chatCompletion

Sends a query to a configured LLM and returns the generated text. Optionally assembles a structured prompt from context data — extracted facts, reference citations, source documents, and metadata. Actual API cost is tracked per execution.

Parameters

Param	Type	Required	Default	Description
query	string	Yes	—	User query or question. Minimum length: 1
system_prompt	string	No	"You are a helpful AI assistant that provides accurate, well-reasoned responses."	System prompt to guide LLM behavior
context_data	object?	No	null	Optional context to inject into the prompt (see below)
model	string	No	"gpt-4o-mini"	LLM model identifier
provider	string?	No	null	LLM provider. Auto-detected from model when not set
temperature	float	No	0.1	Range: 0.0–2.0
max_tokens	integer?	No	null	Maximum tokens in response
response_format	string	No	"text"	One of: text, markdown, json
include_citations	boolean	No	true	Count citation references that appear in the response

context_data object:

Field	Type	Default	Description
facts	object[]	[]	Extracted facts with optional fact, supporting_citations, and confidence_score keys
citations	object[]	[]	Reference citations with optional id, title, and doi keys
documents	string[]	[]	Source document texts
metadata	object	{}	Arbitrary key-value pairs appended to the prompt

Output

Field	Type	Description
response	string	Generated response text from the LLM
model_used	string	The model name passed to the LLM
provider_used	string	Provider resolved from the model, or "error" on failure
token_count	integer	Estimated token count (word_count × 1.3)
context_used	boolean	true if context_data was provided
citations_included	integer	Count of [cite-id] references found in the response
execution_time_seconds	float	Wall-clock time in seconds

Example

{
  "id": "llmCompletion",
  "type": "chatCompletion",
  "data": {
    "label": "Chat Completion",
    "isExecuted": false,
    "handles": ["inputs", "outputs"],
    "schema": {},
    "params": {
      "query": { "value": "{{ $input.question }}", "isExpression": true, "isAttachedToInputNode": false },
      "system_prompt": { "value": "You are a scientific assistant. Answer using only the provided context.", "isExpression": false, "isAttachedToInputNode": false },
      "model": { "value": "gpt-4o-mini", "isExpression": false, "isAttachedToInputNode": false },
      "context_data": {
        "value": {
          "documents": "{{ @vectorSearch.documents }}",
          "citations": [],
          "facts": [],
          "metadata": {}
        },
        "isExpression": true,
        "isAttachedToInputNode": false
      }
    },
    "inputs": [], "outputs": [], "errors": []
  },
  "position": { "x": 600, "y": 0 },
  "isSelected": false,
  "isDragging": false
}