Skip to main content

​
v1.3.2 — December 22, 2025

​
A more observable, governed, and secure Genesis platform

​
Digital Twin & Lineage: Richer end‑to‑end visibility

We started building some of the foundational components of Genesis’ Digital Twin.
  • Introduced multi‑level lineage visualization (System → Component) with drill‑down views for complex pipelines (#2197).
  • Added a Tableau workbook scanner to extract lineage directly from .twb/.twbx dashboards and feed it into the Digital Twin graph (#2197).
  • Enabled multi‑hop pipeline chain detection and chart generation to highlight critical paths and dependencies (#2197).
  • Added a human corrections layer so analysts can adjust lineage and preserve edits across rescans (#2197).

​
Platform Observability: Metrics Foundational Components

We turned metrics into a first‑class part of the platform so you can monitor health, usage, and performance over time. Still early days, as we experiment with the right approach here.
  • Introduced a unified MetricsService that reports slot acquisition and waiting metrics by thread purpose (user chat, blueprint, etc.), making it easier to tune concurrency and capacity (#2188).
  • Added a SQL‑based metrics backend using OpenTelemetry and SQLAlchemy, so metrics can be stored long‑term and queried with standard tools (#2193).
  • Extended the metrics system with instance‑level attributes and simplified reporting APIs, plus a renamed genesis_metrics table for cleaner analytics (#2194).
  • Documented the full Genesis Metrics stack, including configuration, schemas, and example queries for dashboards in tools like Grafana or Metabase (#2216).

​
Stronger RBAC & Secret Governance

We significantly tightened access control across chat, missions, and secrets so security and compliance teams have clearer guarantees.
  • Enforced agent‑level RBAC on all chat paths (HTTP and WebSocket), plus blueprint mission creation, eliminating a high‑severity gap that allowed sending messages to unauthorized agents (#2201).
  • Completed RBAC rollout v3: removed fragile “demo mode” guards in favor of permission‑based controls, centralized thread visibility logic, and added an admin‑configurable default thread visibility (public vs private) for new conversations (#2173).
  • Enabled bots to access user‑created secrets via RBAC: admins can now grant a role access to specific secrets, assign roles to bots, and manage bot/secret mappings from the Users & Roles UI (#2171).
  • Updated APIs and front‑end wiring so secret and credential endpoints consistently enforce MANAGE_SECRETS / MANAGE_CONNECTIONS permissions (#2173, #2171).

​
Provider Connectivity & Model Configuration Improvements

Connecting Genesis to your LLM providers is now more predictable, with clearer configuration flows and better diagnostics.
  • Unified connection‑test logic for all providers behind a dedicated ConnectionTestService, with stateless provider‑specific testers and updated integration tests (#2221).
  • Simplified Gemini configuration to use only an API key, removed deprecated Vertex AI project/location fields, and updated tests and docs accordingly (#2230).
  • Improved Bedrock integration by enforcing AWS account ID and stricter token validation, adding new supported regions, and tightening activation flows (#2215).
  • Refined LiteLLM‑based provider tests and fixed issues with model name normalization and Bedrock ARN construction to reduce false negatives in connection checks (#2224, #2226).
  • Fixed Azure OpenAI integration to use deployment names correctly and clarified the configuration vs activation experience in the UI (#2163).

​
Mission & Blueprint Reliability

Blueprint execution and mission replay are now more robust, especially for complex multi‑phase flows.
  • Fixed multiple phase‑ID and awaiting‑input issues so blueprints behave consistently for both numeric and non‑numeric phase IDs (fork phases, branch phases, etc.) (#2164, #2159).
  • Introduced a dedicated phase status service and a PhaseVisitHistory helper to centralize status tracking and visited‑phase logic, improving loop detection and mission‑flow introspection (#2174, #2177).
  • Ensured pre‑phase artifacts are attributed to the correct first entered phase, eliminating phantom “Phase 0” nodes in mission flow graphs (#2164).
  • Released blueprint agent slots immediately upon blueprint completion by adjusting cooldown behavior, increasing capacity for back‑to‑back runs (#2137).
  • Added the new Changelog blueprint itself, which automates PR‑based changelog generation and PR creation for future releases — You’re reading the results of it now!

​
Thread Search, Titles, and History

Finding and organizing conversations in Genesis is now much easier.
  • Added new agent tools find_threads and read_thread, enabling Eve and other agents to search prior conversations (hybrid semantic + lexical) and read thread history while preserving tool output formatting (#2160).
  • Indexed thread titles in the search backend so users can find conversations by name, not just content (#2148).
  • Switched thread rename to a direct storage update (no AI involvement), making title changes instant and avoiding extra messages in the thread (#2148).
  • Ensured title updates propagate consistently between sidebar and chat header via WebSocket events, keeping all open views in sync (#2148).
  • Fixed relative time display and sorting in the thread list for legacy threads by adding a robust timestamp fallback chain and using server time for calculations (#2200).

​
Security & Access Controls

Beyond RBAC, we shipped several smaller but important security and hardening updates.
  • Removed a deprecated Anthropic prompt-caching header that some enterprise endpoints rejected, restoring compatibility while keeping prompt caching functional (#2217).
  • Ensured agent deletion flows use the correct identifiers and guardrails, preventing partial or inconsistent deletes from the UI (#2147).
  • Tightened pre‑commit hooks to block new tests from being added to legacy test directories, keeping the test tree maintainable and predictable (#2149, #2144).

​
UI & UX Polishing

We made a set of UX improvements that you and your teams will notice day‑to‑day.
  • Refined the New Connection experience with a new StyledConnectLink component and polished hover/active states for a more consistent look and feel (#2237).
  • Fixed the artifacts tree in Mission Results so it fills its panel, shows scrollbars correctly, and avoids content being cut off, especially with resizable panels (#2151).
  • Improved small‑sidebar and mission layout behavior to better reflect scrollable areas and reduce visual ambiguity (#2151).
  • Ensured AG Grid licensing keys are correctly wired and environment‑driven, preventing unpleasant surprises in grid behavior (#2195).
— The Genesis Team