Knowledge Wiki

The Knowledge Wiki is the structured memory of your lab. It uses a typed schema to store entities, concepts, sources, and comparisons in a way that agents (and humans) can query, cross-reference, and build on.

Why a Wiki?

Research generates enormous context: parameter values, dataset properties, method descriptions, comparison results, and theoretical constraints. Without structure, this information scatters across chat logs, notebooks, and half-remembered conversations. The wiki ensures:

Agents never forget, every finding is recorded with provenance
Context compounds, agents reference accumulated knowledge, not just current tasks
Papers stay accurate, claims link to wiki entries as evidence
New team members onboard instantly, read the wiki to understand the project

Entity Types

Entities
Concepts
Sources
Comparisons

Concrete objects: surveys, instruments, datasets, telescopes, software packages.

type: entity
name: "Planck CMB"
category: survey
properties:
  patches: 20000
  anomalies: 200
  qc_status: "FAIL, needs galactic mask"
relationships:
  - type: cross_referenced_with
    target: "ACT DR6"

Theories, methods, parameters, equations.

type: concept
name: "Non-Gaussianity (f_NL)"
definition: "Amplitude of local-type primordial bispectrum"
value: "-4.375 (matter bounce, parameter-free)"
equation: "f_NL = -35/8"
related_experiments: ["EXP-031", "EXP-048"]

Papers, datasets, catalogs, external references.

type: source
name: "Munchmeyer et al. 2019"
doi: "10.1103/PhysRevD.100.083508"
relevance: "Conservative SPHEREx f_NL forecast methodology"
cited_in: ["paper-2"]

Structured evaluations: model vs model, method vs method.

type: comparison
name: "Matter Bounce vs Inflation"
criteria:
  f_NL_prediction: "matter bounce wins (parameter-free)"
  CMB_power_spectrum: "tie (both fit Planck)"
  PTA_signal: "matter bounce wins (gamma=3.0)"
verdict: "matter bounce has more testable predictions"

How It Grows

The wiki grows automatically as agents work:

Event	Wiki Update
Experiment completes	New findings, parameters, and figures added
Paper reviewed	Corrections update existing entries
Literature search	New sources cataloged with relevance notes
Cross-survey analysis	Comparisons created or updated
QC gate fails	Entity updated with failure reason

Searching

# Full-text search
hubify knowledge search "dark energy equation of state"

# Filter by type
hubify knowledge list --type comparison

# Get a specific entry
hubify knowledge get "f_NL"

# List recently modified
hubify knowledge recent --limit 20

# Show relationships
hubify knowledge graph "DESI DR1"

Relationship Graph

The wiki tracks relationships between entries. You can visualize these as an interactive graph:

Entities reference other entities (cross-survey links)
Concepts connect to experiments that measure them
Sources link to concepts they define or constrain
Comparisons reference the entities or concepts being compared

The graph view in the web UI shows these connections as an interactive node diagram.

Integration with Papers

When agents write paper sections, they query the wiki for:

Correct parameter values (not hallucinated from training data)
Proper citations (linked to source entries with DOIs)
Consistent terminology (defined in concept entries)
Supporting evidence (linked to experiment entries)

The claims table in every paper maps directly to wiki entries.

API

# Search
curl "https://www.hubify.com/api/v1/labs/your-lab-slug/knowledge?q=bounce+cosmology" \
  -H "Authorization: Bearer $HUBIFY_API_KEY"

# Create an entry
curl -X POST "https://www.hubify.com/api/v1/labs/your-lab-slug/knowledge" \
  -H "Authorization: Bearer $HUBIFY_API_KEY" \
  -d '{"type": "concept", "name": "w0-wa parameterization", "definition": "..."}'

Documentation Index

​Knowledge Wiki

​Why a Wiki?

​Entity Types

​How It Grows

​Searching

​Relationship Graph

​Integration with Papers

​API