docs: update roadmap and distributions for April 2026 (spiceai#10367)

* docs: update roadmap and distributions for April 2026

- Add Schema Evolution to v2.1 roadmap.
- Add OIDC Token Verification to DISTRIBUTIONS.md (Cloud + Enterprise).
- Link Spice.ai Enterprise mentions to https://docs.spice.ai/docs/enterprise.

* docs(criteria): audit and normalize release criteria

- Fix naming: Spice.AI -> Spice.ai, Azure Openai -> Azure OpenAI.
- Fix typos: Embdedded -> Embedded, 'connector' -> 'accelerator' in accelerators/stable.md, 'Component ensure' -> 'Components ensure'.
- connectors/beta.md: correct schema-inference legend (add native classification).
- connectors/stable.md: rename 'Schema Detection' header to 'Schema Inference' for consistency.
- connectors: add ADBC, Azure BlobFS (sorted), Debezium, DuckLake, GCS, MongoDB, NFS, ScyllaDB, SMB rows across all tiers; restore Iceberg, Glue, Kafka, Oracle dropouts in beta/rc/stable; alphabetize all tables.
- catalogs: add ADBC, Cayenne, DuckLake, MS SQL, MySQL, Oracle, PostgreSQL, Snowflake rows across all tiers; restore Glue in beta/rc/stable; alphabetize.
- models: add AWS Bedrock, Databricks, Google (Gemini); restore Nvidia NIM in beta/rc/stable; alphabetize.
- embeddings: add AWS Bedrock, Databricks; remove xAI (no embeddings API); alphabetize.
- grading.md: fix heading hierarchy (Grade C/B/A as H3) and naming consistency.
- PRINCIPLES.md: expand to clarify scope and interaction between component and feature criteria.
- definitions.md: add SF5 to TPC-H and TPC-DS parallel-query tables; clarify stop-light percentile boundaries.
- features/README.md: add H1 title and reference to PRINCIPLES.md.

* docs(criteria): require Deployment Guide for GA/Stable components

Adds a Deployment Guide requirement to Stable release criteria for connectors, accelerators, catalogs, models, embeddings, and generic features. References the Databricks Deployment Guide as a canonical example: https://spiceai.org/docs/next/components/data-connectors/databricks/deployment

* docs(criteria): add Databricks sql_warehouse mode

Tracks the sql_warehouse mode alongside delta_lake and spark_connect across alpha/beta/rc/stable connector criteria tables. New mode defaults to not-started (no DRI).

* docs(criteria): rename AWS Bedrock to Amazon Bedrock

Use the official 'Amazon Bedrock' branding. Anthropic and Amazon Bedrock are both listed across models and (Bedrock) embeddings tiers.

* docs(criteria): standardize accelerator tables format across all criteria documents

* docs(criteria): add SpiceBench criteria for accelerators and connectors

Author: lukekim

docs/DISTRIBUTIONS.md

@@ -4,7 +4,7 @@ The Spice open source project provides multiple distribution variants to support
 
 > **Note:** The Spice runtime is **64-bit only**. 32-bit platforms are not supported.
 
-> **Note:** Variant distributions (data, allocators, CUDA) are only available in **nightly images** for the open source project. All features and distributions are available in the [Spice Cloud Platform](https://spice.ai/pricing) and [Spice.ai Enterprise](https://spice.ai/pricing).
+> **Note:** Variant distributions (data, allocators, CUDA) are only available in **nightly images** for the open source project. All features and distributions are available in the [Spice Cloud Platform](https://spice.ai/pricing) and [Spice.ai Enterprise](https://docs.spice.ai/docs/enterprise).
 
 ## Supported Platforms & Hardware Requirements
 
@@ -19,15 +19,16 @@ The Spice open source project provides multiple distribution variants to support
 
