Skip to contents

Chat Completion with LM Studio

Source: R/chat.R
lms_chat.Rd

Send a prompt to a locally running LM Studio model. This wrapper automatically routes your request to the appropriate subfunction based on the selected API type.

Usage

lms_chat(
  model,
  input,
  system_prompt = NULL,
  host = "http://localhost:1234",
  api_type = c("openresponses", "openai", "native"),
  logprobs = FALSE,
  simplify = TRUE,
  ...,
  schema = NULL,
  ttl = NULL,
  previous_response_id = NULL,
  store = NULL,
  token = NULL
)

Arguments

model

Character. The name of the loaded model. Must be one name, given as a single string. The string must be valid in its declared encoding and not marked "bytes". A class, names, and the S4 bit are removed before the name is sent.

input

Character. The user prompt to send to the model, as one string. A character vector must hold no missing values, and its length must be one. Any other character value aborts before the check for a running server. To send several prompts, one request each, use lms_chat_batch(). A list is sent as given, as the input field or, on the "openai" route, as the content of the user message.

system_prompt

Character. An optional system prompt to guide model behavior.

host

Character. The base URL of the LM Studio server. Default is "http://localhost:1234".

api_type

Character. The LM Studio API endpoint to use. Options are "openresponses" (default), "openai", or "native".

logprobs

TRUE or FALSE. Whether to return the log probabilities of the generated tokens. Default is FALSE. Any other value, NULL and NA included, aborts before the check for a running server.

simplify

TRUE or FALSE. If TRUE, extracts the core text response. Default is TRUE. Any other value, NULL and NA included, aborts before the check for a running server.

...

Additional arguments passed to the selected API body. The package checks a stream here. A stream other than FALSE or NULL aborts before the call checks for a running server, because the package reads a whole reply and not a streamed one. An instructions here with api_type = "openresponses", the default, aborts before the request, because lms_chat() sets instructions from system_prompt on that route. A messages here with api_type = "openai" also aborts, because lms_chat() builds messages from system_prompt and input on that route. The message names the argument, and the abort has no condition class. Only these exact names abort. On the other routes, each goes into the request body.

schema

A JSON Schema that the reply must match, or NULL. It needs api_type = "openai", and any other api_type aborts before the request. See lms_chat_openai() for its form and for what is returned.

ttl

A whole number of seconds from 1 to .Machine$integer.max, or NULL. It is how long the model stays loaded with no request. It needs api_type = "openai", and any other api_type, the default included, aborts before the request. It has an effect only on a model that this request loads. The server loads a model that is not loaded yet when its just-in-time loading setting is on. A model that is already loaded keeps its idle time.

previous_response_id

One string, a value that carries a response_id attribute, or NULL. It names the stored reply that this chat continues. Pass the earlier reply itself, such as first after first <- lms_chat(...), or its id, attr(first, "response_id"). A value that carries a response_id attribute sends that attribute in place of the value, also when the value is a list or is itself an id. Only that exact attribute name is read. A value with no such attribute is sent as the id itself. So a reply that came back with no id, such as a native reply sent with store = FALSE, goes out as its own text, and the call raises rlmstudio_api_error with status 400. A reply text that breaks the id rules below, such as an empty text or a text of whitespace only, aborts before the request, as such an id does. The character vector that lms_chat_batch() returns with format = "vector" and the output column of its data frame carry no attribute. The data frame holds the ids in its response_id column.

