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

Revert "Merge remote-tracking branch 'origin/main' into release-v1.7.1"

Browse source 
This reverts commit bd74cce86a, reversing
changes made to d2f77e88bd.
This commit is contained in:
Tamo 2024-03-19 16:56:21 +01:00
parent bd74cce86a
commit 567194b925
34 changed files with 614 additions and 1788 deletions
Download patch file Download diff file Expand all files Collapse all files

39
milli/src/vector/error.rs
View file

@ -2,7 +2,6 @@ use std::path::PathBuf;
use hf_hub::api::sync::ApiError;
use super::ollama::OllamaError;
use crate::error::FaultSource;
use crate::vector::openai::OpenAiError;
@ -72,17 +71,6 @@ pub enum EmbedErrorKind {
OpenAiRuntimeInit(std::io::Error),
#[error("initializing web client for sending embedding requests failed: {0}")]
InitWebClient(reqwest::Error),
// Dedicated Ollama error kinds, might have to merge them into one cohesive error type for all backends.
#[error("unexpected response from Ollama: {0}")]
OllamaUnexpected(reqwest::Error),
#[error("sent too many requests to Ollama: {0}")]
OllamaTooManyRequests(OllamaError),
#[error("received internal error from Ollama: {0}")]
OllamaInternalServerError(OllamaError),
#[error("model not found. Meilisearch will not automatically download models from the Ollama library, please pull the model manually: {0}")]
OllamaModelNotFoundError(OllamaError),
#[error("received unhandled HTTP status code {0} from Ollama")]
OllamaUnhandledStatusCode(u16),
}
impl EmbedError {
@ -141,26 +129,6 @@ impl EmbedError {
pub fn openai_initialize_web_client(inner: reqwest::Error) -> Self {
Self { kind: EmbedErrorKind::InitWebClient(inner), fault: FaultSource::Runtime }
}
pub(crate) fn ollama_unexpected(inner: reqwest::Error) -> EmbedError {
Self { kind: EmbedErrorKind::OllamaUnexpected(inner), fault: FaultSource::Bug }
}
pub(crate) fn ollama_model_not_found(inner: OllamaError) -> EmbedError {
Self { kind: EmbedErrorKind::OllamaModelNotFoundError(inner), fault: FaultSource::User }
}
pub(crate) fn ollama_too_many_requests(inner: OllamaError) -> EmbedError {
Self { kind: EmbedErrorKind::OllamaTooManyRequests(inner), fault: FaultSource::Runtime }
}
pub(crate) fn ollama_internal_server_error(inner: OllamaError) -> EmbedError {
Self { kind: EmbedErrorKind::OllamaInternalServerError(inner), fault: FaultSource::Runtime }
}
pub(crate) fn ollama_unhandled_status_code(code: u16) -> EmbedError {
Self { kind: EmbedErrorKind::OllamaUnhandledStatusCode(code), fault: FaultSource::Bug }
}
}
#[derive(Debug, thiserror::Error)]
@ -227,13 +195,6 @@ impl NewEmbedderError {
}
}
pub fn ollama_could_not_determine_dimension(inner: EmbedError) -> NewEmbedderError {
Self {
kind: NewEmbedderErrorKind::CouldNotDetermineDimension(inner),
fault: FaultSource::User,
}
}
pub fn openai_invalid_api_key_format(inner: reqwest::header::InvalidHeaderValue) -> Self {
Self { kind: NewEmbedderErrorKind::InvalidApiKeyFormat(inner), fault: FaultSource::User }
}

