collection_naming_v1 — Trellison Institute

Format

{domain}_{entity}

Separator: _

Examples

corpus_persons

corpus_documents

corpus_atlas

ops_executions

tools

Domains

corpus

World data — everything we know. People, places, organizations, documents, research, government data, financial data, patents, education, entities. The information itself.

Any data that represents knowledge about the real or virtual world. If you could query it to answer a question about reality, it belongs here.

corpus.persons — Unified person profiles (identity-resolved from all sources)

corpus.documents — All documents, research, intelligence, knowledge assets

corpus.atlas — World data — World Bank, IMF, census, county, state, federal datasets

corpus.atlas_catalog — Schema registry for atlas data sources with parameter mappings

corpus.education — Schools, districts, enrollment, assessments, graduation, civil rights

corpus.entities — Government entities, local entities, organizations, companies

corpus.patents — Patent families, claims, prior art, FTO, moat strategies, filings

corpus.ventures — Venture profiles, evaluations, lifecycle, assessments, org charts

corpus.social — LinkedIn activity, outreach, social media, network data

corpus.web — Pages, templates, layouts, themes, share sites, DSR content

corpus.har — HAR captures, endpoints, journeys, site profiles, auth patterns

corpus.catalogs — Product catalogs, dataset catalogs, model catalogs, inventories

corpus.grants — Grant opportunities, applications, funder profiles, alignments

corpus.comms — Email, messages, notifications, correspondence, campaigns

tools

The capability layer — what we can do. Tool registry is a single collection. Capabilities link tools to what they enable.

Tool definitions, capability registry, tool KPIs, tool healing records. NOT execution records (those are ops).

tools — The tool registry (existing, stays as-is)

tools.capabilities — Capability registry linking tools to ventures and gaps

tools.gaps — Detected capability gaps

tools.audit — Capability extraction audit trail

analyses

Outputs from applying tools to corpus data. Investigations, 10-K reports, performance reports, evaluations, assessments. The conclusions.

Any document that represents an analytical output — something generated by examining corpus data. If a human or model drew a conclusion, it belongs here.

analyses.all — Unified analyses — tagged by analysis_type (agi_evaluation, financial_analysis, content_quality, governance, outcome_analysis, investigation, etc.)

connectors

Data acquisition infrastructure — how we get data into the corpus. Harvest configs, pipeline state, ingestion logs, connector configs.

Anything related to acquiring external data. Connector configs, harvest jobs, pipeline DAGs, ingestion runs. The plumbing.

connectors.all — Unified connectors — tagged by connector_type (harvest, pipeline, ingestion, connector_config, harvested_model, harvested_source, harvested_tool)

ops

Operational machinery — execution records, traces, logs, configs, sessions, caches, security. The runtime.

System state that supports execution but is not corpus data, not analysis output, and not tool definition. Logs, configs, sessions, execution records.

ops.executions — Tool execution records

ops.traces — Execution traces and spans

ops.security — Security audit log, alerts, events, controls

ops.sessions — User sessions, chat sessions, auth sessions

ops.configs — System configs, environment configs, feature flags

ops.caches — Cache tables, materialized views

ops.uptime — Health checks, uptime monitoring

ops.threads — Dev threads, thread learnings, agent tasks

ops.selection — Model selection events, MAB state

archived

Superseded collections. Data migrated to a corpus/tools/analyses/connectors/ops collection. Prefixed with archived_ for discoverability.

Any collection whose data has been fully migrated to a unified collection. Never deleted — renamed with archived_ prefix and migration metadata.

Enforcement

method: contract_precondition
description: Any tool that creates or renames a collection must validate the name against this standard. The contract evaluator checks the domain prefix before allowing execution.
violations: {'wrong_prefix': 'Collection name does not match any defined domain prefix', 'missing_provenance': 'Migrated documents missing required provenance fields', 'orphan_collection': 'Collection exists outside any domain prefix and is not archived'}