searchengines-web/MeiliSearch
searchengines-web/MeiliSearch
1
0
Fork 0
mirror of https://github.com/meilisearch/MeiliSearch synced 2025-07-04 20:37:15 +02:00
Code Issues Releases Wiki Activity

Merge a76a3e8f11 into fef089c7b6

Browse source
This commit is contained in:
Clément Renault 2025-07-03 18:25:58 +02:00 • committed by GitHub
commit 792b7833e0
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 61 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

27
crates/meilisearch/src/metrics.rs
View file

@ -15,6 +15,33 @@ lazy_static! {
"Meilisearch number of degraded search requests"
))
.expect("Can't create a metric");
pub static ref MEILISEARCH_CHAT_SEARCH_REQUESTS: IntCounterVec = register_int_counter_vec!(
opts!(
"meilisearch_chat_search_requests",
"Meilisearch number of search requests performed by the chat route itself"
),
&["type"]
)
.expect("Can't create a metric");
pub static ref MEILISEARCH_CHAT_PROMPT_TOKENS_USAGE: IntCounterVec = register_int_counter_vec!(
opts!("meilisearch_chat_prompt_tokens_usage", "Meilisearch Chat Prompt Tokens Usage"),
&["workspace", "model"]
)
.expect("Can't create a metric");
pub static ref MEILISEARCH_CHAT_COMPLETION_TOKENS_USAGE: IntCounterVec =
register_int_counter_vec!(
opts!(
"meilisearch_chat_completion_tokens_usage",
"Meilisearch Chat Completion Tokens Usage"
),
&["workspace", "model"]
)
.expect("Can't create a metric");
pub static ref MEILISEARCH_CHAT_TOTAL_TOKENS_USAGE: IntCounterVec = register_int_counter_vec!(
opts!("meilisearch_chat_total_tokens_usage", "Meilisearch Chat Total Tokens Usage"),
&["workspace", "model"]
)
.expect("Can't create a metric");
pub static ref MEILISEARCH_DB_SIZE_BYTES: IntGauge =
register_int_gauge!(opts!("meilisearch_db_size_bytes", "Meilisearch DB Size In Bytes"))
.expect("Can't create a metric");

40
crates/meilisearch/src/routes/chats/chat_completions.rs
View file

@ -13,9 +13,9 @@ use async_openai::types::{
ChatCompletionRequestDeveloperMessageContent, ChatCompletionRequestMessage,
ChatCompletionRequestSystemMessage, ChatCompletionRequestSystemMessageContent,
ChatCompletionRequestToolMessage, ChatCompletionRequestToolMessageContent,
ChatCompletionStreamResponseDelta, ChatCompletionToolArgs, ChatCompletionToolType,
CreateChatCompletionRequest, CreateChatCompletionStreamResponse, FinishReason, FunctionCall,
FunctionCallStream, FunctionObjectArgs,
ChatCompletionStreamOptions, ChatCompletionStreamResponseDelta, ChatCompletionToolArgs,
ChatCompletionToolType, CreateChatCompletionRequest, CreateChatCompletionStreamResponse,
FinishReason, FunctionCall, FunctionCallStream, FunctionObjectArgs,
};
use async_openai::Client;
use bumpalo::Bump;
@ -48,7 +48,11 @@ use crate::analytics::Analytics;
use crate::error::MeilisearchHttpError;
use crate::extractors::authentication::policies::ActionPolicy;
use crate::extractors::authentication::{extract_token_from_request, GuardedData, Policy as _};
use crate::metrics::MEILISEARCH_DEGRADED_SEARCH_REQUESTS;
use crate::metrics::{
MEILISEARCH_CHAT_COMPLETION_TOKENS_USAGE, MEILISEARCH_CHAT_PROMPT_TOKENS_USAGE,
MEILISEARCH_CHAT_SEARCH_REQUESTS, MEILISEARCH_CHAT_TOTAL_TOKENS_USAGE,
MEILISEARCH_DEGRADED_SEARCH_REQUESTS,
};
use crate::routes::chats::utils::SseEventSender;
use crate::routes::indexes::search::search_kind;
use crate::search::{add_search_rules, prepare_search, search_from_kind, SearchQuery};
@ -286,7 +290,7 @@ async fn process_search_request(
let output = output?;
let mut documents = Vec::new();
if let Ok((ref rtxn, ref search_result)) = output {
// aggregate.succeed(search_result);
MEILISEARCH_CHAT_SEARCH_REQUESTS.with_label_values(&["internal"]).inc();
if search_result.degraded {
MEILISEARCH_DEGRADED_SEARCH_REQUESTS.inc();
}
@ -488,6 +492,7 @@ async fn streamed_chat(
let (tx, rx) = tokio::sync::mpsc::channel(10);
let tx = SseEventSender::new(tx);
let workspace_uid = workspace_uid.to_string();
let _join_handle = Handle::current().spawn(async move {
let client = Client::with_config(config.clone());
let mut global_tool_calls = HashMap::<u32, Call>::new();
@ -497,6 +502,7 @@ async fn streamed_chat(
let output = run_conversation(
&index_scheduler,
&auth_ctrl,
&workspace_uid,
&search_queue,
&auth_token,
&client,
@ -534,6 +540,7 @@ async fn run_conversation<C: async_openai::config::Config>(
Data<IndexScheduler>,
>,
auth_ctrl: &web::Data<AuthController>,
workspace_uid: &str,
search_queue: &web::Data<SearchQueue>,
auth_token: &str,
client: &Client<C>,
@ -543,13 +550,34 @@ async fn run_conversation<C: async_openai::config::Config>(
global_tool_calls: &mut HashMap<u32, Call>,
function_support: FunctionSupport,
) -> Result<ControlFlow<Option<FinishReason>, ()>, SendError<Event>> {
use DbChatCompletionSource::*;
let mut finish_reason = None;
chat_completion.stream_options = match source {
OpenAi | AzureOpenAi => Some(ChatCompletionStreamOptions { include_usage: true }),
Mistral | VLlm => None,
};
// safety: unwrap: can only happens if `stream` was set to `false`
let mut response = client.chat().create_stream(chat_completion.clone()).await.unwrap();
while let Some(result) = response.next().await {
match result {
Ok(resp) => {
let choice = &resp.choices[0];
if let Some(usage) = resp.usage.as_ref() {
MEILISEARCH_CHAT_PROMPT_TOKENS_USAGE
.with_label_values(&[workspace_uid, &chat_completion.model])
.inc_by(usage.prompt_tokens as u64);
MEILISEARCH_CHAT_COMPLETION_TOKENS_USAGE
.with_label_values(&[workspace_uid, &chat_completion.model])
.inc_by(usage.completion_tokens as u64);
MEILISEARCH_CHAT_TOTAL_TOKENS_USAGE
.with_label_values(&[workspace_uid, &chat_completion.model])
.inc_by(usage.total_tokens as u64);
}
let choice = match resp.choices.first() {
Some(choice) => choice,
None => break,
};
finish_reason = choice.finish_reason;
let ChatCompletionStreamResponseDelta { ref tool_calls, .. } = &choice.delta;