chatAI4R (version 1.3.1)

multiLLMviaionet: multiLLMviaionet: Execute ALL available LLM models simultaneously via io.net API

Description

This function automatically runs the same prompt across ALL currently available LLM models on the io.net API (by default it now selects one random model to reduce API load). It supports running multiple models (configurable) in parallel with streaming responses and comprehensive error handling.

The function dynamically fetches the current list of available models from the io.net API and executes a configurable subset (default 1 random model; set max_models or random_selection to control). Results are cached for 1 hour to improve performance. If the API is unavailable, it falls back to a static model list.

Typical models observed on io.net (30-Nov-2025) include: - moonshotai/Kimi-K2-Thinking, moonshotai/Kimi-K2-Instruct-0905 - zai-org/GLM-4.6 - meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8, Llama-3.3-70B-Instruct, Llama-3.2-90B-Vision-Instruct - Qwen/Qwen3-235B-A22B-Thinking-2507, Qwen3-Next-80B-A3B-Instruct, Qwen2.5-VL-32B-Instruct, Intel/Qwen3-Coder-480B-A35B-Instruct-int4-mixed-ar - deepseek-ai/DeepSeek-R1-0528 - mistralai/Devstral-Small-2505, Magistral-Small-2506, Mistral-Large-Instruct-2411, Mistral-Nemo-Instruct-2407 - openai/gpt-oss-120b, openai/gpt-oss-20b

Usage

multiLLMviaionet(
  prompt,
  max_models = 1,
  streaming = FALSE,
  random_selection = TRUE,
  api_key = Sys.getenv("IONET_API_KEY"),
  max_tokens = 1024,
  temperature = 0.7,
  timeout = 300,
  parallel = TRUE,
  verbose = TRUE,
  refresh_models = FALSE,
  retries = 1,
  retry_wait = 2,
  monitor_timeout = 120
)

Value

A list containing:

results

List of responses from each model with metadata

summary

Summary statistics including execution times and token usage

errors

Any errors encountered during execution

models_used

Character vector of models that were actually executed

Arguments

prompt

A string containing the input prompt to send to all selected models.

max_models

Integer specifying maximum number of models to run (1-50). Default is 1 (single model).

streaming

Logical indicating whether to use streaming responses. Default is FALSE.

random_selection

Logical indicating whether to randomly select models from all available models. If TRUE (default), randomly selects up to max_models from all available models. If FALSE, uses models in order (up to max_models limit).

api_key

A string containing the io.net API key. Defaults to the environment variable "IONET_API_KEY".

max_tokens

Integer specifying maximum tokens to generate per model. Default is 1024.

temperature

Numeric value controlling randomness (0-2). Default is 0.7.

timeout

Numeric value in seconds for request timeout per model. Default is 300.

parallel

Logical indicating whether to run models in parallel. Default is TRUE.

verbose

Logical indicating whether to show detailed progress. Default is TRUE.

refresh_models

Logical indicating whether to force-refresh the io.net model list before execution. Default is FALSE.

retries

Integer number of retries for transient errors (429/5xx/timeout/connection). Default is 1.

retry_wait

Base wait in seconds before retry (multiplied by attempt count). Default is 2.

monitor_timeout

Maximum seconds to wait for all async futures in the progress loop (prevents hangs). Default is 120.

Author

Satoshi Kume

Details

Multi-LLM via io.net API

Examples

if (FALSE) {
  # Set your io.net API key
  Sys.setenv(IONET_API_KEY = "your_ionet_api_key_here")

  # Basic usage - single random model (default)
  result_one <- multiLLMviaionet(
    prompt = "Explain quantum computing in simple terms"
  )
  print(result_one$summary)
  print_multiLLM_results(result_one)

  # Run first 5 models in list order (no random selection)
  result_top5 <- multiLLMviaionet(
    prompt = "What is machine learning?",
    max_models = 5,
    random_selection = FALSE
  )
  print_multiLLM_results(result_top5)

  # Random selection of 10 models from all available
  result_random10 <- multiLLMviaionet(
    prompt = "Write a Python function to calculate fibonacci numbers",
    max_models = 10,
    random_selection = TRUE,
    temperature = 0.3,
    streaming = FALSE
  )
  print_multiLLM_results(result_random10)
}