continuum

Deterministic state machine and multi-agent narrative execution pipeline with GitOps cascades, RAG vector memory, and asynchronous asset synthesis.

Continuum Banner

Deterministic state machine and multi-agent narrative execution pipeline with GitOps cascades, RAG vector memory, and asynchronous asset synthesis.

Version 0.1.0 Crystal >= 1.21.0 GitLab CI Passing Specs Passing Ameba Clean Flaw Clean

Sphinx Documentation Crystal API Documentation License: AGPL-3.0-or-later Donate using Liberapay

GitLab Project Metadata:

  • Description: Deterministic state machine and multi-agent narrative execution pipeline with GitOps cascades, RAG vector memory, and asynchronous asset synthesis.
  • Topics / Tags: crystal, crystal-lang, story-engine, multi-agent, dag, gitops, rag, cli, podman, ameba, flaw, sphinx, agplv3

Overview

Continuum is a compiled, high-performance deterministic story engine and multi-agent execution pipeline written in Crystal. It acts as an immutable physics orchestrator, semantic memory manager, and Git repository controller, coordinating a Directed Acyclic Graph (DAG) of LLM subagents for continuity analysis, research, narrative drafting, asynchronous visual asset synthesis, and adversarial quality assurance.

Key Highlights

  1. Deterministic State Machine: Mutex-protected YAML/JSON state management with path-sandboxed, delta-aware mutations (add, subtract, set) committed only after passing adversarial QA.
  2. Multi-Agent Execution DAG: Coordinates Continuity, Researcher (Google Search + RAG vector cosine similarity), Writer, Asset Synthesis, and Auditor agents with dynamic temperature escalation.
  3. GitOps Retroactive Cascades: Treats historical timeline divergence like a Git rebase—branching into isolated cascade-<slug> branches, restoring state checkpoints, and recompiling downstream chapters.
  4. Asynchronous Asset Synthesis: Fiber-based background rendering pipeline synchronized via WaitGroup barriers for local diffusion workflows (sdcli.py / ComfyUI).
  5. Zero-Defect Quality Gates: Enforces 100% spec verification, Ameba static analysis, Flaw SAST security scanning, and cRSTLint documentation validation.

Quickstart

Prerequisites

  • Crystal: >= 1.21.0
  • Shards: Bundled with Crystal
  • GNU Make & Git
  • Podman (optional, for container runs)
  • cRSTLint (for documentation linting)

Build and Run Locally

# 1. Clone repository
git clone https://gitlab.com/renich/continuum.git
cd continuum

# 2. Setup dependencies and pre-commit hooks
make setup

# 3. Build the binary
make build

# 4. Run application CLI
./bin/continuum --help

Usage Examples

# Initialize a new story workspace
./bin/continuum init my_story
cd my_story

# Generate a narrative scene from timeline or prompt
continuum generate --prompt "Describe the discovery of the monolith."

# Trigger a retroactive multiversal recompilation cascade
continuum cascade --event=slug-of-the-divergence-event

# Validate or format project YAML/JSON state and timeline files
continuum data validate
continuum data format

# Compile master outline from timeline ledger
continuum outline

Running with Podman

# Build container image with Podman
make containers-build

# Run containerized CLI
make containers-run

Build System Targets

The repository includes a comprehensive, standard GNUmakefile:

Target Description
make build Compiles development binary in bin/continuum
make release Builds stripped, optimized production binary
make static Builds static binary using Alpine container in Podman
make test Executes the complete spec test suite (crystal spec)
make lint Runs Ameba static analysis and Flaw SAST security scanner
make doc-check Lints all reStructuredText files using crstlint and rstcheck
make check Aggregates test, lint, and doc-check
make install Installs binary to $(DESTDIR)$(bindir) (FHS compliant)
make install-local Installs binary to ~/.local/bin/continuum
make containers-build Builds rootless container image using Podman
make clean Cleans build artifacts in bin/
make distclean Removes dependencies, cache directories, and build outputs

Documentation

Full architectural blueprints, functional specifications, and operational playbooks are maintained under docs/:

Guide Scope & Highlights Entry Point
Business Context & Strategy Stakeholder requirements, ROI objectives, and user personas. Business Specs
Functional Specifications Requirements [FUNC-001]–[FUNC-006] and behavioral contracts. Functional Specs
Technical Architecture DAG execution pipeline, state mutations, RAG memory, and cascades. Technical Specs
Release Playbook SemVer, Keep a Changelog, and GPG-signed GitLab release steps. Release Procedure
Architecture Decisions (ADRs) Immutable logs of architectural decisions and trade-offs. ADR Index
Project Roadmap Phased milestone tracking and delivery roadmaps. Project Roadmap
User Guide Timeline schemas, cascade workflows, and CLI command reference. User Guide
Crystal API Reference Compiler-generated API reference. API Reference

Contributing & Code of Honor

All contributions must adhere to the Universal Code of Honor and Contributing Guidelines.


License

  • Software: GNU Affero General Public License v3.0 or later (LICENSE).
  • Documentation: GNU Free Documentation License v1.3 or later (LICENSE-DOCS).

Copyleft © 2026 Rénich Bon Ćirić renich@evalinux.com.


Support & Donations

If you find this project useful and wish to support its ongoing development, please consider donating:

Donate using Liberapay

Repository

continuum

Owner
Statistic
  • 0
  • 0
  • 0
  • 0
  • 3
  • about 1 hour ago
  • October 15, 2026
License

GNU General Public License v3.0 or later

Links
Synced at

Thu, 15 Oct 2026 14:23:20 GMT

Languages