The id, the attribute when there is one, must be valid in its declared encoding and not marked "bytes". A class, names, and the S4 bit are removed before the id is sent. NULL, the default, starts a new thread. It needs api_type = "native" or api_type = "openresponses". With api_type = "openai", an id or a value that carries one aborts before the request, because the OpenAI chat endpoint keeps no thread. As the id, NA, an empty string, a string of whitespace only, a value that is not a string, and more or fewer than one string abort before the check for a running server. For an attribute, the message names the attribute. Only this exact argument name is checked. A shortened name, such as previous, goes into the request body unchecked, under the name you wrote. An id that the server does not hold raises rlmstudio_api_error with status 400.

store

TRUE, FALSE, or NULL. Whether the server stores the reply, so that a later call can continue from its id. NULL, the default, sends no store field, so the server default applies, and that default stores the reply. TRUE and FALSE go out as a plain JSON true or false, also when the value has names, dimensions, or a class. Any other value, NA included, aborts before the check for a running server. A TRUE or FALSE needs api_type = "native" or api_type = "openresponses". NULL passes on every route. With api_type = "openai", TRUE or FALSE aborts before the request, because the OpenAI chat endpoint keeps no thread. Only this exact name is checked. A shortened name, such as sto, goes into the request body unchecked, under the name you wrote.

token

Character or NULL. An API token for a server that requires authentication. NULL reads the rlmstudio.token option and then the RLMSTUDIO_API_TOKEN environment variable. See rlmstudio_token.

Value

Depending on the arguments provided:

  • If simplify = FALSE, returns a parsed list of the raw JSON response.

  • If simplify = TRUE and logprobs = FALSE, returns a single character string containing the model's text response. With a schema, it returns the reply parsed into an R value instead.

  • If simplify = TRUE and logprobs = TRUE (and the chosen API type supports it), returns an object of class lms_chat_result containing both the text and a data.frame of token probabilities.

With simplify = TRUE and api_type = "native" or api_type = "openresponses", the value carries the id of the reply in a response_id attribute. Pass the value itself as previous_response_id to continue the thread, and the attribute is sent. The id is the response_id field of a native reply and the id field of an OpenResponses reply. The value has no attribute when that field is absent or is not one string. With api_type = "openai", or with simplify = FALSE, the value has no response_id attribute.

A native reply sent with store = FALSE has no response_id field, so its value has no attribute. An OpenResponses reply sent with store = FALSE still has an id, so its value carries the attribute, but the server does not hold that reply. A later call on the OpenResponses route that passes that id as previous_response_id raises rlmstudio_api_error with status 400 and the code "previous_response_not_found".

Details

This function calls lms_chat_openresponses(), lms_chat_openai(), or lms_chat_native(), according to api_type. It runs no request of its own. It can raise rlmstudio_no_server and rlmstudio_api_error through lms_chat_openresponses(), lms_chat_openai(), or lms_chat_native(). With simplify = TRUE, it can raise rlmstudio_bad_response through any of the three, for a reply that holds no readable answer text. With either setting of simplify, it raises rlmstudio_bad_response through any of the three for a status-200 body that does not parse as JSON. With either setting of simplify, it raises rlmstudio_model_mismatch through lms_chat_openresponses() or lms_chat_openai() for a reply from a model other than the one asked for.

Server not running

Functions that call the LM Studio REST API open a TCP connection to the hostname and port named in host before they send the request. A function that checks its own arguments does that first, so a bad model, job_id, input, inputs, messages, schema, ttl, previous_response_id, or batch_size, a bad type of list_models() or list_instances(), a bad TRUE or FALSE argument such as simplify, logprobs, quiet, force, or store, a store on the "openai" route of lms_chat() or lms_chat_batch(), a stream in the ... of a chat function, a logprobs in the ... of lms_chat_batch() or lms_chat_native(), a value in the ... of lms_chat_batch() that lms_chat() reads as an argument already given, an instructions or messages in the ... of lms_chat() or lms_chat_batch() on the route where lms_chat() sets it, or a name, id, or type string that is not valid in its encoding or is marked as bytes, aborts with an argument message and no condition class even when the server is down. A condition of class rlmstudio_no_server is raised when that connection cannot be opened. A refused connection raises it. So do an address the package cannot parse and a hostname that does not resolve. An address that neither accepts nor refuses the connection also raises it. That case waits for the operating system to give up, which can take a minute. Start the server with lms_server_start(), or give host the address that your server listens on.

