Public docs

Robotron LLM APIs

Use Robotron as a public-facing task gateway for chat, session orchestration, and long-running AI jobs. As part of task execution, Robotron can also coordinate specialized model workflows, validation steps, and other agent-driven pipelines.

Back to overview Docs Automation kit

Why Robotron

One public surface for language models and execution-heavy AI tasks.

Robotron does more than pass prompts through to a model. It can queue work, persist session history, and hand off specialized jobs to background agents that validate, route, summarize, or enrich business records.

Task execution

Built for more than just chat.

Robotron can run workflows where an LLM initiates or structures work, while agents perform downstream execution such as processing files, checking policy rules, or coordinating multi-step automations.

  • LLM chat
  • TLM workflows
  • Policy validation
  • Document processing
  • Background agents
  • Streaming updates

Authentication

Simple app-key access

  • Send x-app-key on public API requests.
  • Use one key for client calls and agent coordination.
  • Keep session creation and retrieval scoped to your app.

Operational model

Designed for asynchronous jobs

  • Create a session or task from your UI.
  • Let Robotron persist state and message history.
  • Allow connected agents to pick up long-running work.
  • Poll or stream status until the result is ready.

Output formats

Text in, multimodal results out

A single Robotron workflow can start from text prompts and finish with generated copy, summaries, classifications, recommendations, or structured outputs depending on the task type.

Base schema

OpenAPI document

Use the public OpenAPI schema as the canonical starting point for LLM session integration.

GET /api/public/openapi.json

Returns an OpenAPI 3.1 document describing the public chat session endpoints.

Endpoint 1

Create a chat session

Start a session that an agent can later process, whether the job is simple LLM chat or a longer execution pipeline.

curl -X POST http://localhost:3000/api/public/chat/sessions \ -H 'Content-Type: application/json' \ -H 'x-app-key: demo-app-key' \ -d '{ "model": "gemma3:4b", "systemPrompt": "You are Robotron.", "metadata": {"channel":"public-docs"} }'

Endpoint 2

Append a user message

Add the next user turn to an existing session and let Robotron continue orchestration.

curl -X POST http://localhost:3000/api/public/chat/sessions/<sessionId>/messages \ -H 'Content-Type: application/json' \ -H 'x-app-key: demo-app-key' \ -d '{ "content": "Review this policy exception and summarize the approval risks." }'

Endpoint 3

Inspect a session

Retrieve session state, metadata, and message history so your client can render the latest status.

curl -H 'x-app-key: demo-app-key' \ http://localhost:3000/api/public/chat/sessions/<sessionId>

Execution notes

How LLM APIs connect to heavier task execution.

  • Use public chat endpoints when you want a conversational entry point.
  • Use Robotron tasks when the request should trigger a background job.
  • Attach agents to process compute-intensive work outside the request-response cycle.
  • Return generated artifacts such as summaries, decisions, structured records, or reports.