Skip to main content

85 posts tagged with "release"

View All Tags

Release 0.5.2 — RAG Access Control and Gateway Routing

Sri Aradhyula
Sri Aradhyula

Released: 2026-05-29 Chart: oci://ghcr.io/cnoe-io/charts/ai-platform-engineering:0.5.2 Previous release: 0.5.1

Highlights​

0.5.2 completes the OpenFGA relationship-based access model for RAG and Knowledge Bases — knowledge bases can now be shared with teams, every Knowledge sidebar tab is gated on a real OpenFGA capability, and the RAG server and BFF enforce deny-by-default. AgentGateway gains an MCP route bridge so platform MCP endpoints (including the built-in Knowledge Base server) can be rendered straight from Helm, and the default now routes all gateway-managed MCP servers through the gateway. Keycloak adds strict client-secret reconciliation for production installs and a migration-health panel in the admin UI, while a RAG ingestor SSRF fix and pinned TLS dependencies harden the web loader.

Release 0.5.3 — CRD-Free Gateway and Release-Name Defaults

Sri Aradhyula
Sri Aradhyula

Released: 2026-05-29 Chart: oci://ghcr.io/cnoe-io/charts/ai-platform-engineering:0.5.3 Previous release: 0.5.2

Highlights​

0.5.3 makes AgentGateway CRD-free by default. The new global.agentgateway.routingMode defaults to static, so enabling the gateway renders the standalone proxy's own config instead of Gateway API / AgentGateway custom resources — no cluster-scoped CRDs and no controller install required, which keeps helm diff and helm upgrade clean on clusters you do not own. The chart also stops hardcoding ai-platform-engineering-* service URLs: in-cluster defaults are now computed from the Helm release name in the deployment templates, so installs under a custom release name resolve their supervisor, Keycloak, OpenFGA, skill-scanner, and bot URLs correctly without manual overrides. Rounding out the release, every agent gains proper Kubernetes startup/liveness/readiness probe semantics, and a Twisted bump closes a DNS-compression DoS in the RAG web loader.

Release 0.5.4 — Admin UI Polish and Keycloak Realm Fix

Sri Aradhyula
Sri Aradhyula

Released: 2026-05-29 Chart: oci://ghcr.io/cnoe-io/charts/ai-platform-engineering:0.5.4 Previous release: 0.5.3

Highlights​

0.5.4 is a small maintenance release focused on admin UI consistency and a Keycloak fix. Native agent and team <select> controls are replaced with the shared searchable picker components, and the Slack and Webex admin panels are consolidated into a single ConnectorAdminPanel, so connector management looks and behaves the same everywhere. Team cards now surface knowledge-base, agent, and tool counts at a glance, the team Knowledge tab's create links point at the correct /knowledge-bases route, and Keycloak renders the configured realm name instead of assuming caipe. There are no values.yaml changes — this is a drop-in upgrade from 0.5.3.

Release 0.5.1 — Fine-Grained Knowledge Base RBAC

Sri Aradhyula
Sri Aradhyula

Released: 2026-05-27 Chart: oci://ghcr.io/cnoe-io/charts/ai-platform-engineering:0.5.1 Previous release: 0.5.0

Highlights​

0.5.1 makes OpenFGA the source of truth for human access to the RAG knowledge base surfaces — Search, Data Sources, Graph, and MCP Tools — and lets admins share individual knowledge bases with teams through relationship-based grants. Under the hood, team membership is consolidated into a single canonical store so every reader and writer agrees on who belongs to a team, eliminating the drift that caused authorized users to fall back to the Default CAIPE Supervisor instead of their configured agent. The release also adds a searchable team picker across admin panels, a friendlier first-run experience that seeds a working Hello-World agent, and an intentionally noisy emergency flag to bypass UI RBAC while repairing an OpenFGA/Keycloak stack.