The check reads the port and nothing else. Any process holding that port accepts the connection, so the condition is not raised even though no LM Studio server is there. The call then does not raise rlmstudio_no_server, and what it does depends on what answers. On a status-200 body that does not parse as JSON, the chat functions, lms_embed(), list_models(), list_instances(), lms_load(), lms_download(), lms_download_status(), and lms_unload_all() raise rlmstudio_bad_response. lms_unload() does not read the body, so it can report success. A body that parses as JSON but has another shape can come back unchanged with simplify = FALSE. With simplify = TRUE, the chat functions and lms_embed() raise rlmstudio_bad_response for it. list_models() and list_instances() raise it for a model list with another shape, and so do lms_unload_all() and lms_load() without force = TRUE, which read that list. lms_chat_openai() and lms_chat_openresponses() raise it for such a model list too, with either setting of simplify, through the model lookup that a reply from another model starts. lms_chat() raises it through them. lms_load(), lms_download(), and lms_download_status() raise it for a reply of their own with another shape, such as {}. A process that does not answer in HTTP gives an httr2_failure error. Use lms_server_ready() for the stronger test: it asks the host for a model list and reports TRUE only for a model list that list_models() can read.

lms_chat_batch() checks the server once before its first input, and lms_chat() checks it again for each input. If that check finds the server gone during the batch, the batch aborts with rlmstudio_no_server, and no request goes out after that. The condition then carries a results field, a list as long as inputs. Its elements before the lost input hold the values that format = "list" returns for those inputs. The element of the lost input and every element after it are NULL. The check before the first input adds no results field. A connection that fails after the check passes, such as a server that stops during a request, raises an httr2_failure error instead. That error aborts the batch and carries no results field.

lms_embed() checks the server before each request, and a request carries at most batch_size inputs. If a check after the first request finds the server gone, the call aborts with rlmstudio_no_server. Once a request has succeeded, the condition carries a results field. With simplify = TRUE, results is a matrix with NA in each row whose embedding did not arrive. With simplify = FALSE, it is a list with one element per batch, with NULL in the element of the request that ended the call and in every element after it.

API failure

A condition of class rlmstudio_api_error is raised when a REST call returns a response that the wrapper treats as a failure. The condition carries a status field, which holds the HTTP response status as an integer. It also carries a code field. The field holds the string at error.code of the response body, such as "model_not_found". It is NULL when the body does not parse, when error is not a JSON object, or when its code is not one string.

lms_chat_batch() aborts on it when its status is 401, 403, or 404, and when its status is 400 and its code is "model_not_found". No request goes out after that input. The condition then carries a results field that follows the rule for a lost server in the "Server not running" section: its elements before the failed input hold the values that format = "list" returns for those inputs, and the element of the failed input and every element after it are NULL. For any other status, the element of the failed input holds the condition, or NA where the result is text, and the batch warns once and goes on. See the details of lms_chat_batch().

lms_embed() follows the same rule for each request. A 401, 403, or 404 aborts the call, and once a request has succeeded the condition carries a results field, a matrix or a list as the "Server not running" section describes. Any other status fails the inputs of that request alone, and the call warns once and goes on. If every request fails, the call aborts with the first condition. See the details of lms_embed().

Malformed response

A condition of class rlmstudio_bad_response is raised when the server answers with a status the wrapper accepts and a body the wrapper cannot read. It is raised where a wrapper checks the body before it reshapes it, rather than indexing straight into whatever arrived. Eleven functions raise it for a body they cannot read: lms_embed(), lms_chat(), lms_chat_native(), lms_chat_openresponses(), lms_chat_openai(), list_models(), list_instances(), lms_load(), lms_download(), lms_download_status(), and lms_unload_all(). lms_chat() raises it through the chat function it calls. lms_unload_all() raises it through list_models(), and so does lms_load() unless force = TRUE. lms_chat_batch() raises it only as rlmstudio_model_mismatch, which the "Reply from another model" section of lms_chat_openai() describes.

