Chat & AI¶

The chat interface is Synaptiq's primary interaction surface — where natural language becomes rich, interactive business applications.

Sending Messages¶

Type your query in the message input at the bottom of the screen. Synaptiq supports:

Query Type	Example	Response
Data queries	"Show me Q1 revenue by region"	KPI cards + bar chart + data table
Dashboard requests	"Create a sales dashboard"	Composite view with multiple components
Workflow triggers	"Generate therapy goals for this client"	Multi-agent workflow execution
Knowledge questions	"What's our refund policy?"	RAG-powered answer with source citations
Admin commands	"Update the AI persona to be more formal"	Configuration changes

Streaming Responses¶

Synaptiq uses Server-Sent Events (SSE) for real-time response streaming:

sequenceDiagram
    participant User
    participant Frontend as Angular Shell
    participant Backend as Spring Boot
    participant LLM as Gemini / OpenAI

    User->>Frontend: Sends message
    Frontend->>Backend: POST /api/v1/chat/message<br/>(Accept: text/event-stream)
    Backend->>LLM: System prompt + schema + query

    loop Streaming tokens
        LLM-->>Backend: Token chunk
        Backend-->>Frontend: SSE event
        Frontend-->>User: Render incrementally
    end

    Backend-->>Frontend: Component DSL spec
    Frontend->>Frontend: Parse & render components
    Frontend-->>User: Interactive dashboard

Responses appear in real-time as the LLM generates them. Component DSL specs are rendered into rich UI components after the full spec is received.

Component DSL Rendering¶

When the LLM determines that a rich UI component is appropriate, it emits a Component DSL specification. The frontend renders these into native Angular Material 3 components:

Data Visualization¶

Component	When Used
KPI Card	Single metrics (revenue, count, percentage) with trend indicators
Bar Chart	Comparing values across categories
Line Chart	Trends over time
Pie / Donut Chart	Proportional breakdowns (powered by ECharts)
Stat Grid	Multiple metrics in a compact grid

Catalog & Lists¶

Component	When Used
Item Card	Product/entity display with image, title, metadata
Item Grid	Grid layout of item cards
Data Table	Sortable, filterable tabular data
Comparison Table	Side-by-side comparison of entities

Interactive¶

Component	When Used
Dynamic Form	Data entry with validation and conditional fields
Kanban Board	Status-based workflow visualization
Timeline	Chronological event display
Progress Tracker	Step-by-step completion status

Session Management¶

Each conversation is tracked as a session:

Create session — automatically created on first message, or manually via the UI
Session history — view past conversations in the left sidebar
Context continuity — follow-up questions reference previous context within the session
Multi-session — run multiple conversations with independent context

API Endpoints¶

Method	Endpoint	Description
`POST`	`/api/v1/chat/sessions`	Create a new chat session
`GET`	`/api/v1/chat/sessions`	List sessions for the current tenant
`POST`	`/api/v1/chat/message`	Send a message (SSE streaming response)
`GET`	`/api/v1/chat/sessions/{id}/history`	Get conversation history

RAG Context¶

When a knowledge base is configured, Synaptiq automatically enriches chat responses with relevant document context:

User's message is embedded using the configured embedding model
Similar document chunks are retrieved from MongoDB Atlas Vector Search
Retrieved context is injected into the system prompt
The LLM generates responses grounded in your organization's documents
Source citations are included in the response

Improving RAG Quality

Upload domain-specific documents (policies, procedures, product catalogs)
Use clear, well-structured documents for better chunking
Review and curate the knowledge base regularly