 ## Distribution Availability
 
-| Distribution / Variant | Open Source      | Spice Cloud | Enterprise |
-| ---------------------- | ---------------- | ----------- | ---------- |
-| Default (Data + AI)    | ✅                | ✅           | ✅          |
-| Data-only              | Nightly only     | ✅           | ✅          |
-| NAS (SMB + NFS)        | Nightly only     | ❌           | ✅          |
-| Metal (macOS)          | ✅                | ✅           | ✅          |
-| CUDA (Linux)           | Nightly only     | ✅           | ✅          |
-| Allocator variants     | Nightly only     | ✅           | ✅          |
-| ODBC connector         | Local build only | ✅           | ✅          |
+| Distribution / Variant  | Open Source      | Spice Cloud | Enterprise |
+| ----------------------- | ---------------- | ----------- | ---------- |
+| Default (Data + AI)     | ✅                | ✅           | ✅          |
+| Data-only               | Nightly only     | ✅           | ✅          |
+| NAS (SMB + NFS)         | Nightly only     | ❌           | ✅          |
+| Metal (macOS)           | ✅                | ✅           | ✅          |
+| CUDA (Linux)            | Nightly only     | ✅           | ✅          |
+| Allocator variants      | Nightly only     | ✅           | ✅          |
+| ODBC connector          | Local build only | ✅           | ✅          |
+| OIDC Token Verification | ❌                | ✅           | ✅          |
 
 ## Default Distribution
 
@@ -61,7 +62,7 @@ docker pull spiceai/spiceai:latest
 
 The data distribution excludes AI/ML model support, resulting in a smaller binary size and reduced attack surface. Use this when you only need data federation and acceleration capabilities without AI features.
 
-> **Open Source:** Available in nightly builds only. **[Cloud Platform & Enterprise](https://spice.ai/pricing):** Production-ready data distribution available.
+> **Open Source:** Available in nightly builds only. **[Cloud Platform](https://spice.ai/pricing) & [Enterprise](https://docs.spice.ai/docs/enterprise):** Production-ready data distribution available.
 
 **Included Features:**
 
@@ -108,7 +109,7 @@ make install-metal
 
 For Linux systems with NVIDIA GPUs, CUDA distributions enable GPU-accelerated AI/ML inference. Multiple CUDA compute capability versions are available.
 
-> **Open Source:** Available in nightly builds only. **[Cloud Platform & Enterprise](https://spice.ai/pricing):** Production-ready CUDA distribution available.
+> **Open Source:** Available in nightly builds only. **[Cloud Platform](https://spice.ai/pricing) & [Enterprise](https://docs.spice.ai/docs/enterprise):** Production-ready CUDA distribution available.
 
 **Included Features:**
 
@@ -139,7 +140,7 @@ CUDA_COMPUTE_CAP=89 make install-cuda
 
 The NAS (Network Attached Storage) distribution adds support for SMB and NFS data connectors, enabling federated queries against data stored on network file shares.
 
-> **[Enterprise](https://spice.ai/pricing):** The NAS distribution is available in nightly builds and with Spice.ai Enterprise.
+> **[Enterprise](https://docs.spice.ai/docs/enterprise):** The NAS distribution is available in nightly builds and with Spice.ai Enterprise.
 
 **Included Features:**
 
@@ -157,7 +158,7 @@ make install-nas
 
 Different memory allocators can significantly impact performance depending on workload characteristics.
 
-> **Open Source:** Allocator variants are available in nightly builds only for testing and benchmarking. **[Cloud Platform & Enterprise](https://spice.ai/pricing):** Production-ready allocator variants available.
+> **Open Source:** Allocator variants are available in nightly builds only for testing and benchmarking. **[Cloud Platform](https://spice.ai/pricing) & [Enterprise](https://docs.spice.ai/docs/enterprise):** Production-ready allocator variants available.
 
 ### snmalloc (Default)
 
@@ -203,7 +204,7 @@ docker pull ghcr.io/spiceai/spiceai-nightly:latest-sysalloc
 | Windows (WSL)                 | ✅       | Nightly         | Nightly         | ❌     | Nightly         |
 | Windows (Native)              | ❌       | Enterprise only | Enterprise only | ❌     | Enterprise only |
 
-> **Note:** Native Windows support for the Spice runtime is available with the [Spice Cloud Platform and Spice.ai Enterprise](https://spice.ai/pricing). Open source users on Windows should use Windows Subsystem for Linux (WSL).
+> **Note:** Native Windows support for the Spice runtime is available with the [Spice Cloud Platform](https://spice.ai/pricing) and [Spice.ai Enterprise](https://docs.spice.ai/docs/enterprise). Open source users on Windows should use Windows Subsystem for Linux (WSL).
 