All eleven raise it for a status-200 body that does not parse as JSON, such as an HTML page from a proxy, JSON text that stops part way, or an empty body. In the functions that take simplify, the body is parsed before simplify is read, so the condition is raised whatever simplify is. The body is parsed by its content and not by its Content-Type header, so valid JSON under text/plain is read as JSON. The body is read as JSON text and nothing else. A body whose text is a URL or the path of a file does not parse, and the package does not fetch the URL or read the file. The message says that the body did not parse as JSON and that something other than LM Studio may be answering on the host. It does not hold the body text.

The condition carries a status field, which holds the HTTP response status as an integer. Today the status is always 200: each of these functions reads the body only after a 200, and reports every other status as an rlmstudio_api_error instead.

Malformed model list

list_models() also raises it for a status-200 model list with the wrong shape. lms_unload_all() and lms_load() without force = TRUE raise it through list_models(). lms_chat_openai() and lms_chat_openresponses() apply these rules to the model list of their model lookup, as the "Reply from another model" section of lms_chat_openai() describes. A model list must follow four rules. Each field is read by its exact name, so a field named keyX does not stand in for key.

  1. The body is a JSON object whose models field is an array. The array can be empty.

  2. Each entry of models is a JSON object. Its type and key are strings, and its loaded_instances is an array.

  3. The size_bytes of an entry is a number, or absent, or null. So is its max_context_length.

  4. Each entry of loaded_instances is a JSON object whose id is a string with a character that is not whitespace.

The rules are checked before the type and loaded filters, so an entry that the filters drop can still raise the condition. The message names the field or entry that broke a rule. lms_server_ready() applies the same rules and returns FALSE for a body that breaks one.

list_instances() applies the four rules too, and two more. They cover only an entry whose type is in its type argument and that has at least one loaded instance. The display_name of such an entry is a string, or absent, or null. The config of each of its instances is a JSON object, or absent, or null. list_models() and lms_server_ready() do not apply these two rules.

Malformed chat reply

lms_chat_native() and lms_chat_openresponses() raise it with simplify = TRUE when the reply holds no readable answer text. Both read the answer from the items of type "message" in the output array. They raise it when output is missing, empty, or not an array, or when an item in it is not a JSON object. They also raise it when no item has the type "message", as in a reply that holds only reasoning or a tool call. The text of a message must be one string. For lms_chat_native() that is the content of the item. For lms_chat_openresponses() it is the text of each part of type "output_text". For lms_chat_openresponses() only, the content of each message must be an array of JSON objects, and the messages together must hold at least one "output_text" part.

Apart from a body that does not parse as JSON, lms_chat_openai() raises it in four cases, all only with simplify = TRUE. The first case is a response whose choices field is missing, empty, or not an array, or whose first element is not a JSON object with a message object in it, so there is no reply to read. This case is raised with or without a schema, and with logprobs = TRUE as well. The second case is a reply that does not parse. A schema was given, logprobs = FALSE, and the reply content is not one string of valid JSON. The third case is reply content that is not one string, such as null, a missing content field, a number, or an array. A reply that holds only a tool call has null content. This case is raised without a schema, and with logprobs = TRUE with or without one. The fourth case is a cut-off reply. A schema was given, logprobs = FALSE, and the server reports the finish reason "length". This case is raised also when the reply content parses, because a reply that stops part way can parse to a wrong value, such as the first digit of a longer number. In the second, third, and fourth cases, if the server reports the finish reason "length", a length limit ended the reply. The limit is max_tokens or the context length of the model. The message then says so and names both.

