Genesis v1.19.61 Release Notes
June 4, 2026 Faster chat on long threads, MongoDB and Power BI integrations, Claude Opus 4.8 as the new premiere model, Tailscale egress controls for Snowflake, and broad reliability fixes across connections, compression, and agent tooling.Data Connections and Context Graph
New database and analytics integrations expand what agents can query and how lineage is discovered.- MongoDB is now a first-class connection type. Agents can run collection queries and aggregations with interactive Data Table results, browse collections in the database explorer, and scan MongoDB assets into the Context Graph for lineage.
- Microsoft Power BI is available as a new connection type. Agents can list workspaces and reports, inspect dataset refresh status, trigger refreshes, and update data source bindings or parameters after upstream schema changes. A Context Graph scanner maps Power BI lineage alongside other sources.
- Data-context search now cross-checks each result against the live connection registry, so deleted or unreachable connections are no longer shown as queryable.
- Bare Connections-API connection names resolve correctly in cached query paths and the dashboard.
- dbt Cloud connection discovery surfaces failures clearly instead of failing silently during setup.
Chat, Threads, and Compression
Chat responsiveness and thread continuity improved substantially, especially on long or restarted threads.- Cold-path latency after a server restart dropped sharply on large threads (receive-to-first-LLM roughly fourfold faster in typical cases). Warm turns on long threads also respond much faster before the model starts generating.
- Thread log compaction and automatic snapshotting reduce restart overhead on busy deployments without changing how conversations are stored.
- The thread details panel stays open when switching threads, shows an immediate “Thinking…” indicator on send, and presents a clearer per-turn cost breakdown with additive line items.
- New-thread composer text and attachment drafts persist across navigation, with attachment drafts retained for a limited time so accidental tab changes do not lose work in progress.
- Long-thread compression (V6) gained reactive guards, a health probe, and clearer status in the Pulse UI, with quieter background logging when compression is not needed.
- When a shared thread has an active run, Genesis warns before interrupting it.
- Mid-stream tool-call truncation on Cortex is detected more reliably; agents receive guidance to retry in smaller chunks when output hits the cap.
- Mid-stream connection drops on supported providers are retried automatically; if retries are exhausted, chat shows a recoverable notice instead of a silent failure.
- LLM provider errors appear as a clear in-chat notice rather than empty or confusing responses.
- Cortex thread summaries handle refusals and empty fallbacks gracefully.
- Context budget warnings use clearer wording so operators understand what triggered them.
- The compression history panel labels the “Messages” row correctly when stored message counts grow across compressions.
Genesis Apps, Dashboards, and Managed Sources
App authoring, dashboards, and repository-backed artifacts are easier to operate at scale.- Dash dashboard scaffolds derive KPI cards from the bound dataset’s real columns instead of hardcoded demo metrics.
- Managed Sources recovers from malformed frontmatter in repository files and removes an obsolete guide scan type from the admin UI.
- Genesis Apps that commit through the API use the connection’s stored credentials on the commit path, fixing authentication failures for API-driven git operations.
- State export omits regenerable Genesis App artifacts from raw snapshots and resolves path accessors correctly, avoiding export failures.
Connections, LLM Providers, and Cortex
Connection setup and provider configuration gained auth modes and admin visibility.- The premiere (highest-capability) model is now Claude Opus 4.8, used by default for bots on the premiere tier — with a 300K-token context window and vision support.
- Cortex connections support Snowflake key-pair authentication in the configuration UI and across embedding paths.
- Cortex embedding can be set explicitly to REST or SQL, with fields forwarded consistently through save and test flows.
- The git clone dialog lists every git and code connection type, not only GitHub.
- Git operations inherit the configured CA certificate bundle for corporate TLS inspection environments.
- LiteLLM proxy OAuth header fields are visible in the connection UI for troubleshooting delegated auth.
- File uploads respect scope and client identity so attachments correlate correctly with the active bot and thread.
Jira Integration
Jira agents can perform more of the workflow actions operators expect.- Agents can link issues, upload evidence attachments from scoped Genesis files, and transition issues by status name or transition ID.
