Pipeline topology, resource graph, and core traits for the ECL pipeline runner.

This crate defines:

• The resolved pipeline topology (PipelineTopology, ResolvedStage)
• Resource graph computation and parallel schedule derivation
• Core traits (SourceAdapter, Stage) and their supporting types (PipelineItem, SourceItem, ExtractedDocument, StageContext)

The topology is computed from a PipelineSpec (from ecl-pipeline-spec) at init time and is immutable during execution.

Textrynum

A workspace and tools for weaving knowledge into a hyper-connected, searchable fabric

What is Textrynum?

Textrynum is a workspace for building knowledge systems. It houses two complementary layers:

1. Fabryk — A modular knowledge fabric: ingest content, build knowledge graphs, index for full-text and semantic search, and serve it all via MCP tools. Production-ready, 24,000+ lines, 20 crates.

2. ECL (Extract, Cogitate, Load) — A workflow orchestration engine for durable AI agent pipelines with feedback loops, validation, and managed serialism. Early stage, building on solid foundations.

(generated with Dendryform)

Using Fabryk in Your Project

Fabryk is designed around two umbrella crates. Most projects need only one or two [dependencies] lines:

[dependencies]
# Core knowledge fabric — content, graph, search, auth, acl
fabryk = { version = "0.2", features = ["full"] }

# MCP server tools — all tool suites + server infrastructure
fabryk-mcp = { version = "0.2", features = ["http"] }

What each umbrella includes

fabryk provides the core library:

fabryk-mcp provides the MCP toolkit:

Vendor-specific crates are added separately as needed:

fabryk-auth-google = "0.2"  # Google OAuth2 / JWKS
fabryk-gcp = "0.2"          # GCP credential detection

See the Fabryk README for the full crate map and feature flags.

HOWTOs

Step-by-step guides for common integration tasks with Fabryk MCP servers:

• Connecting Fabryk MCP Servers to Claude Code — Set up Claude Code to talk to your Fabryk MCP server over Streamable HTTP
• MCP Async Startup Pattern — Start MCP transport instantly while expensive services initialize in the background
• MCP Health Endpoint How-To — Add a service-aware /health endpoint to your Fabryk MCP server
• MCP Metadata & Discoverability How-To — Give AI agents rich metadata when connecting to your MCP server

Project Status

v0.2.0 — Fabryk is functional; ECL is in progress.

Completed

• Knowledge graph with traversal algorithms (fabryk-graph)
• Full-text search with Tantivy backend (fabryk-fts)
• Vector/semantic search with LanceDB (fabryk-vector)
• Markdown content parsing and frontmatter extraction (fabryk-content)
• MCP server infrastructure and tool suites (fabryk-mcp-*)
• OAuth2 authentication with Google provider (fabryk-auth-*)
• CLI framework with graph commands (fabryk-cli)
• Configuration infrastructure with TOML support
• Restructured crate hierarchy with two clean umbrella crates
• CI/CD pipeline

In Progress

• ECL workflow primitives
• Step abstraction layer with feedback loops
• LLM integration
• Connecting ECL workflows to Fabryk persistence

Planned

• Additional MCP tool suites
• Example workflows
• Published crate documentation

Getting Started

Prerequisites

• Rust 1.85+

Building

git clone https://github.com/oxur/textrynum
cd textrynum
cargo build

Testing

cargo test --workspace --all-features

Contributing

We're not yet accepting external contributions, but will open the project once the core architecture stabilizes.

License

Apache-2.0