, fill in the connection form, and click Connect to HiveMind:

Once the handshake completes, type a message and it is sent to hivemind-core as a recognizer_loop:utterance. Spoken replies are rendered back in the chat log.

The browser negotiates the highest HiveMind protocol version both peers support (WIRE-1). Against a protocol v3 hivemind-core instance it runs the Noise handshake over the default Noise_XXpsk2_25519_ChaChaPoly_SHA256 suite, with full cipher parity with hivemind-core, and against older hivemind-core versions it falls back to the legacy v1 password handshake (PBKDF2-HMAC-SHA256 key derivation + AES-GCM). Either way, all traffic after the handshake is encrypted end to end, entirely in the browser via HiveMind-js, which pairs native Web Crypto with the pure-JS @noble/ciphers + @noble/hashes bundle for the two primitives Web Crypto lacks (ChaCha20-Poly1305 and argon2id).

Protocol v3 (Noise)

Against a v3 hivemind-core instance (hivemind-bus-client 0.10.1a1 / hivemind-core 4.7.0a1 or newer) the browser negotiates the default ChaChaPoly suite and derives the PSK as argon2id(password, SHA-256(node_id)) in-browser, byte-for-byte identical to hivemind-core. So the Password field alone is enough:

• Password (default): the client stretches it with argon2id on-device to the Noise PSK. No server-side KDF change and no provisioning required.
• Provisioned PSK (optional): paste a 64-hex-char PSK (equal to argon2id(password, SHA-256(node_id))) into the Protocol v3 section of the connect dialog to skip on-device derivation. An optional server key pin enables KKpsk0 TOFU pinning.
• PBKDF2 (fallback): if a hivemind-core instance explicitly advertises the PBKDF2 PSK KDF, the client derives the PSK with PBKDF2 from the password instead.

The one caveat: a minimal page bundle shipped without the @noble primitives degrades to the Web-Crypto-only AES-GCM (25519_AESGCM_SHA256) + PBKDF2 subset, and then needs a provisioned PSK or a PBKDF2-advertising hivemind-core instance. If no PSK is available for a v3 hivemind-core instance at all, the client warns and falls back to the legacy v1 handshake, so the UX keeps working against every hivemind-core instance.

Command-line options

usage: hivemind-webchat [-h] [--port PORT]

Start HiveMind WebChat

options:
  -h, --help   show this help message and exit
  --port PORT  HTTP port to serve the webchat on (default 9090)

How it works

• hivemind_webchat.WebChat is a threading.Thread wrapping a Tornado HTTPServer. It serves templates/index.html at / and the chat assets under /static.
• index.html loads the HiveMind-js client from jsDelivr and app.js wires the connection form to JarbasHiveMind.connect(host, port, user, accessKey, password, options), where options carries the optional v3 provisioned PSK / server key pin.
• All HiveMind traffic (handshake, encryption, message routing) happens in the browser inside HiveMind-js. The Python side never touches hivemind-core.

Tests

WebChat is a Python + JavaScript hybrid, so it has two test suites.

Python (tests/): tests/test_smoke.py covers the Tornado server and the bridge construction. tests/e2e/ boots a real loopback hivemind-core instance via hivescope and drives the real WebchatBridge over a real HiveMessageBusClient. A chat message goes to hivemind-core and a speak reply is routed back. Only the browser/websocket frontend is mocked. No importorskip/skipif.

uv pip install -e ".[e2e]"
pytest tests/

JavaScript (tests/e2e.mjs): a Node end-to-end test that loads the exact HiveMind-js client the page ships, connects to a real loopback hivemind-core instance (tests/hub_fixture.py), performs the handshake, and verifies hivemind-core decrypts and receives an encrypted utterance. A second test (tests/v3_negotiation.test.mjs) drives the browser client's protocol-v3 negotiation path directly. It feeds the client a synthetic v3 ServerHello and asserts that it selects the AES-GCM Noise suite and derives a valid PSK from the password via the PBKDF2 KDF. (The loopback hivemind-core floors an older stack predating the v3 suite, so full v3-over-the-wire is not exercised end to end.)

npm install     # ws
npm run test:v3 # node --test tests/v3_negotiation.test.mjs  (no hivemind-core needed)
npm test        # v3 negotiation + node tests/e2e.mjs  (e2e needs python [e2e] stack)

Full details in docs/testing.md.

Security

The Tornado server is plain HTTP and out of scope for hardening here. For any non-local exposure, put it behind a reverse proxy (nginx, Caddy) with TLS, for example via Let's Encrypt. The HiveMind connection itself is always encrypted end to end between the browser and hivemind-core, independent of how this page is served.

Online demo

A static build is published from the gh-pages branch: https://jarbashivemind.github.io/HiveMind-webchat. It is the same page served by this server, pointed at whatever hivemind-core instance you enter.

Credits

Original WebChat UI: jcasoft.

License

Apache 2.0. See LICENSE.