¶ Knowledge Resources
Resources are the foundational building blocks of the Knowledge Framework in Fuse. They represent structured, composable units of knowledge that support real-time access, embedding, search, and AI-driven orchestration.
¶ Overview
In Fuse, a Resource encapsulates any form of content — whether static or dynamic — and makes it available in a secure, governable, and reusable format across workflows, AI agents, pipelines, and external systems.
Resources may represent:
- Files (e.g., documents, PDFs, CSVs)
- URLs (external links or APIs)
- Raw content (inline text or instructions)
- Workflows (pipelines that generate data on execution)
- Knowledge systems (e.g., Wiki.js, SharePoint)
All Resources are interoperable across Fuse domains and can expose runtime methods, caching behaviors, trust scores, and access control enforcement.
¶ Key Capabilities
| Capability | Description |
|---|---|
 Unified Knowledge Abstraction |
Encapsulate diverse content sources under a consistent model |
 Runtime Injection |
Deliver content to AI prompts, workflow steps, or external calls |
 Composable Graph Integration |
Serve as the source for pages, embeddings, nodes, and links |
 Scoped Access Control |
Enforce identity-aware permissions and domain-level security |
 Versioning & Trust |
Track changes, trustworthiness, and audit trails for every resource |
 Real-Time Behavior |
Support dynamic retrieval from APIs, workflows, or providers |
¶ Resource Types
| Type | Description |
|---|---|
File |
A file stored in Fuse (e.g., PDF, DOCX, CSV) |
URL |
External web resource or API endpoint |
Content |
Raw inline content entered by a user or system |
Workflow |
A pipeline that generates data when invoked |
Knowledge |
A dynamic provider-backed knowledge source (e.g., Wiki.js) |
Each type may define custom runtime methods (e.g., search, getLatest) depending on source capability.
¶ Interoperability with Other Capabilities
Resources serve as the entry point to the rest of the Knowledge Framework:
- Feed KnowledgeBases for embedding, graphing, and filtering
- Generate KnowledgePages, KnowledgeNodes, and KnowledgeLinks
- Drive Prompt Context during RAG workflows
- Enable graph traversal starting from a document, query, or concept
- Allow assistants to access live or cached results based on trust and access
¶ Access & Governance
- Access is enforced per Resource, per user/session, and per consumer (e.g., AI assistant)
- Permissions use
ClientMethod,fnGetOperations, andResourceAccessRulelogic - Trust scores and visibility (public, private, tenant-scoped) are configurable
- All interactions with Resources are logged and auditable
- Versioning supports traceable knowledge use and answer attribution
¶ Composability & Reuse
Resources can be:
- Linked to multiple KnowledgeBases
- Referenced in multiple workflows or AI assistants
- Used in prompt injection with fallback or summary behavior
- Composed dynamically (e.g., chaining multiple sources via a Workflow Resource)
¶ Example Scenarios
- Use a
WorkflowResource to retrieve data from multiple systems for each assistant query - Wrap a
KnowledgeResource pointing to Wiki.js and exposesearch()for dynamic prompt support - Store internal SOPs as
FileResources, versioned and trust-scored, used across multiple departments - Combine multiple
ContentResources into a single KnowledgeBase for semantic embedding
¶ Future Enhancements
- Dynamic resource chaining and resolution rules
- Resource fallback strategies by assistant role or trust level
- Integrated preview and transformation UI
- Graph overlays per Resource for explanation or tagging
Next: Knowledge Embeddings — Explore how Fuse uses embeddings to power semantic retrieval, ranking, and injection during inference.