Skip to content

Latest commit

 

History

132 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Elixis

License: MIT Python 3.12 CI GitHub stars

A local-first AI pattern synthesis engine for turning raw references, repositories, and folders into Source Corpora, Market Kits, identity, brand voice, design systems, and naming direction.

Elixis is a cross-domain synthesis engine. It transforms raw brain dump text, GitHub repositories, and local folders — references, influences, values, works, people, places, aesthetics, constraints, documentation, metadata, and public interfaces — into a curated Source Corpus and pattern graph, then resolves that graph into usable market, brand, design, naming, and identity outputs. The identity lens generates SOUL.md documents for OpenClaw and Soul Spec, but SOUL.md is one output mode, not the whole scope.

TL;DR

Elixis takes a messy constellation of references or a real product repository and turns it into usable structure: Corpus Signals, beads, threads, archetypal patterns, tensions, and resolved outputs. Use Source Corpus ingestion to gather quality-ranked evidence, Market Kit orchestration for naming/positioning/messaging/design direction, the brand lens for voice guidelines, the design lens for design tokens and principles, the naming tools for semantically aligned names, or the identity lens for SOUL.md. It is bigger than identity-file generation: SOUL.md is a compatibility surface, while the core product is a reusable synthesis engine.

AI discovery: llms.txt provides a compact project summary for AI assistants and search crawlers.

Public landing page: https://kyanitelabs.github.io/Elixis/

Canonical Names

Elixis has one active product and package name:

Surface Name
Product and repository Elixis
Python distribution, import package, CLI, Docker service elixis
Compatibility targets, not package names OpenClaw, SoulClaw, Soul Spec, SOUL.md

Historical planning notes are preserved under archive/ for continuity only. They do not define the current package, product, or public positioning.

What it does

Phase / Surface Function
Declaration Parses references, people, works, concepts, and values into beads
Elaboration Enriches beads with local knowledge and external context
Connection Builds threads, cross-domain bridges, tensions, and archetypal patterns
Resolution Uses lenses to generate identity, brand, design, or future output forms
Naming Generates and scores names against the same pattern graph
Source Corpus Ingests GitHub repositories or local folders into quality-scored evidence
Market Kit Orchestrates naming, positioning, brand, marketing, and design direction
Validation Input sanitization, prompt-injection filtering, output validation
Traces & Backups Preserves synthesis runs and enables rollback

Repository map

elixis/              Core Python package
  __main__.py           Operator CLI: serve, run, extract, patterns, name, ingest, corpus, mcp
  entities.py           Entity extraction and structured terms
  engine.py             Four-phase pattern synthesis orchestrator
  lenses/               Output lenses: identity, brand, design
  patterns.py           Archetypal pattern discovery and graph
  synthesis.py          SOUL.md identity lens implementation
  translate.py          Text localization helpers and language detection
  research.py           Wikipedia entity enrichment
  naming.py             Name research and variant generation
  llm.py                LLM interface (Ollama / OpenAI-compatible / Anthropic)
  validation.py         Input sanitization and validation
  ingest.py             Source Corpus ingestion for GitHub repositories and local folders
  market.py             Market Kit orchestration from Source Corpus evidence
  traces.py             Run trace capture and diagnostics
  backup.py             Backup and restore helpers
  logging_config.py     Structured logging setup
  mcp_server.py         Optional stdio adapter for MCP-compatible assistants
  templates/            Landing page UI with live multi-lens demo
tests/                  Unit and integration tests
app.py                  HTTP server entry point (stdlib, port 3110)
openapi.yaml            HTTP API contract

Quick start

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
python app.py

Operator surfaces

MCP is an integration adapter, not the product scope. The supported operator contract is the same engine behind three local-first surfaces:

Surface Command / contract Use
CLI python -m elixis run --text "Athena, Batman, Musashi" --lens brand Local scripting and one-shot synthesis
HTTP API python -m elixis serve --port 3110; see openapi.yaml Browser UI, services, and external clients
MCP stdio python -m elixis mcp AI assistants that need tool access to the same engine

Useful CLI commands:

python -m elixis --version
python -m elixis serve --port 3110
python -m elixis run --file references.txt --lens design --json
python -m elixis extract --text "Athena, Batman, Miyamoto Musashi"
python -m elixis patterns --stdin < references.txt
python -m elixis name --name Elixis --context "local AI synthesis tool"
python -m elixis ingest --path . --kit --artifact markdown
python -m elixis ingest --github https://github.com/KyaniteLabs/Elixis --kit --artifact html
python -m elixis corpus inspect <run_id>
python -m elixis mcp

Source Corpus and Market Kits

Elixis can point at a GitHub repository or local folder and build a Source Corpus: a curated, auditable evidence bundle selected because it improves output quality. The corpus is not a raw dump of every readable file. Each Corpus Signal is scored for relevance, distinctiveness, authority, lens utility, provenance, freshness, and noise risk.

