Skip to main content

Collate 2.0: What Changed

Collate 2.0 is a major release. It:
  • Introduces AI Mode.
  • Retires the thread-backed collaboration model.
  • Replaces the Explore experience.
  • Converts the agent apps into AI Automations.
  • Moves LLM and embedding configuration into one place.
  • Reshapes several API contracts.
This section covers everything that can break an existing client, script, deployment or user workflow, broken out by component so you only read the parts that apply to you.

Start here: AI Mode

In 2.0, the headline change is a second, AI-native interface. Every new capability lands there first. Learn what is in it and how to turn it on for yourself, a persona, or the whole tenant.
Take a full database backup before upgrading. The 2.0.0 migration renames tables, creates more than 15 new ones, deletes the three agent applications, and rewrites application, service-connection and tag rows in place. No automated downgrade path exists.

How to read this section

Each entry states what changed, who is affected and what to do.

The changes most likely to affect you

Deprecation notices

These are not 2.0 breaking changes, but they change what you should build on now.
Airflow as the internal orchestrator is deprecated and will be removed in a future release. The native Kubernetes Orchestrator released in 1.12 is its replacement. New deployments should standardise on the Kubernetes Orchestrator. Existing Airflow-based deployments keep working in 2.0 but should plan a migration.
Great Expectations 0.x is no longer supported. 2.0 requires Great Expectations 1.3 or later. See Ingestion & Connectors.

Browse by component

AI Mode

The new AI-native interface, what is in it, and the three ways to enable it.

AI Agents & Automations

Agent apps replaced by AI Automations, Skills, Context Plugins, AI Dashboards and AICharts.

Collate AI & Credits

Conversation sharing and collaborators, plus the reworked credit and billing model.

Discovery & Search

The redesigned Explore page, the browse-and-filter query bar and staged ranking.

Collaboration

Task redesign, Suggestions API removal, standalone Announcements and the Activity Stream.

Data Governance

Policies and task operations, glossary and ontology, classification, workflows, Data Access Requests and the Policy Agent.

Data Quality & Observability

Incidents as Tasks, inlined incident status, auto-close and the removed Data Insights DQ module.

Lineage

FQN-based edge APIs, batch hydration and time-window traversal semantics.

Ingestion & Connectors

pipelineStatuses, Databricks Pipeline auth, log-stream and progress payloads, Great Expectations 1.x and new connectors.

API & Schema

Removed endpoints, changed request and response shapes, tightened validation and enum changes.

Platform, Config & Security

llmConfiguration, session management, database timeouts and the 2.0.0 migration.

Platform Applications

Search Indexing app config, RDF Index defaults, Data Insights module changes and MCP tool contracts.

UI & Customization

Design-system consolidation, per-user preferences and removed components.

Upgrade checklist

1

Inventory your API clients

Search scripts, dashboards, Terraform and CI jobs for /v1/suggestions, /v1/feed, /v1/agents/abilities, /v1/knowledgeCenter, /v1/contextCenter/memories, search_after=, pipelineStatuses and recreateIndex.
2

Record your agent app configuration

The Documentation, Tier and Quality agent apps are deleted during the upgrade. Note their filters and schedules. The migration preserves the filter but the replacement automation arrives disabled and unscheduled.
3

Port your LLM configuration

The elasticsearch.naturalLanguageSearch.* provider blocks no longer exist. Move them to llmConfiguration before you upgrade or semantic search silently degrades.
4

Check entity names for newly invalid characters

>, " and ASCII control characters are rejected by name validation in 2.0.
5

Back up the database and run the upgrade

Expect these data migrations: suggestions and thread tasks into the Task entity, system feed threads into the Activity Stream, announcements into their own entity, agent apps into AI Automations, and the Data Access Request and AutoPilot workflows reseeded.
6

Allow a full search reindex

Reindexing in 2.0 always recreates the index. 2.0 has no incremental mode, and ranking configuration is backfilled during the migration.
7

Re-enable your migrated automations

Open AI Automations, find the entries suffixed (migrated), review the carried-over filter, then enable and schedule each one.
8

Reissue tokens for removed agent bots

CollateAIApplicationBot and the Tier and Quality agent bots are deleted with their apps. Reissue any API or SDK token that used them.
9

Review AI Dashboard visibility

The upgrade sets every dashboard to Public to preserve access. Owners should restrict the ones that should not be org-wide.
10

Rebuild custom ingestion images

Ingestion moves to Python 3.12. cp310 wheels will not load.
11

Decide your AI Mode rollout

Choose whether to leave the tenant default unset, pin a pilot persona to AI, or default everyone to AI Mode. See AI Mode.

Version scope

Where this guide names a file, environment variable, package or HTTP header, it uses the literal identifier as it appears in the product, for example openmetadata.yaml, openmetadata-ingestion or /api/v1/. Those names are unchanged in 2.0 and renaming them here would make the instructions wrong.