With simplify = TRUE, lms_chat_native(), lms_chat_openresponses(), and lms_chat_openai() also raise it for a body that is a bare JSON value, such as 5, "s", or true. The message says that the response body is not a JSON object. A body of null gets the message about its missing output or choices field instead. lms_chat() can raise the condition through all three chat functions.

lms_chat_batch() does not abort on it, except on the subclass rlmstudio_model_mismatch, as the "Reply from another model" section of lms_chat_openai() says. The element of the failed input holds the condition, or NA where the result is text, and the batch warns once and goes on. See the details of lms_chat_batch().

A condition from lms_chat_openai() about its reply also carries two more fields. A condition from the model-list lookup does not, as the "Reply from another model" section of lms_chat_openai() says. The content field holds the reply content of the first choice, and the finish_reason field holds the finish reason of the first choice. Both are NULL for a response with no choices. In the third case, content holds the value that was read, which is NULL for null or missing content. For the second, third, and fourth cases, the message names the content field, so you can read what the model wrote without a second request. The other messages of the chat functions about the reply name simplify = FALSE, which returns the body unchanged, with one exception. A body that did not parse as JSON is checked before that argument is read, so its message points at the host instead. For such a body, the content and finish_reason fields of a condition from lms_chat_openai() are NULL. The messages of rlmstudio_model_mismatch and of the model-list lookup do not name simplify = FALSE, because the check runs with either setting of simplify.

Malformed logprobs

With simplify = TRUE and logprobs = TRUE, lms_chat_openresponses() also raises it for a logprobs value that breaks one of these rules. The logprobs value of each "output_text" part is checked. A logprobs value, token, logprob, or top_logprobs that is null or absent passes its rule. A null step or candidate breaks rule 2 or rule 6.

  1. The value is an array.

  2. Each step in the array is a JSON object.

  3. The token of a step is a string.

  4. The logprob of a step is a number.

  5. The top_logprobs of a step is an array.

  6. Each candidate in top_logprobs is a JSON object.

  7. The token and logprob of each candidate follow rules 3 and 4.

The parts are checked in order, then the steps of a part, then the candidates of a step, one at a time. Within a step, rules 3, 4, and 5 come before the candidates. The message names the first broken rule that this order reaches. These checks run only after the text of every "output_text" part is read, so a reply that also has a bad text in any part gets the text message. Parts of other types, such as a refusal, are not checked, and with logprobs = FALSE no part is checked. Fields are read by their exact names, so a field whose name only starts with the one asked for, such as tokenX, reads as absent and gives NA in the data frame.

Cut-off reply

A warning of class rlmstudio_reply_cut_off is given when a length limit ended a reply that lms_chat_openai() returns as text. The server reports this with the finish reason "length". The limit is max_tokens or the context length of the model, and the reply does not say which one. The message names both. The warning is given with simplify = TRUE for reply content that is one string, in three settings: no schema, logprobs = TRUE, and a schema with logprobs = TRUE. The call returns what the same reply returns without the cut-off. A schema reply with logprobs = FALSE raises rlmstudio_bad_response instead, as the "Malformed chat reply" section says. The warning and that abort read the finish reason of the first choice, which is the choice that the call reads. No other element of choices is read. With simplify = FALSE, the call returns the body with every choice and no warning.

lms_chat() gives the warning through lms_chat_openai() with api_type = "openai". lms_chat_batch() gives one warning of this class for the whole batch in place of one for each input. It names the count and the positions of the cut-off inputs, and each of those elements keeps its reply. It comes after the warning about failed inputs. A batch that aborts with a results field gives this warning before the abort. It names the cut-off inputs whose replies the results field of the abort holds. The native and OpenResponses routes give no such warning.

The warning shows whatever quiet and the rlmstudio.quiet option say, because it is the only sign that an answer is not complete.

Reply from another model

lms_chat_openai() and lms_chat_openresponses() read the model field of each status-200 reply, with either setting of simplify. On LM Studio 0.4.25+1 with one chat model loaded, both routes answered a model name that the server could not find with status 200 and a reply from the loaded model. With two chat models loaded, they answered with status 400 and the code "model_not_found", as the "API failure" section describes. The model field names the loaded instance that answered. Its id can differ from the key that the call asked for.