18
milli/src/vector/mod.rs
View file

@ -10,8 +10,6 @@ pub mod manual;
pub mod openai;
pub mod settings;
pub mod ollama;
pub use self::error::Error;
pub type Embedding = Vec<f32>;
@ -78,7 +76,6 @@ pub enum Embedder {
HuggingFace(hf::Embedder),
OpenAi(openai::Embedder),
UserProvided(manual::Embedder),
Ollama(ollama::Embedder),
}
#[derive(Debug, Clone, Default, serde::Deserialize, serde::Serialize)]
@ -130,7 +127,6 @@ impl IntoIterator for EmbeddingConfigs {
pub enum EmbedderOptions {
HuggingFace(hf::EmbedderOptions),
OpenAi(openai::EmbedderOptions),
Ollama(ollama::EmbedderOptions),
UserProvided(manual::EmbedderOptions),
}
@ -148,10 +144,6 @@ impl EmbedderOptions {
pub fn openai(api_key: Option<String>) -> Self {
Self::OpenAi(openai::EmbedderOptions::with_default_model(api_key))
}
pub fn ollama() -> Self {
Self::Ollama(ollama::EmbedderOptions::with_default_model())
}
}
impl Embedder {
@ -159,7 +151,6 @@ impl Embedder {
Ok(match options {
EmbedderOptions::HuggingFace(options) => Self::HuggingFace(hf::Embedder::new(options)?),
EmbedderOptions::OpenAi(options) => Self::OpenAi(openai::Embedder::new(options)?),
EmbedderOptions::Ollama(options) => Self::Ollama(ollama::Embedder::new(options)?),
EmbedderOptions::UserProvided(options) => {
Self::UserProvided(manual::Embedder::new(options))
}
@ -176,10 +167,6 @@ impl Embedder {
let client = embedder.new_client()?;
embedder.embed(texts, &client).await
}
Embedder::Ollama(embedder) => {
let client = embedder.new_client()?;
embedder.embed(texts, &client).await
}
Embedder::UserProvided(embedder) => embedder.embed(texts),
}
}
@ -194,7 +181,6 @@ impl Embedder {
match self {
Embedder::HuggingFace(embedder) => embedder.embed_chunks(text_chunks),
Embedder::OpenAi(embedder) => embedder.embed_chunks(text_chunks),
Embedder::Ollama(embedder) => embedder.embed_chunks(text_chunks),
Embedder::UserProvided(embedder) => embedder.embed_chunks(text_chunks),
}
}
@ -203,7 +189,6 @@ impl Embedder {
match self {
Embedder::HuggingFace(embedder) => embedder.chunk_count_hint(),
Embedder::OpenAi(embedder) => embedder.chunk_count_hint(),
Embedder::Ollama(embedder) => embedder.chunk_count_hint(),
Embedder::UserProvided(_) => 1,
}
}
@ -212,7 +197,6 @@ impl Embedder {
match self {
Embedder::HuggingFace(embedder) => embedder.prompt_count_in_chunk_hint(),
Embedder::OpenAi(embedder) => embedder.prompt_count_in_chunk_hint(),
Embedder::Ollama(embedder) => embedder.prompt_count_in_chunk_hint(),
Embedder::UserProvided(_) => 1,
}
}
@ -221,7 +205,6 @@ impl Embedder {
match self {
Embedder::HuggingFace(embedder) => embedder.dimensions(),
Embedder::OpenAi(embedder) => embedder.dimensions(),
Embedder::Ollama(embedder) => embedder.dimensions(),
Embedder::UserProvided(embedder) => embedder.dimensions(),
}
}
@ -230,7 +213,6 @@ impl Embedder {
match self {
Embedder::HuggingFace(embedder) => embedder.distribution(),
Embedder::OpenAi(embedder) => embedder.distribution(),
Embedder::Ollama(embedder) => embedder.distribution(),
Embedder::UserProvided(_embedder) => None,
}
}

