Orchestration Layer
Orchestration is not a standalone model. It analyzes the incoming request, routes it to the right model or agent path, and combines partial results into a single response.
01How It Works
The decision flow first identifies the request type and difficulty, then selects the required models or agents. In the final stage, outputs pass through a consistency check and are returned as a single answer.
User Request
Intent, context, tool requirements, and output format are collected in one input.
Decision Engine
Unified Response
Selected step outputs are validated, simplified, and returned as one answer.
02Routing Examples
The same endpoint can run different requests with different strategies. Simple work goes to a fast model; multi-step work goes to agents and pipeline structure.
03Why Orchestration?
Automatic Routing
Request complexity, modality, and tool requirements are read; the task is transferred to the most suitable model or agent path.
Cost Control
Simple work is not sent to heavy models. Each step runs with the most efficient capacity that is sufficient.
Parallel Execution
Independent subtasks run at the same time, and their results are combined into one response.
Single API
The integration side keeps one call; background selection and distribution are handled by the orchestration layer.
04API Usage
curl https://api.briqmind.com/v1/chat/completions \
-H "Authorization: Bearer $BRIQ_API_KEY" \
-d '{
"workflow": "orchestration",
"messages": [
{
"role": "user",
"content": "Scan the codebase and produce a security report."
}
]
}'orchestration is selected, model choice and task distribution are handled automatically.