Skip to main content

Maritime intelligence system — distills cloud AI into compiled edge tiles

Project description

LucidDreamer — Maritime Intelligence System

PyPI version Python License: MIT Tests

A thinking boat that learns its captain, tiles its knowledge, and compiles its intelligence.

What This Is

LucidDreamer is a modular AI system that starts with cloud intelligence and distills it down to edge-ready compiled code. It learns your voice, your commands, your fish, your waters. Every interaction creates a tile. Every tile makes the system smarter. Over time, the system compiles what it knows into deterministic code that needs no inference — zero tokens, zero latency, zero cloud.

The Big Idea

Most AI systems stay soft forever — they always need inference, always cost tokens, always depend on the cloud. LucidDreamer is different. It treats the soft model as a teacher, not a worker. The cloud API's job is to explore the space of possible commands, responses, and decisions. Then it compiles what it learns into hard code. The edge device just runs the compiled code.

Think of it like learning to drive. At first, every action requires conscious thought (inference). Over time, the actions become automatic (compiled). You don't "infer" how to turn the steering wheel — you just do it. LucidDreamer does the same thing for boat operations.

What Can It Do?

Autopilot Voice Control

Tell the autopilot what to do in natural language. The system learns your dialect, your command style, and your preferred responses. Starts with cloud API, ends up compiled on a Raspberry Pi.

Fish Sorting Vision

Camera watches the deck, classifies every fish, counts per species per hold. When uncertain, alerts the deck crew. Every fish gets a photo and an identity. Captain reviews low-confidence calls during transit. Buyer's count is the final anchor.

Chart Intelligence

Reads the navigation display (via screen capture or NMEA data), understands depth contours, predictor lines, radar overlays. The captain can ask "are we going to remain deep enough for the next 10 minutes?" and the system extrapolates from speed, heading, and bathymetric data.

Cocapn Standalone Chatbot

A chatbot that lives on the boat. Learns from every interaction. Tiles its knowledge. Can output mouse movements to the navigation computer (deliberate, visual, captain can stop if wrong). Understands the 5-minute predictor line, radar overlay, color-coded contours, and whatever else the captain teaches it.

Architecture

CLOUD (Starlink)          EDGE (Boat)              HUMAN (Captain)
┌──────────────┐    ┌──────────────────┐    ┌──────────────────┐
│ Heavy Model  │    │ Compiled Tiles   │    │ Voice + Review   │
│ (GPT-4 etc)  │───▶│ (regex+lookup)   │◀──▶│                  │
│              │    │ Gemma 1B fallback│    │ Ground Truth     │
│ Distills to  │    │ Audio tiles      │    │ Oracle           │
│ tiles        │    │ Vision classifier│    │                  │
└──────────────┘    │ Chart reader     │    └──────────────────┘
                    │ Cocapn chatbot   │
                    └──────────────────┘

Installation

pip install luciddreamer

Quick Start

See docs/GETTING-STARTED.md for the full walkthrough.

Documentation

Modules

Module Description
luciddreamer.tiles Tile types, storage, and retrieval
luciddreamer.compiler Rigid structure finder and compiler
luciddreamer.bathymetry Confidence mapping and coverage tracking
luciddreamer.router Route to compiled tiles vs fallback inference
luciddreamer.audio Audio tile management (pre-generated responses)
luciddreamer.chart Chart interpretation and depth prediction
luciddreamer.vision Fish sorting vision pipeline
luciddreamer.cocapn Standalone chatbot integration
luciddreamer.simulators Test simulators for all subsystems

Ecosystem Integration

LucidDreamer integrates with the SuperInstance ecosystem:

  • eisenstein-embed — Bitvector matching for fast tile lookup
  • tensor-spline — Compress tile data for edge storage
  • device-router — Route inference to right device
  • training-throttle — Manage cloud API usage
  • plato-core — Tile provenance and lifecycle
  • flux-lucid — Dream reconstruction and constraint compilation

License

MIT

Related

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

luciddreamer-0.1.0.tar.gz (66.1 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

luciddreamer-0.1.0-py3-none-any.whl (62.4 kB view details)

Uploaded Python 3

File details

Details for the file luciddreamer-0.1.0.tar.gz.

File metadata

  • Download URL: luciddreamer-0.1.0.tar.gz
  • Upload date:
  • Size: 66.1 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.10.12

File hashes

Hashes for luciddreamer-0.1.0.tar.gz
Algorithm Hash digest
SHA256 439857b61db75524707e6f3f1a1de158a41699246c87265d7643b92936fe7e14
MD5 7193da56a7476d1cba2654036e2faa7a
BLAKE2b-256 0939c823470ce24ea14989b65f0181e04d4f7396319fb7869e3851e70bb08185

See more details on using hashes here.

File details

Details for the file luciddreamer-0.1.0-py3-none-any.whl.

File metadata

  • Download URL: luciddreamer-0.1.0-py3-none-any.whl
  • Upload date:
  • Size: 62.4 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.10.12

File hashes

Hashes for luciddreamer-0.1.0-py3-none-any.whl
Algorithm Hash digest
SHA256 ded12698a7ce1f59263307138417d0f01aed25d2a72b5ba0b42cd63231ed215d
MD5 7f687efa24bde788e441ed8351965f1c
BLAKE2b-256 d61ff6652a48e2cd4d9d93ea005aee8bae0643ec1fda41ef014228de3121e4e2

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page