This area will become the controlled collaboration gateway between Eden and the SVTV
Platform. Eden can now report real backend provider state while remaining read-only.
Diagnostics refresh automatically every 15 seconds and read the server health endpoint. No
API keys, tool execution, writes, or persistent memory are exposed here.
Eden Diagnostics
Planned
Reads the live SVTV server state from /api/eden/health. This panel does not use
hardcoded provider values and does not silently fall back to Local when the health request
fails.
Provider
Not checked
Model
Not checked
API Key
Not Loaded
Env Source
missing
Health
Not checked
Fallback
None
Last health check
Not checked
Health not checked yet.
Live Chat
Draft
Shows the current gated provider selection for /api/eden/chat. This card does not
call OpenAI and does not affect the conversation transcript.
Provider
Not checked
Model
Not checked
Gate
Disabled
Current Chat Provider
Local
Fallback Active
Yes
Fallback Reason
Not checked
API Key
Not Loaded
Last Status Check
Not checked
Isolated Live Eden Smoke Test
Draft
Sends one tiny manual test message through /api/eden/smoke-test. This does not run
automatically, does not affect Hub rendering, and does not wire live Eden into the main panel.
Not run yet.
Provider Information
Planned
Current Provider
Not checked
Configured Provider
unset
Current Model
Not checked
Provider Status
Not checked
Connection Status
Not checked
API Endpoint
/api/eden/chat
API Key Loaded
Not Loaded
Env Source
missing
Read Only
Yes
Memory
Not Connected
Tools
Disabled
Streaming
Planned
Last Health Check
Not checked
Environment
Local
Fallback Reason
None
Server Environment
Planned
EDEN_PROVIDER
Not checked
EDEN_MODEL
Not checked
OPENAI_API_KEY_LOADED
Not checked
EDEN_LIVE_CHAT_ENABLED
Not checked
Env Source
Not checked
Server Health
Planned
Endpoint Reachable
Not checked
Provider Loaded
Not checked
Configuration Loaded
Not checked
Connection Ready
Not checked
Future Memory
Planned
Future Tools
Planned
Request Pipeline
Planned
BrowserReady
EdenClientReady
API EndpointReady
ProviderPlanned
ResponseReady
Connection Log
Active
Local metadata-only events. Request content is never logged here.
No connection events yet.
Provider Switcher
Draft
Runtime switching is not enabled. Provider selection reflects the real server environment.
Local PlaceholderOpenAI
Connection Hardening
Active
MCP exposes provider diagnostics, health checks, connection logs, and request-pipeline visibility before live Eden receives tool access.
Diagnostics read /api/eden/health instead of hardcoded placeholder values
Health checks use real backend state and ignore browser Safe Mode preference
Connection Lost is shown when the health request fails
Provider details show endpoint, mode, memory, tools, streaming, and environment
Server Health reports configuration readiness without exposing secrets
Read-Only Live Eden
Active
Live Eden chat can route through the server provider while remaining read-only and approval-aware.
Browser uses EdenClient only
/api/eden/chat selects the provider server-side behind EDEN_LIVE_CHAT_ENABLED=true
OpenAI activates only when EDEN_PROVIDER, OPENAI_API_KEY, EDEN_MODEL, and the live-chat gate are configured
LocalProvider remains the permanent safety net for chat when provider configuration is unavailable or OpenAI fails
No tool execution, writes, deployments, database changes, or memory persistence are allowed
Provider Safety Boundary
Active
The browser never receives provider keys. All live Eden calls go through the SVTV server route.
Provider selection is server-side only
OPENAI_API_KEY stays in private environment variables
Eden receives StudioState as read-only approved context
Attachment handling is metadata-only
Tools, actions, memory, writes, and MCP execution remain disabled
StudioState / Studio Brain
Draft
StudioState will become the unified context snapshot that the platform prepares before Eden responds.
current route, area, page, application, project, and selected item
current conversation, current revision, and conversation status
recent local Studio Content edits, knowledge, assets, tasks, deployment state, and permissions
future live Eden should receive StudioState or an approved subset instead of querying platform areas directly
EdenSession
Draft
EdenSession is the future request envelope carrying conversation, permissions, attachments, revision, and StudioState context into live Eden.
sessionId and conversationId
route, page, project, application, and selectedItem
memoryMode, permissions, conversationStatus, and revisionId
attachments and StudioState context snapshot
no execution, API calls, or real MCP actions yet
Tool Registry
Draft
Future Eden/MCP tools are registered as names, descriptions, required permissions, and readiness statuses before any execution exists.
Read Current Page
Read Studio Context
Read Knowledge
Read Assets
Search Studio
Draft Prompt
Generate Documentation
Review Architecture
Permission Layer
Draft
Future Eden actions will pass through an approval model before reads, writes, commands, or MCP tools are allowed.
Read Page
Read Knowledge
Read Assets
Read Conversations
Generate Text
Modify Documentation
Modify Assets
Run Commands
Execute MCP Tools
Memory Model Draft
Draft
Future Eden memory should live inside the SVTV Platform as controlled, searchable, project-aware context.
Conversations should be archived inside SVTV Platform
Buddy controls what is stored and what Eden can access
Memory should support summaries, tags, timestamps, source links, and project/context links
Future Eden should prefer approved platform memory over vague hidden memory
Safety Rules
Draft
No silent destructive changes. No public-facing changes without approval. No permanent writes without clear confirmation. Prefer drafts and previews before applying changes. Maintain history where possible.