 ## Choosing a Distribution
 
@@ -218,7 +219,7 @@ docker pull ghcr.io/spiceai/spiceai-nightly:latest-sysalloc
 
 ## Additional Connectors
 
-Some connectors require additional dependencies and are available with the [Spice Cloud Platform and Spice.ai Enterprise](https://spice.ai/pricing):
+Some connectors require additional dependencies and are available with the [Spice Cloud Platform](https://spice.ai/pricing) and [Spice.ai Enterprise](https://docs.spice.ai/docs/enterprise):
 
 - **ODBC** - Connect to any ODBC-compatible data source
 
@@ -228,6 +229,12 @@ These can be built locally for development and testing:
 make install-odbc
 ```
 
+## Enterprise Authentication
+
+The following authentication features are available with the [Spice Cloud Platform](https://spice.ai/pricing) and [Spice.ai Enterprise](https://docs.spice.ai/docs/enterprise):
+
+- **OIDC Token Verification** - Validate identity tokens from enterprise providers (Okta, Azure AD, Auth0, Google, etc.) for secure access to Spice runtime endpoints.
+
 ### Linux arm64 Notes
 
 - **FP16 (FEAT_FP16)** is required because the `gemm` matrix multiplication library (used by the Candle ML framework) contains half-precision ARM inline assembly that requires the `fullfp16` CPU feature. This is supported on AWS Graviton2+, Ampere Altra, Apple M-series (via Linux VM), and most ARMv8.2-A+ processors.

docs/ROADMAP.md

@@ -15,56 +15,51 @@ To propose features or report issues, please [file an issue](https://github.com/
 
 ## Release Timeline
 
-### v2.0 (March 2026)
+### v2.0 (April 2026)
 
 **Focus:** Production-Grade HA.
 
 - **Spice Cayenne (GA)**: Production-ready columnar storage format for accelerated datasets.
-- **Active-Active HA (GA)**: Production-ready multi-node deployment with zero-downtime failover.
+- **Multi-Active HA (GA)**: Production-ready multi-node deployment with zero-downtime failover.
 - **Distributed Query (GA)**: Stable multi-node query execution for large-scale workloads.
-- **Backup & Point-in-Time Recovery**: Snapshot restore for accelerated datasets.
 - **Accelerated Dataset Distribution**: Replicate accelerated datasets across executor nodes.
 
 ### v2.1 (H2 2026)
 
-**Focus:** Schema Management & Search.
+**Focus:** Schema Management, Search, and Policy.
 
+- **Elasticsearch**: Integration for search and analytics workloads.
 - **Distributed Search (Alpha)**: Federated vector and full-text search across multiple nodes.
 - **Schema Registry (Initial)**: Versioning and backward compatibility checks.
+- **Schema Evolution**: Safe, non-breaking schema changes for accelerated datasets (add/drop/rename columns, type widening) with automatic migration.
 - **Hybrid Search Ranking**: Configurable fusion strategies (Reciprocal Rank Fusion (RRF) weights, learned ranking).
+- **Policy Engine (Cedar-based) Beta**:
+  - Role-based security for fine-grained access control.
+  - Data access policies to enforce compliance and governance.
 
 ### v2.2 (H2 2026)
 
 **Focus:** Reactive Actions & Event Processing.
 
-- **Actions (Drasi-based)**: Reactive event-driven actions triggered by data changes.
 - **Webhooks & Event Notifications**: Push-based data change alerts for downstream consumers.
+- **Actions (Drasi-based)**: Reactive event-driven actions triggered by data changes.
 
 ### v2.3 (H2 2026)
 
-**Focus:** Enterprise Security & Compliance.
+**Focus:** Enterprise Security, Compliance, & Governance.
 
 - **Data Masking & Anonymization**: Dynamic PII redaction at query time.
-- **OIDC Token Verification**: Validate identity tokens from enterprise providers (Okta, Azure AD, etc.).
-
-### v2.4 (2027)
-
-**Focus:** Governance.
-
-- **Policy Engine (Cedar-based)**:
-  - Role-based security for fine-grained access control.
-  - Data access policies to enforce compliance and governance.
 - **Audit Logging**: Persistent, immutable query and access logs for compliance.
 - **Resource Quotas**: Per-user/tenant query limits and throttling.
 
-### v2.5 (2027)
+### v2.4 (2027)
 
