# Help
EigenVertex Studio helps humans and agents turn scattered materials into a usable knowledge system.
It lets you collect sources, stage them, build corpora, group them into constellations, and query them with the right retrieval strategy. In short: Studio is where raw material becomes trusted answers, reusable knowledge, and agent-ready infrastructure.
Context
Workspace none
Constellation none
# Get started
Choose your entry point.
Humans work from evidence. Agents work from structure. The product supports both without forcing the same workflow on everyone.
Use Studio as a knowledge cockpit: ask questions, inspect citations, compare sources, and keep human judgment in the loop.
- Ask across corpora or constellations
- Review sources and staging
- Trust citations and precise outputs
Use Studio as an execution layer: shape corpora, enable retrieval strategies, generate projections, and wire knowledge into downstream agents.
- Compose retrieval layers
- Compile Agent / OKF bundles
- Route knowledge into agentic workflows
# Definitions
Corpus, corpora, constellation, sources, staging, ingestion.
These are the core nouns of the app. Once they are clear, Ask, graph views, and ingestion choices become much easier to reason about.
Corpus / corpora
A corpus is one retrieval workspace. Corpora is the plural: several workspaces that can be kept separate or grouped together.
Constellation
A constellation is a curated group of related corpora. It lets Ask widen the search scope while keeping the same user experience.
Sources
Sources are the original materials: PDFs, URLs, notes, scans, emails, audio, and anything else that can be ingested or staged.
Staging
Staging is the review layer before ingestion or delivery. It is where you inspect, validate, and decide what should enter the corpus.
Ingestion
Ingestion is the process that turns a source into searchable, structured, and traceable corpus material.
# OKF projections
A portable knowledge pack for agents.
OKF is a compiled projection, not a second source of truth. It packages canonical knowledge, provenance, evidence references and machine-readable indexes into a versioned bundle that can be downloaded, shared with an agent or published to a controlled external destination. It is useful when an agent needs a stable snapshot; live Project Ask should continue to use the current corpus, vectors, Wiki and temporal graph.
# Project map
Read the project from its facts or from its sources.
Semantic
The default view. It focuses on the people, decisions, milestones, work and relationships that describe the project.
Sources & provenance
Use this when you need to check where a fact came from. It adds the source records and their links to the map.
Current shows facts that are still valid. Current & history also includes earlier facts, so you can understand how the project evolved.
# Ingestion lifecycle
From source to searchable corpus.
01
Bring a source into the workspace.
02
Stage it if human review is needed.
03
Ingest it into the retrieval corpus.
04
Index it with the right strategy layers.
05
Expose it in Ask, Graph, or downstream surfaces.
# Ingestion strategies
The strategy changes how the corpus is built and queried.
Vector / Qdrant
Best for dense semantic retrieval, citation-bearing chunks, and general ask flows.
Documentary GraphRAG
Best when relationships between documents, chunks, and entities matter.
Flat search
Best for directory-like sources where the file tree itself is already meaningful.
LLM-Wiki / OKF
Best for curated knowledge pages, long-form references, and stable prose answers.
Temporal memory
Best for project or session memory that evolves over time rather than static reference content.
# Constellations
Use a constellation when related corpora should answer together.
A constellation groups several corpora from the same domain or workflow. In Ask, it lets the runtime default to a broader but still curated scope when one corpus alone is not enough.
Practical rule
If the corpus belongs to a constellation, Ask should usually start from the constellation scope and let the user narrow down only when needed.
# Query scope
Corpus scope is narrow. Constellation scope is grouped.
Corpus
Best when you know the exact corpus you want and want to keep retrieval tight.
Constellation
Best when the answer may live in another corpus of the same family, such as manufacturer history plus tuning references.