Skip to main content

​
v1.3.3 — January 8, 2026

​
Smarter embeddings, richer metrics, and smoother observability

​
Smarter Embeddings with LiteLLM and Bedrock

We’ve hardened how Genesis handles vector embeddings so they “just work” even in complex proxy setups and across multiple providers.
  • Embedding calls now bypass LiteLLM Proxy automatically when your embeddings are configured to talk directly to OpenAI or Azure, avoiding confusing “invalid model” errors.
  • Added support for a dedicated litellm_proxy embedding provider, so you can explicitly route embeddings through your proxy when that’s what you want.
  • Bedrock embedding provider names are now normalized correctly, so LiteLLM can route requests without extra manual configuration.
  • Additional unit tests cover provider selection, credential handling, and proxy detection so future changes remain safe.

​
Server-Side Row Model (SSRM) for Query Results

Large query results in the Genesis UI now stay fast and responsive, even at scale.
  • Query results grids (in the Data page and embedded in chat) now use server-side pagination, sorting, and filtering instead of loading everything into the browser.
  • You can filter, sort, group, and pivot directly in the grid, with all heavy lifting pushed down to the database.
  • A new backend SSRM service wraps your original SQL and applies AG Grid operations safely, including for complex queries (CTEs, unions, window functions).
  • Row counts are cached to keep performance snappy on large datasets.

​
Transparent Configuration “Sources” View

Understanding where a configuration value really comes from is now much easier.
  • New Config Sources views let you see exactly which source (profile, env var, file, default, etc.) is driving each config value.
  • A dedicated changes modal summarizes what will be added, updated, deleted, or skipped when applying a profile.
  • CLI and backend APIs expose the same “sources” information, so you can reason about configuration both in UI and automation.
  • This supports safer rollouts and more predictable behavior across environments.

​
More Robust Authentication-Aware UX

The frontend now waits for authentication to be ready before firing key data calls and WebSocket connections.
  • Many hooks (blueprints, dashboards, notifications, docs, bots, user info, etc.) now defer requests until auth config is loaded and the user is authenticated.
  • WebSocket connections to backend services now gracefully wait for tokens rather than failing early during page load.
  • Better logging and readiness checks make it easier to diagnose authentication-related issues.

​
Stronger Snowflake Deployment & Changelog Workflows

We’ve smoothed out release and deployment automation, especially around Snowflake and GitHub-powered changelogs.
  • The Snowflake deploy workflow now fetches shared actions and scripts from main when needed so older tags can still be deployed reliably.
  • After each Snowflake deployment, both Docker images and the git repo are tagged with a Snowflake-specific version (e.g., snowflake-dev-V1_3_3-5) for clear traceability.
  • The promote-to-GA workflow gained a more robust automatic changelog generation, plus:
    • Better error handling for the first GA release vs. incremental releases.
    • A new changelog preview workflow that shows changes between any two refs in the GitHub Actions summary.
  • Workflow permissions now explicitly allow read access to pull requests and wire in GITHUB_TOKEN where needed so changelog jobs can query the GitHub API successfully.
  • Redundant git-tag steps and unnecessary complexity were removed from Snowflake deployment workflows.

​
Improved Metrics Model & Observability

Metrics storage and schema have been upgraded to support richer analysis and safer evolution.
  • Each metric record now includes a globally unique unique_id for more reliable correlation and updates across systems.
  • The attributes column is now native JSON (JSONB where supported), enabling richer structured metric metadata.
  • SQLAlchemy and related adapters were upgraded to 2.0-compatible versions, and unused Redshift-specific dependencies were removed.
  • Redundant indexes on the metrics model were cleaned up, reducing clutter and keeping database schemas tidy.

​
Better Embedding Stability in Proxy/Hybrid Environments

Beyond the big features, there are multiple incremental embedding improvements that reduce production surprises.
  • Embedding calls now explicitly pass provider-specific endpoints and keys when global proxy settings would otherwise interfere.
  • Normalization and mapping between Genesis provider labels and LiteLLM provider names ensure that Bedrock and other providers are recognized correctly.
  • Additional tests validate that mixed setups (proxy for LLMs, direct providers for embeddings) behave predictably.

​
Observability & Logs UX Refinements

The Pulse logs experience is easier to use for real-time troubleshooting.
  • The Pulse Live Logs page now supports horizontal scrolling so long lines (stack traces, JSON, SQL) are fully visible.
  • Log lines use clear color-coding by level (INFO, ERROR, WARNING, telemetry) with theme-aware styling instead of ad-hoc string matching.
  • Layout and theming fixes improve readability in both light and dark modes.

​
Dark Mode & Theming Polish

We continued to refine the new theming system introduced with dark mode.
  • The dashboard, chat, missions views, config pages, and Pulse screens now use consistent theme tokens for backgrounds, borders, text, and states.
  • Dark mode and light mode behavior is more consistent, including inputs, notifications, accordions, tables, and “testing” banners.
  • Many one-off hardcoded colors have been replaced with centralized theme variables for easier future updates.

​
Under the Hood

Under-the-hood improvements that support reliability and maintainability:
  • Snowflake helper scripts and actions were cleaned up (removing stale imports and redundant tags) to eliminate brittle failure points in long-lived workflows.
  • Demo SQLite database connections are now auto-repaired on startup to avoid broken sample connections when paths change.
  • Multiple new and refactored test suites cover SSRM behavior, configuration sources, embedding providers, and deployment helpers.

​
What This Means for You

🔐 Safer, clearer infrastructure Your release, deployment, and metrics pipelines are easier to reason about, with clearer tags, richer metric schemas, and more robust Snowflake workflows. 📊 Faster, more scalable analytics UX Query results and logs remain responsive at larger scales, and you can interact with them using powerful server-side filtering, sorting, grouping, and pivoting. 🧠 More reliable AI behavior Embedding behavior is more predictable across proxies and providers, reducing production surprises in vector search and AI-powered experiences.
Have feedback or want to see a capability expanded in a future release? We’d love to hear from you. — The Genesis Team