SLURP is a comprehensive system designed for managing, ingesting, normalizing, and structuring knowledge within advanced AI and infrastructure projects, especially those utilizing the UCXL framework. Its primary purpose is to act as a context curator and steward of contextual metadata, enabling semantic use, provenance tracking, decision reasoning, and knowledge synthesis across multiple projects and domains.

**Curator + policy gate.** SLURP merges RAG outputs, reasoning chains, and API traces into citation-backed bundles, then **filters by role/business-intent** so each agent receives the right slice at the right time — with justifications attached.

SLURP manages artefacts and decision records (DRs) submitted by AI agents when they complete tasks. It enforces provenance by maintaining citations, reasoning chains, and decision history, and it prepares context for both human interpretation and agent reasoning. SLURP plays a crucial role in content delivery, ensuring the right context reaches the right agent at the right time, while maintaining permissions, need-to-know policies, and audit trails through role-aware filtering and access control.

Functionally, SLURP interacts with several components. It ingests contextual data from sources like files, APIs, and decision nodes, and manages dependencies, constraints, and metadata associated with files and decisions. It also integrates with tools like BUBBLE for citation tracking and provenance graphs, and supports decision publishing, validation, and recovery. It communicates with other system components such as COOEE for messaging, SHHH for secrets management, and WHOOSH for search indexing, playing an integral role in building and maintaining a structured, traceable, and semantic knowledge graph within the ecosystem.

# Feeding HMMM safely

SLURP emits **scope-clean** thought bundles for HMMM so agents can critique/share intermediate reasoning _without_ cross-project contamination; HMMM writes back signed notes that BUBBLE can cite.


References:

- [KG] Entity 1, Description, and Relationships
- [DC] Document 1 content describing SLURP's purpose, features, and interactions

## TODO

- Leader-only mode: Keep SLURP embedded in CHORUS and codify leader-only execution via elections; add guards and tests.
- BUBBLE integration: Use `/decision/bundle` in the live flow; integrate role filtering and policy enforcement.
- DHT writes: Store curated bundles with role-based encryption and verify end-to-end retrieval.
- Temporal analysis: Elevate decision-hop analysis from demos/scripts into an operational component with APIs.