The resulting Ingestion Result is the shared structured contract across CLI, HTTP API, and MCP. It contains the Source Target, included and rejected signals, scoring summary, process trace, and, when requested, a Market Kit with naming directions, positioning, audience/category notes, brand voice, messaging pillars, landing-page copy angles, color palettes, typography, spacing, borders, shadows, visual motifs, and design-system direction.

HTML/CSS are optional artifact tiers for high-quality human review. They are generated only when explicitly requested or when visual presentation is critical; structured JSON remains the source of truth.

Configuration

Copy the example environment file and set provider/runtime values as needed:

cp .env.example .env

For GLM 5.1 cloud inference, use the OpenAI-compatible provider shape:

LLM_PROVIDER=openai
LLM_BASE_URL=https://api.z.ai/api/paas/v4
LLM_MODEL=glm-5.1
ZAI_API_KEY=...

Development

pip install -r requirements.txt
python -m unittest discover tests/ -v

Best for

  • Mapping a constellation of cultural references into beads, threads, patterns, and tensions.
  • Turning a GitHub repository or local folder into a quality-scored Source Corpus.
  • Generating a Market Kit with naming, positioning, messaging, brand, and visual design direction.
  • Resolving the same pattern graph into identity, brand voice, design direction, or naming options.
  • Creating AI agent personas for OpenClaw, SoulClaw, or any SOUL.md-compatible system.
  • Inspecting process evidence: model, phases, pattern scores, entity support, bridges, timings, and fallbacks.

Works With

System How
OpenClaw Drop SOUL.md into ~/.openclaw/workspace/
SoulClaw Native Soul Spec v0.5 support
Claude Code Paste into ~/.claude/CLAUDE.md
Cursor Paste into .cursorrules
Windsurf Paste into .windsurfrules
Any LLM Prepend SOUL.md to system prompt

FAQ

What is Elixis? Elixis is a local-first AI pattern synthesis engine. It turns raw reference material into a pattern graph and resolves that graph through lenses for identity, brand voice, design systems, naming research, and lens-specific outputs.

How do I create an output? Start the server with python app.py, open http://localhost:3110, enter references and influences, choose a lens, and run the synthesis. The identity lens generates SOUL.md; the brand and design lenses generate different documents from the same graph.

What LLM providers does Elixis support? Elixis works with local LLMs via Ollama, OpenAI-compatible APIs, and Anthropic's Messages API. It falls back to template-based synthesis when no LLM is available.

Is Elixis free and open source? Yes. Elixis is MIT-licensed and uses zero external Python dependencies (stdlib only).

What is the Soul Spec? Soul Spec is an open standard for AI agent personas. Elixis can generate SOUL.md documents compatible with Soul Spec v0.5, OpenClaw, and any framework that reads markdown-based identity files, but Elixis is not limited to identity files.

License

See LICENSE.

What is Elixis?

Elixis is a product/tooling surface in the Kyanite Labs stack that helps builders using Kyanite Labs tooling use the Elixis product surface with clear install and intent.

Product Elixis
Category product/tooling surface in the Kyanite Labs stack
Best for builders using Kyanite Labs tooling
Not unrelated to Kyanite Labs
Source GitHub · Forgejo
Keywords Elixis, Kyanite Labs tool

Who it's for

  • Primary: builders using Kyanite Labs tooling
  • Use when you need to use the Elixis product surface with clear install and intent
  • Skip if you need unrelated to Kyanite Labs

FAQ

What is Elixis?

Elixis is a product/tooling surface in the Kyanite Labs stack. It helps builders using Kyanite Labs tooling use the Elixis product surface with clear install and intent.

Who should use Elixis?

builders using Kyanite Labs tooling.

How is Elixis different?

See README body for how Elixis differs from adjacent tools in the stack.

Is Elixis production software?

Treat the README status and release tags as source of truth for maturity. Validate against your own requirements before production use.

Status

  • Maintained as of 2026 on the default branch
  • Prefer release tags when pinning dependencies
  • Report issues on the canonical remote listed above

Agent surface

  • Coding agents: read this README first, then repo docs/AGENTS.md if present
  • Prefer machine-readable briefs (llms.txt) when the repo ships one
  • MCP or skill entrypoints are documented in-repo when applicable

Contributing

Issues and PRs welcome on the canonical remote. Keep public docs free of secrets and machine-local paths.

License

See LICENSE in this repository (or package metadata if license is package-only).

Releases

Packages

Contributors

Languages