Hi all,
Recording: https://fathom.video/share/Q811Jrv5sBiKsUVBeFbz6y_x2gx_qkFy

*Purpose*
Review 2026 roadmap priorities, finalize 0.9 release (largest since
open-sourcing), and align on Iceberg V3 integration, AI positioning, Rust
client evolution & Security.

0.9 Release (Early Feb Target)
Timeline: Code freeze this weekend → RCs next week → final release
pre-Lunar New Year.
*Key Features*
● Complex types: full array/map/nested row support with schema evolution
sync to tiered tables
● Auto-increment columns for global dictionaries and high-performance
distinct counts
● Compacted log format for efficient storage of large vectors (row-based,
no column projection overhead)
● Cluster rebalance automation during scaling
● Change Data Feed (CDF) system tables for decision tracking
● Spark integration: catalog + batch/streaming reads/writes
● Azure Blob Storage support

2026 Roadmap: Streaming Lakehouse Strategy
Iceberg  Integration (7 Pillars)

   - Iceberg V3 Core
   - Catalog Ecosystem
   - Performance
   - Multi-Engine Support
   - In-Place Lakehouse
   - Maintenance Operations
   - Benchmarking



*Fluss for AI Positioning*
Core Strategy: Unified open storage layer for raw/columnar/vector data;
eliminates complex multi-destination pipelines (Kafka + KV + ClickHouse).
*Key Capabilities:*
● Feature/context store with external computation engines (e.g., LanceDB
for vector search)
● Decision tracking via Virtual tables (trace  decisions through changelog)
● Zero-copy schema evolution for model adaptation without pipeline breaks
● Compacted log format optimized for large vector storage
● Engine-agnostic with multi-language SDK support (Rust/Java/Go)

*Rust Client Progress*
Milestone: 80 PRs closed in 2 months (~10 contributors); 0.1.0 release on
track.
Recent Additions:
● Column pruning, partitioned KV tables, admin APIs, log poll batching,
leader availability handling, C++ bindings
● Pending: partitioned scan/append features for 0.1.0
Impact: Strategic backbone for unified, low-maintenance ecosystem.

*Security, PG & Flight*

● PostgreSQL API is the top priority. It unlocks the entire Postgres
ecosystem (e.g., psql, pgAdmin), significantly expanding the user base.
● AeroFlight SQL is lower priority. Its user base is unknown, and it can be
implemented as an independent gateway to avoid core dependency issues.
● TLS support is confirmed for v0.10. Lorenzo will draft a spec, mirroring
Kafka's user experience to leverage existing user familiarity.



*Action Items*
● Mehul: Create Iceberg catalog tickets (REST/Glue/Unity/Hive/JDBC);
● Giannis: Sync with Victor on Lake Keeper REST catalog + Flink support
● Community: Email dev list with Rust client progress summary
● Nicholas: Open discussion  for DirectIO



*Current Objective*
Ship cohesive 0.9 with complex types, Changelog, Spark integration, and
cluster rebalancing, Fixes on Streaming lakehouse

Best Regards,
Mehul Batra

Reply via email to