 **Focus:** Extensibility & Plugin Architecture.
 
 - **Extensible Middleware**: Pluggable extensions for dynamic customization.
 - **Custom Tool Providers**: User-defined MCP tool implementations.
 
-### v2.6 (2027)
+### v2.5 (2027)
 
 **Focus:** Encryption.
 
@@ -80,9 +75,6 @@ These are prioritized based on community demand and strategic alignment. Share y
 ### Data Connectors
 
 - **Delta Lake**: Write support for transactional data lakes.
-- **Elasticsearch**: Integration for search and analytics workloads.
-- **MotherDuck**: Hybrid cloud DuckDB connector.
-- **BigQuery**: Support for Google Cloud's data warehouse.
 - **Google Docs**: Experimental connector for collaborative data sources.
 
 ### APIs
@@ -94,7 +86,7 @@ These are prioritized based on community demand and strategic alignment. Share y
 ### AI & Analytics
 
 - **Vision Processing**: Support for image and video.
-- **Custom Model Integration**: Framework for user-defined ML models.
+- **Custom ML Model Integration**: Framework for user-defined ML models.
 - **Model Versioning & A/B Testing**: Canary deployments and version management for models.
 - **Hallucination Detection**: Fact-checking LLM responses against source data.
 
@@ -105,13 +97,6 @@ These are prioritized based on community demand and strategic alignment. Share y
 ### Data Platform
 
 - **Data Lineage**: Track data provenance and transformations across the pipeline.
-- **Schema Registry & Evolution**: Versioning and backward compatibility checks.
-
-### Platform Enhancements
-
-- **Observability**: Comprehensive metrics, logging, and tracing.
-- **CLI & SDKs**: Streamlined developer experience with improved tooling.
-- **Data Accelerators**: Additional engines for specialized workloads.
 
 ---
 

docs/criteria/PRINCIPLES.md

@@ -1,23 +1,47 @@
 # Principles
+
 ## Overview
- - Beta internally QA complete; RC externally QA complete.
- - Documentation quality should match code quality standard.
+
+- Beta internally QA complete; RC externally QA complete.
+- Documentation quality should match code quality standard.
+
+## Scope
+
+A "component" refers to any of the following:
+
+- Data Connectors — see [connectors/](./connectors/).
+- Data Accelerators — see [accelerators/](./accelerators/).
+- Catalog Connectors — see [catalogs/](./catalogs/).
+- Models (LLM providers) — see [models/](./models/).
+- Embedding providers — see [embeddings/](./embeddings/).
+
+Runtime subsystems and cross-cutting capabilities that are not a single component (e.g. search, HA, distributed query) use the generic [features/](./features/) criteria.
+
+The generic feature criteria and the component-specific criteria are complementary:
+
+- **Component criteria apply first.** Connector/accelerator/catalog/model/embedding releases are governed by the files in their respective subdirectories.
+- **Feature criteria apply to component releases as well**, wherever the component-specific criteria do not already cover the same requirement (e.g. preview-warning removal, deprecation policy). Where the two conflict, the component criteria win.
 
 ## Per Level
+
 ### Alpha
+
 - Code complete at quality benchmark for most basic, valuable functionality.
 - Limitations & edge case failures expected.
 
 ### Beta
-  - Functionally complete.
-  - Satisfies minimal performance bar. Resource efficiency (memory, CPU, and I/O usage) is measured.
-  - Metrics & tracing in place for observability.
+
+- Functionally complete.
+- Satisfies minimal performance bar. Resource efficiency (memory, CPU, and I/O usage) is measured.
+- Metrics & tracing in place for observability.
 
 ### RC
-  - Ready for production: security & privacy
-  - Performance tested at scale
-  - Feature complete; API stable.
-  - Component ensure correctness.
+
+- Ready for production: security & privacy.
+- Performance tested at scale.
+- Feature complete; API stable.
+- Components ensure correctness.
 
 ### Stable
-  - Verified in production.
+
+- Verified in production.

