Conversation tools

​

Contract status

Derived implementation contract These features are documented as product and UX specs rather than fixed route contracts, but they define clear service responsibilities for the conversation layer.

​

Source documents

• cognition console UI design
• collaborative personas planning
• computer use for personas
• chat cleanup system
• persona skill slots
• adaptive UI tutorials
• in-chat navigation
• context windows in AI
• fluid UI architecture
• conversation split and route
• forget this feature
• persona meeting mode

​

Core interaction services

​

Cognition console

Required capabilities:

• Display multi-layer reasoning or context traces
• Surface persona state and tool activity
• Support safe operator oversight

​

Collaborative personas

Required capabilities:

• Run multi-persona collaboration on the same thread or task
• Preserve authorship and persona boundaries
• Support explicit invocation and shared outcome capture

​

Computer use

Required capabilities:

• Represent browser or computer-use actions as governed tool calls
• Log actions for audit and replay
• Route approvals when actions exceed safe automation limits

​

Chat cleanup

Required capabilities:

• Collapse or archive stale context
• Preserve important references and pinned material
• Improve future retrieval without losing meaningful history

​

Skill slots

Required capabilities:

• Attach skill bundles to personas
• Expose available skills by role or context
• Control enablement per persona and instance

​

Adaptive tutorials

Required capabilities:

• Detect user context and show relevant learning flows
• Track progress and dismiss state

​

In-chat navigation

Required capabilities:

• Jump to prior topics, references, and workspace objects from the chat surface
• Preserve conversation map metadata

​

Context windows

Required capabilities:

• Rank and inject only the most relevant memory and workspace context
• Respect token budgets and persona priorities

​

Split and route

Required capabilities:

• Detect when a conversation should branch into separate threads or tasks
• Route the branch to the right persona, module, or workspace destination

​

Forget

Required capabilities:

• Remove or suppress selected memories or references in a governed way
• Keep audit and policy metadata around deletion or redaction decisions

​

Meeting mode

Required capabilities:

• Support multi-party, persona-aware operating mode
• Persist meeting artifacts, decisions, and action items

​

Shared design requirements

• Everything must be auditable.
• Persona authorship must remain visible.
• High-risk actions need explicit governance.
• Cleanup and forgetting must never silently destroy critical records.
• Context injection should be deliberate and ranked, not indiscriminate.

​

Implementation note

These features form the OS interaction layer. Treat them as orchestration services on top of personas, memory, and workspace objects rather than as isolated one-off UI features.