Chat Sessions, Smarter Optimization, and Deployment Controls

You can now keep ongoing conversations with the Hyphen Agent in dedicated chat sessions, give the Optimizer more application context from connected GitHub repositories, and fine tune deployments with more granular container resource controls.

This release also adds support for selecting the latest preview build when triggering preview deployment runs.

What’s new

• Start and continue chat sessions with the Agent
• View session history with titles, message previews, statuses, and attached references
• Close chat sessions when work is complete
• Let the Agent create tasks from a chat and attach created references back to the session
• Use connected GitHub repositories to provide additional application context for Optimizer analysis
• Detect likely language, framework, build system, and workload type for deployed apps
• Trigger preview deployment runs using the latest preview build option

How it works

Chat sessions create a persistent conversation thread that stores message history, references, and created entities. Sessions are scoped to the requesting member, but any tasks created by the Agent during a session are visible to all members of the organization.

When the Optimizer runs, Hyphen inspects connected GitHub repositories for the apps in a deployment and builds a lightweight application profile from repository manifests and directory structure. That profile is used alongside deployment metrics and recent activity to improve recommendation quality.

Advanced deployment scaling now supports provider-aware CPU and memory overrides at the container level. For preview runs, selecting latest preview automatically resolves the newest preview build for the targeted app.

Why this matters

Agent conversations are easier to continue and easier to turn into tracked work. Optimization recommendations become more context aware, and container sizing can be tuned more precisely for performance and cost.

• Keep ongoing Agent work organized in a single session
• Share resulting Agent-created tasks across the organization
• Improve optimization recommendations with repository context
• Tune container resources more precisely
• Deploy the newest preview artifact without manually selecting a build

Things to know

• Chat sessions are scoped to the requesting member within the organization
• Tasks created from chat sessions are visible to all organization members
• Optimizer repository profiling requires a connected GitHub repository for the app
• Repository signals support recommendation quality but do not independently trigger changes
• AWS advanced overrides currently support CPU and memory but not per-region min/max instance counts
• The latest preview build option applies to preview deployment runs and requires a matching preview context