Release 0.5.0 — Enterprise RBAC and OpenFGA Authorization

Sri Aradhyula
Sri Aradhyula

Released: 2026-05-26 Chart: oci://ghcr.io/cnoe-io/charts/ai-platform-engineering:0.5.0 Previous release: 0.4.18

Highlights​

0.5.0 is a headline minor release that lands CAIPE's enterprise authorization foundation: a full cutover from legacy role-based access control to relationship-based access control (ReBAC) backed by OpenFGA, with Keycloak as the identity provider and an AgentGateway ext_authz bridge enforcing per-call MCP decisions. The release introduces a new Webex bot alongside ReBAC-aware Slack routing, an envelope-encrypted credential store and OAuth connector platform, and a hardened secrets bootstrap that fails loudly on dev placeholders instead of silently shipping insecure defaults. The entire authorization stack is opt-in and disabled by default — for a stock 0.4.18 deployment this is a drop-in upgrade with no values.yaml edits required. Operators who adopt RBAC get team-based access control, per-document RAG ACLs, channel-derived team binding for bots, and a redesigned admin UI for managing teams, identity-group sync, and onboarding defaults.

Release 0.4.17 — Scrapy 2.16 Crawl Fix

Sri Aradhyula
Sri Aradhyula

Released: 2026-05-22 Chart: oci://ghcr.io/cnoe-io/charts/ai-platform-engineering:0.4.17 Previous release: 0.4.16

Highlights​

0.4.17 is a targeted fix for RAG web ingestion. Scrapy 2.16.0 added validation that raised AttributeError whenever a spider defined a singular start_url attribute without a populated start_urls, which broke every crawl mode — single, recursive, and sitemap — resulting in zero pages crawled for any web ingestion request. This release renames the conflicting attribute, adds the new async spider entrypoint, and pins a stable Twisted to resolve a Scrapy 2.16 TLS bug.

Release 0.4.18 — Dynamic Agent Chat Regression Fix

Sri Aradhyula
Sri Aradhyula

Released: 2026-05-22 Chart: oci://ghcr.io/cnoe-io/charts/ai-platform-engineering:0.4.18 Previous release: 0.4.17

Highlights​

0.4.18 fixes a regression introduced in 0.4.17 where dynamic agent configs stored with null fields in MongoDB raised a pydantic ValidationError on every chat request, effectively breaking chat for affected deployments. A new _strip_nulls() helper now removes explicit null values before pydantic construction so default_factory defaults apply correctly. This release also fixes Jira internal-comment creation, which was failing with a JSM 404 authentication error.

Release 0.4.16 — RAG Truncation and Seed Persistence Fixes

Sri Aradhyula
Sri Aradhyula

Released: 2026-05-21 Chart: oci://ghcr.io/cnoe-io/charts/ai-platform-engineering:0.4.16 Previous release: 0.4.15

Highlights​

0.4.16 is a maintenance release focused on correctness. The RAG_MAX_OUTPUT_CHARS cap now actually fires for MCP tool results (previously a type check silently skipped truncation), seed agent configs once again persist their ui, features, and interrupt_on fields to MongoDB, and two reopened CodeQL code-scanning alerts are closed. The bundled CAIPE API reference docs were also refreshed to match the current BFF and RAG surfaces.

Release 0.4.14 — Slack and Chat Reliability

Sri Aradhyula
Sri Aradhyula

Released: 2026-05-20 Chart: oci://ghcr.io/cnoe-io/charts/ai-platform-engineering:0.4.14 Previous release: 0.4.13

Highlights​

0.4.14 is a maintenance release focused on Slack and chat reliability. It fixes several cases where the Slack bot silently failed to respond — overthink skip logic that never fired on auto-routed channel messages, and alert integrations that put all their content inside attachment blocks. The /invoke chat endpoint gains optional conversation-history persistence, the skill scanner now returns actionable validation errors for malformed skills, and Jira agents can post internal (non-customer-visible) service desk comments.

Release 0.4.15 — Curl Private Endpoints

Sri Aradhyula
Sri Aradhyula

Released: 2026-05-20 Chart: oci://ghcr.io/cnoe-io/charts/ai-platform-engineering:0.4.15 Previous release: 0.4.14

Highlights​

0.4.15 is a small follow-up release. The curl builtin tool gains an opt-in allow_non_public_urls flag so dynamic agents can reach private or internal endpoints when explicitly configured, while domain ACLs still apply. The release also quiets noisy Jira MCP logs by default and closes several CodeQL findings by removing sensitive data from logs and stdout.