If the model field equals the name that the call sent, the reply is accepted. If it differs, the call sends one request for the model list to api/v1/models on the same host, with the token of the call, and prints no message. The reply is accepted if a model whose key equals the asked name in any letter case has a loaded instance whose id is the model field. Otherwise the call aborts with a condition of class rlmstudio_model_mismatch. That condition is also an rlmstudio_bad_response, and its status is 200. Its model field holds the asked name, and its reply_model field holds the model field of the reply. The message names both. A condition from lms_chat_openai() also carries the content and finish_reason fields, both NULL.

The check runs before the reply is read, so it also runs with logprobs = TRUE, with a schema on lms_chat_openai(), and for a reply with no answer text. A reply is not checked if its body is not a JSON object, if it has no model field, or if its model is not one string or holds only whitespace. A body that is not a JSON object is returned with simplify = FALSE, and with simplify = TRUE it raises the error that the "Malformed chat reply" section describes. If the instance that answered is unloaded before the model-list request, the call aborts, also when that instance belongs to the asked model.

If the model-list request fails, the call raises the condition of that failure. The message opens with the label of the chat function, followed by "because the model-list lookup failed". A status other than 200 raises rlmstudio_api_error. A body that does not parse as JSON, or that breaks a rule of a model list in the "Malformed model list" section, raises rlmstudio_bad_response. A server that the port check before the request cannot reach raises rlmstudio_no_server. A condition from the lookup does not carry the content and finish_reason fields, also when lms_chat_openai() raises it.

lms_chat() runs the check through lms_chat_openai() and lms_chat_openresponses(). lms_chat_native() does not check the reply. On LM Studio 0.4.25+1, /api/v1/chat answered a name that it could not find with status 404, which raises rlmstudio_api_error.

lms_chat_batch() aborts at the first input that raises rlmstudio_model_mismatch, on the "openai" and "openresponses" routes. It also aborts at an rlmstudio_api_error with status 400 whose code is "model_not_found". No request goes out after that input. In both cases, the condition carries a results field that follows the rule for a lost server in the "Server not running" section.

Long prompts

A prompt longer than the context length of the loaded model fails. On LM Studio 0.4.25+1, with google/gemma-3-1b loaded at a context_length of 512, the /v1/responses and /api/v1/chat routes answered a longer prompt with status 500. The /v1/chat/completions route answered it with status 400. Each message began "The number of tokens to keep from the initial prompt is greater than the context length". The chat functions raise such a reply as an rlmstudio_api_error, and the message holds the server text. lms_chat_batch() fails that input alone and goes on to the next input. With format = "list", the element of that input holds the condition. With format = "vector", it holds NA.

To fit a longer prompt, load the model with a larger context_length in lms_load(). When the server reports it, the max_context_length column of list_models(detailed = TRUE) gives the largest context length that the model list reports for each model. If context_length is larger, the server loads the model with the asked value and gives no message. On LM Studio 0.4.25+1, it loaded google/gemma-3-1b at 65536 tokens, above its maximum of 32768. So lms_load() gives a warning of class rlmstudio_context_above_max that names both numbers, and then sends the load. The rlmstudio.quiet option does not hide it. The warning needs the model list, so it comes only with force = FALSE, and only when the list has a maximum for the model and the model is not loaded yet.

On LM Studio 0.4.25+1, the load endpoint answered a rope_frequency_scale field in ... with status 400 and the code "unrecognized_keys". So lms_load() cannot set RoPE scaling through that field. RoPE scaling stretches the position encoding of a model past its trained length.

To see how many tokens a prompt took, use lms_chat_batch(format = "data.frame"). Its input_tokens column holds the prompt token count that the server reports for each reply, on every route.