Prompt file imported from synccinct/customagents-foundation (.github/prompts/wiring/07-chat-execution-wiring.prompt.md). Copyright stays with the author.

Prompt 7 — Chat Execution Wiring

Wire the agent chat UI to a backend execution endpoint.

Goal

Customers can interact with their deployed agents through the dashboard.

Implement

• conversation model
• message model
• POST message endpoint
• GET conversations endpoint
• GET conversation messages endpoint
• UI conversation state
• message persistence
• execution adapter placeholder
• structured execution response
• error display

Execution Architecture

UI ChatWindow
-> API /agents/:agentId/chat
-> Auth/customer scope check
-> Agent lookup
-> ExecutionService
-> Provider/model adapter placeholder
-> Message persistence
-> Response returned to UI

Rules

• Do not expose provider API keys.
• Do not expose system prompts.
• Validate message length.
• Rate-limit hook or placeholder.
• Every request should have requestId.
• Log agentId, customerId, conversationId, latency, status.
• Store user and assistant messages.

Acceptance Criteria

• User can send a chat message from UI.
• API validates and stores message.
• Assistant placeholder response is stored and returned.
• UI renders message history.
• Errors are shown safely.
• Execution service can later plug into AIXpertise or provider layer.