docs/criteria/accelerators/alpha.md

@@ -6,13 +6,15 @@ All criteria must be met for the Accelerator to be considered Alpha. As Alpha si
 
 ## Alpha Quality Accelerators
 
-| Accelerator | Alpha Quality | DRI Sign-off |
-| ----------- | ------------- | ------------ |
-| Arrow       | ✅            | @Sevenannn   |
-| Cayenne     | ✅            | @lukekim     |
-| DuckDB      | ✅            | @peasee      |
-| SQLite      | ✅            | @sgrebnov    |
-| PostgreSQL  | ✅            | @sgrebnov    |
+| Accelerator             | Alpha Quality | DRI Sign-off |
+| ----------------------- | ------------- | ------------ |
+| Arrow                   | ✅             | @Sevenannn   |
+| Cayenne                 | ✅             | @lukekim     |
+| DuckDB                  | ✅             | @peasee      |
+| SQLite                  | ✅             | @sgrebnov    |
+| PostgreSQL<sup>\*</sup> | ✅             | @sgrebnov    |
+
+<sup>\*</sup> PostgreSQL is a [Spice.ai Enterprise](https://docs.spice.ai/docs/enterprise)-only accelerator.
 
 ## Alpha Release Criteria
 

docs/criteria/accelerators/beta.md

@@ -6,13 +6,15 @@ All criteria must be met for the accelerator to be considered Beta, with excepti
 
 ## Beta Quality Accelerators
 
-| Accelerator | Beta Quality | DRI Sign-off |
-| ----------- | ------------ | ------------ |
-| Arrow       | ✅            | @Sevenannn   |
-| Cayenne     | ✅            | @phillipleblanc |
-| DuckDB      | ✅            | @peasee      |
-| SQLite      | ✅            | @sgrebnov    |
-| PostgreSQL  | ✅            | @sgrebnov    |
+| Accelerator             | Beta Quality | DRI Sign-off    |
+| ----------------------- | ------------ | --------------- |
+| Arrow                   | ✅            | @Sevenannn      |
+| Cayenne                 | ✅            | @phillipleblanc |
+| DuckDB                  | ✅            | @peasee         |
+| SQLite                  | ✅            | @sgrebnov       |
+| PostgreSQL<sup>\*</sup> | ✅            | @sgrebnov       |
+
+<sup>\*</sup> PostgreSQL is a [Spice.ai Enterprise](https://docs.spice.ai/docs/enterprise)-only accelerator.
 
 ## Beta Release Criteria
 
@@ -28,7 +30,7 @@ The Beta release criteria expand on and require that all [Alpha release criteria
 - [ ] The accelerator supports full federation within a single dataset (e.g. `select * from my_dataset`)
 - [ ] The accelerator supports federation push down across multiple datasets within the same accelerator (e.g. `select * from first_dataset, second_dataset`)
 - [ ] The accelerator supports resolving on conflict behaviors (e.g. Drop/Upsert)
-- [ ] Embdedded accelerators support file-mode storage (e.g. SQLite, DuckDB)
+- [ ] Embedded accelerators support file-mode storage (e.g. SQLite, DuckDB)
 - [ ] [Core Arrow Data Types](../definitions.md) are supported
 - [ ] All known [Major Bugs](../definitions.md) are resolved
 - [ ] The accelerator supports reading datasets with the same max column count as the accelerator source.

docs/criteria/accelerators/rc.md

@@ -6,13 +6,15 @@ All criteria must be met for the accelerator to be considered [RC](../definition
 
 ## RC Quality Accelerators
 
-| Accelerator | RC Quality | DRI Sign-off |
-| ----------- | ---------- | ------------ |
-| Arrow       | ✅          | @sgrebnov    |
-| Cayenne     | ✅          | @peasee      |
-| DuckDB      | ✅          | @peasee      |
-| SQLite      | ✅          | @peasee      |
-| PostgreSQL  | ✅          | @peasee      |
+| Accelerator             | RC Quality | DRI Sign-off |
+| ----------------------- | ---------- | ------------ |
+| Arrow                   | ✅          | @sgrebnov    |
+| Cayenne                 | ✅          | @peasee      |
+| DuckDB                  | ✅          | @peasee      |
+| SQLite                  | ✅          | @peasee      |
+| PostgreSQL<sup>\*</sup> | ✅          | @peasee      |
+
+<sup>\*</sup> PostgreSQL is a [Spice.ai Enterprise](https://docs.spice.ai/docs/enterprise)-only accelerator.
 
 ## RC Release Criteria