307
milli/src/vector/ollama.rs
View file

@ -1,307 +0,0 @@
// Copied from "openai.rs" with the sections I actually understand changed for Ollama.
// The common components of the Ollama and OpenAI interfaces might need to be extracted.
use std::fmt::Display;
use reqwest::StatusCode;
use super::error::{EmbedError, NewEmbedderError};
use super::openai::Retry;
use super::{DistributionShift, Embedding, Embeddings};
#[derive(Debug)]
pub struct Embedder {
headers: reqwest::header::HeaderMap,
options: EmbedderOptions,
}
#[derive(Debug, Clone, Hash, PartialEq, Eq, serde::Deserialize, serde::Serialize)]
pub struct EmbedderOptions {
pub embedding_model: EmbeddingModel,
}
#[derive(
Debug, Clone, Hash, PartialEq, Eq, serde::Serialize, serde::Deserialize, deserr::Deserr,
)]
#[deserr(deny_unknown_fields)]
pub struct EmbeddingModel {
name: String,
dimensions: usize,
}
#[derive(Debug, serde::Serialize)]
struct OllamaRequest<'a> {
model: &'a str,
prompt: &'a str,
}
#[derive(Debug, serde::Deserialize)]
struct OllamaResponse {
embedding: Embedding,
}
#[derive(Debug, serde::Deserialize)]
pub struct OllamaError {
error: String,
}
impl EmbeddingModel {
pub fn max_token(&self) -> usize {
// this might not be the same for all models
8192
}
pub fn default_dimensions(&self) -> usize {
// Dimensions for nomic-embed-text
768
}
pub fn name(&self) -> String {
self.name.clone()
}
pub fn from_name(name: &str) -> Self {
Self { name: name.to_string(), dimensions: 0 }
}
pub fn supports_overriding_dimensions(&self) -> bool {
false
}
}
impl Default for EmbeddingModel {
fn default() -> Self {
Self { name: "nomic-embed-text".to_string(), dimensions: 0 }
}
}
impl EmbedderOptions {
pub fn with_default_model() -> Self {
Self { embedding_model: Default::default() }
}
pub fn with_embedding_model(embedding_model: EmbeddingModel) -> Self {
Self { embedding_model }
}
}
impl Embedder {
pub fn new_client(&self) -> Result<reqwest::Client, EmbedError> {
reqwest::ClientBuilder::new()
.default_headers(self.headers.clone())
.build()
.map_err(EmbedError::openai_initialize_web_client)
}
pub fn new(options: EmbedderOptions) -> Result<Self, NewEmbedderError> {
let mut headers = reqwest::header::HeaderMap::new();
headers.insert(
reqwest::header::CONTENT_TYPE,
reqwest::header::HeaderValue::from_static("application/json"),
);
let mut embedder = Self { options, headers };
let rt = tokio::runtime::Builder::new_current_thread()
.enable_io()
.enable_time()
.build()
.map_err(EmbedError::openai_runtime_init)
.map_err(NewEmbedderError::ollama_could_not_determine_dimension)?;
// Get dimensions from Ollama
let request =
OllamaRequest { model: &embedder.options.embedding_model.name(), prompt: "test" };
// TODO: Refactor into shared error type
let client = embedder
.new_client()
.map_err(NewEmbedderError::ollama_could_not_determine_dimension)?;
rt.block_on(async move {
let response = client
.post(get_ollama_path())
.json(&request)
.send()
.await
.map_err(EmbedError::ollama_unexpected)
.map_err(NewEmbedderError::ollama_could_not_determine_dimension)?;
// Process error in case model not found
let response = Self::check_response(response).await.map_err(|_err| {
let e = EmbedError::ollama_model_not_found(OllamaError {
error: format!("model: {}", embedder.options.embedding_model.name()),
});
NewEmbedderError::ollama_could_not_determine_dimension(e)
})?;
let response: OllamaResponse = response
.json()
.await
.map_err(EmbedError::ollama_unexpected)
.map_err(NewEmbedderError::ollama_could_not_determine_dimension)?;
let embedding = Embeddings::from_single_embedding(response.embedding);
embedder.options.embedding_model.dimensions = embedding.dimension();
tracing::info!(
"ollama model {} with dimensionality {} added",
embedder.options.embedding_model.name(),
embedding.dimension()
);
Ok(embedder)
})
}
async fn check_response(response: reqwest::Response) -> Result<reqwest::Response, Retry> {
if !response.status().is_success() {
// Not the same number of possible error cases covered as with OpenAI.
match response.status() {
StatusCode::TOO_MANY_REQUESTS => {
let error_response: OllamaError = response
.json()
.await
.map_err(EmbedError::ollama_unexpected)
.map_err(Retry::retry_later)?;
return Err(Retry::rate_limited(EmbedError::ollama_too_many_requests(
OllamaError { error: error_response.error },
)));
}
StatusCode::SERVICE_UNAVAILABLE => {
let error_response: OllamaError = response
.json()
.await
.map_err(EmbedError::ollama_unexpected)
.map_err(Retry::retry_later)?;
return Err(Retry::retry_later(EmbedError::ollama_internal_server_error(
OllamaError { error: error_response.error },
)));
}
StatusCode::NOT_FOUND => {
let error_response: OllamaError = response
.json()
.await
.map_err(EmbedError::ollama_unexpected)
.map_err(Retry::give_up)?;
return Err(Retry::give_up(EmbedError::ollama_model_not_found(OllamaError {
error: error_response.error,
})));
}
code => {
return Err(Retry::give_up(EmbedError::ollama_unhandled_status_code(
code.as_u16(),
)));
}
}
}
Ok(response)
}
pub async fn embed(
&self,
texts: Vec<String>,
client: &reqwest::Client,
) -> Result<Vec<Embeddings<f32>>, EmbedError> {
// Ollama only embedds one document at a time.
let mut results = Vec::with_capacity(texts.len());
// The retry loop is inside the texts loop, might have to switch that around
for text in texts {
// Retries copied from openai.rs
for attempt in 0..7 {
let retry_duration = match self.try_embed(&text, client).await {
Ok(result) => {
results.push(result);
break;
}
Err(retry) => {
tracing::warn!("Failed: {}", retry.error);
retry.into_duration(attempt)
}
}?;
tracing::warn!(
"Attempt #{}, retrying after {}ms.",
attempt,
retry_duration.as_millis()
);
tokio::time::sleep(retry_duration).await;
}
}
Ok(results)
}
async fn try_embed(
&self,
text: &str,
client: &reqwest::Client,
) -> Result<Embeddings<f32>, Retry> {
let request = OllamaRequest { model: &self.options.embedding_model.name(), prompt: text };
let response = client
.post(get_ollama_path())
.json(&request)
.send()
.await
.map_err(EmbedError::openai_network)
.map_err(Retry::retry_later)?;
let response = Self::check_response(response).await?;
let response: OllamaResponse = response
.json()
.await
.map_err(EmbedError::openai_unexpected)
.map_err(Retry::retry_later)?;
tracing::trace!("response: {:?}", response.embedding);
let embedding = Embeddings::from_single_embedding(response.embedding);
Ok(embedding)
}
pub fn embed_chunks(
&self,
text_chunks: Vec<Vec<String>>,
) -> Result<Vec<Vec<Embeddings<f32>>>, EmbedError> {
let rt = tokio::runtime::Builder::new_current_thread()
.enable_io()
.enable_time()
.build()
.map_err(EmbedError::openai_runtime_init)?;
let client = self.new_client()?;
rt.block_on(futures::future::try_join_all(
text_chunks.into_iter().map(|prompts| self.embed(prompts, &client)),
))
}
// Defaults copied from openai.rs
pub fn chunk_count_hint(&self) -> usize {
10
}
pub fn prompt_count_in_chunk_hint(&self) -> usize {
10
}
pub fn dimensions(&self) -> usize {
self.options.embedding_model.dimensions
}
pub fn distribution(&self) -> Option<DistributionShift> {
None
}
}
impl Display for OllamaError {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
write!(f, "{}", self.error)
}
}
fn get_ollama_path() -> String {
// Important: Hostname not enough, has to be entire path to embeddings endpoint
std::env::var("MEILI_OLLAMA_URL").unwrap_or("http://localhost:11434/api/embeddings".to_string())
}

