Chat Agents API - SambaNova Documentation

The SambaNova Agent provides access to specialized AI agents for different tasks. Each agent has both a fire-and-forget endpoint (single call) and an interactive endpoint (multi-turn conversation).

Available agents

Agent	Endpoint	Capabilities
Main Agent	`api/agent/mainagent`	General-purpose assistant with access to all subagents
Financial Analysis	`api/agent/financialanalysis`	Stock analysis, financial metrics, market data
Deep Research	`api/agent/deepresearch`	Multi-iteration research with comprehensive reports
Data Science	`api/agent/datascience`	Dataset analysis, visualizations, ML workflows
Coding Agent	`api/agent/coding`	Code execution, debugging, sandboxed Python

Key features

Artifacts System: Files (charts, PDFs, scripts) returned as downloadable IDs
Multi-turn Conversations: Interactive endpoints with thread_id continuity
Auto-approval: Deep research and data science interrupts handled automatically
File Upload: Data science accepts dataset uploads
Bearer Auth: Secure API key authentication

Common patterns

Fire-and-forget pattern

Single API call that processes the request and returns the complete result. Use case: Quick queries, one-shot tasks, automated workflows

curl -X POST "https://<base-url>/agent/mainagent" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Create a bar chart for Q1 sales"
  }'

{
  "status": "success",
  "result": "I've created the bar chart...",
  "artifacts": ["file-id-123"],
  "thread_id": "thread-abc"
}

Interactive pattern

Multi-turn conversation with persistent context using thread_id. Use case: Follow-up questions, iterative development, complex workflows

Step 1: Initial request

curl -X POST "https://<base-url>/agent/mainagent/interactive" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Create a bar chart for Q1 sales",
    "resume": false
  }'

{
  "status": "success",
  "result": "I've created the bar chart...",
  "artifacts": ["file-id-123"],
  "thread_id": "thread-abc"
}

Step 2: Follow-up request

curl -X POST "https://<base-url>/agent/mainagent/interactive" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Now create a pie chart showing the distribution",
    "thread_id": "thread-abc",
    "resume": true
  }'

{
  "status": "success",
  "result": "I've created the pie chart...",
  "artifacts": ["file-id-456"],
  "thread_id": "thread-abc"
}

​Available agents

​Key features

​Common patterns

​Fire-and-forget pattern

​Interactive pattern

​Step 1: Initial request

​Step 2: Follow-up request