- Updating a Jira issue description no longer posts the text as a comment by mistake.
- Jira connector write operations use a scoped service identity consistent with other connectors.
Blueprints and Missions
Blueprint library filtering and operational validation improved.- Selecting multiple tags in the Blueprint library uses OR semantics — blueprints matching any selected tag appear, consistent with Mission Center tag filters.
- A new read-only operational health blueprint runs six phases (preflight, connections, LLM round-trip, prompt-cache telemetry, log sweep, and aggregated report) without mutating data.
- Blueprint missions can be preflight-checked against high-risk policy before execution.
Agents, Flows, and Runtime
Agents and flows can read deployment context safely, and tool configuration is more flexible.- A runtime values tool exposes curated non-secret deployment metadata (such as instance name) to agents and Flow steps without reading raw environment variables.
- MCP-backed tool groups are available when an agent’s tool mode is set to include all tools; legacy tool-add paths were modernized.
- New bots validate RBAC roles before the bot row is persisted, preventing invalid role assignments from being saved.
Snowflake Native App and Private Networking
Snowflake deployments with embedded Tailscale gained operator controls introduced in prior releases.- A guided setup flow in the dashboard walks operators through configuring the embedded Tailscale auth-key secret.
- Operators can manage the Tailscale egress allow list from the dashboard, reset to a safe default, and follow a guided disable flow that unbinds references and surfaces cleanup SQL for consumer-owned objects.
- Tailscale advertise-tags for OAuth-client auth keys are sourced from a single boot configuration read on SPCS.
- Snowflake native app upgrades respect release-channel-enabled package settings.
- Strict CORS mode allows additional headers needed for access-preview and cache expiry in embedded views.
Wiki and Knowledge Management
Wiki writing and maintenance are safer and more transparent for governed deployments.- Flow and agent wiki writes record provenance, and a librarian gate can require review before changes land.
- Operators can choose which model powers wiki extraction from the wiki settings UI.
- The background librarian tolerates a corrupted local wiki database without spamming errors on every tick.
- Wiki databases participate in offline startup restore alongside other local state.
- Models that reject temperature parameters no longer fail wiki or general LLM calls because of a fixed temperature setting.
- When wiki writing is disabled in configuration, agents no longer see the wiki write tool.
Post-Upgrade Guidance and Documentation
- Fresh installations now receive an auto-generated first upgrade summary when release notes are available, instead of an empty summary modal on first boot.
- Documentation can be served locally from the running instance, with offline packaging for air-gapped or restricted networks. In-product doc links default to the bundled internal documentation set, styled consistently with the Genesis UI.
Administration
- User last-login timestamps refresh correctly on activity.
- OAuth token endpoint failures log provider response bodies server-side to speed up identity troubleshooting (no change to end-user sign-in flows).
Upgrade Notes
- MongoDB: Register a MongoDB connection under Config → Connections, run Test Connection, then ask an agent to query a collection or run a Context Graph scan. Confirm interactive Data Table links open from the thread.
- Power BI: Create a Power BI connection with Azure tenant ID, application client ID, and client secret. Verify workspace listing and, if lineage is required, run a Context Graph scan after the connection is healthy.
- Tailscale on SPCS: If embedded Tailscale is enabled, review the egress allow list after upgrade and reconnect the Tailscale external access integration in Snowsight if you change allowed hosts. Use the guided disable flow before removing Tailscale-related secrets.
- Cortex key-pair and embedding: After upgrade, open each Cortex connection and confirm authentication mode and embedding method (REST vs SQL) match your Snowflake deployment. Re-test connection and run a short chat turn.
- Long-thread compression: Open a long thread, confirm compression completes, and check the Pulse UI and compression history panel for expected status labels.
- Validate after upgrade: Send a message on a large thread and confirm faster time-to-first-token; open the thread cost panel and verify the breakdown; run the operational health blueprint on a non-production instance if you rely on it for smoke checks.
— The Genesis Team

