Skip to contents

Report the language-model routes available from this machine

Source: R/llm_native.R
bricklayer_llm_status.Rd

Shows, in one table, every way bricklayer_llm_ask could reach a language model from this machine and whether each is ready: your own endpoint, each provider whose key is set, and a local Ollama server. Use it to find out why a question fails with "no route", or which model a route would use. The command line prints it as part of rmoriebricklayer doctor.

Usage

bricklayer_llm_status()

Value

A data frame with one row per route and three character columns: route ("own endpoint", a provider label such as "Anthropic", "other providers" or "local Ollama"), status ("not set" or "configured" for your own endpoint; "key set" for a provider; "no key" for the shared row; "disabled", "not running", "no models" or "available" for Ollama) and detail (the address and model, or what to set up).

Details

Rows come in the order bricklayer_llm_ask() tries the routes. The providers without a key share one row that names each one with its key variable. Key values are never shown. Nothing is sent to a provider; the only network contact is a GET /api/tags to the Ollama address (OLLAMA_HOST, default http://localhost:11434; 2 second timeout) unless OLLAMA_HOST=off.

A settings file (llm.json next to the credentials file) that cannot be read does not make the call fail: the result is then a single row with route = "settings" and status = "unreadable", with the reason and the command that moves the file aside.

See also

bricklayer_llm_route_info for the one route a question would take; bricklayer_llm_providers.

Examples

old <- Sys.getenv(c("XDG_CONFIG_HOME", "OLLAMA_HOST",
                    "MORIE_LLM_BASE_URL", "MORIE_LLM_MODEL"),
                  unset = NA)
# a scratch settings folder and the local route off: no network at all
Sys.setenv(XDG_CONFIG_HOME = tempfile(), OLLAMA_HOST = "off")
st <- bricklayer_llm_status()
st[, c("route", "status")]
#>             route   status
#> 1    own endpoint  not set
#> 2 other providers   no key
#> 3    local Ollama disabled

# your own OpenAI-compatible server shows as configured (it is not contacted)
Sys.setenv(MORIE_LLM_BASE_URL = "http://localhost:8080", MORIE_LLM_MODEL = "llama3")
bricklayer_llm_status()[1, ]
#>          route     status                               detail
#> 1 own endpoint configured http://localhost:8080  model: llama3

for (v in names(old)) {
  if (is.na(old[[v]])) Sys.unsetenv(v) else do.call(Sys.setenv, as.list(old[v]))
}