20
milli/src/vector/openai.rs
View file

@ -419,12 +419,12 @@ impl Embedder {
// retrying in case of failure
pub struct Retry {
pub error: EmbedError,
struct Retry {
error: EmbedError,
strategy: RetryStrategy,
}
pub enum RetryStrategy {
enum RetryStrategy {
GiveUp,
Retry,
RetryTokenized,
@ -432,23 +432,23 @@ pub enum RetryStrategy {
}
impl Retry {
pub fn give_up(error: EmbedError) -> Self {
fn give_up(error: EmbedError) -> Self {
Self { error, strategy: RetryStrategy::GiveUp }
}
pub fn retry_later(error: EmbedError) -> Self {
fn retry_later(error: EmbedError) -> Self {
Self { error, strategy: RetryStrategy::Retry }
}
pub fn retry_tokenized(error: EmbedError) -> Self {
fn retry_tokenized(error: EmbedError) -> Self {
Self { error, strategy: RetryStrategy::RetryTokenized }
}
pub fn rate_limited(error: EmbedError) -> Self {
fn rate_limited(error: EmbedError) -> Self {
Self { error, strategy: RetryStrategy::RetryAfterRateLimit }
}
pub fn into_duration(self, attempt: u32) -> Result<tokio::time::Duration, EmbedError> {
fn into_duration(self, attempt: u32) -> Result<tokio::time::Duration, EmbedError> {
match self.strategy {
RetryStrategy::GiveUp => Err(self.error),
RetryStrategy::Retry => Ok(tokio::time::Duration::from_millis((10u64).pow(attempt))),
@ -459,11 +459,11 @@ impl Retry {
}
}
pub fn must_tokenize(&self) -> bool {
fn must_tokenize(&self) -> bool {
matches!(self.strategy, RetryStrategy::RetryTokenized)
}
pub fn into_error(self) -> EmbedError {
fn into_error(self) -> EmbedError {
self.error
}
}

29
milli/src/vector/settings.rs
View file

@ -1,7 +1,7 @@
use deserr::Deserr;
use serde::{Deserialize, Serialize};
use super::{ollama, openai};
use super::openai;
use crate::prompt::PromptData;
use crate::update::Setting;
use crate::vector::EmbeddingConfig;
@ -80,15 +80,11 @@ impl EmbeddingSettings {
Self::SOURCE => {
&[EmbedderSource::HuggingFace, EmbedderSource::OpenAi, EmbedderSource::UserProvided]
}
Self::MODEL => {
&[EmbedderSource::HuggingFace, EmbedderSource::OpenAi, EmbedderSource::Ollama]
}
Self::MODEL => &[EmbedderSource::HuggingFace, EmbedderSource::OpenAi],
Self::REVISION => &[EmbedderSource::HuggingFace],
Self::API_KEY => &[EmbedderSource::OpenAi],
Self::DIMENSIONS => &[EmbedderSource::OpenAi, EmbedderSource::UserProvided],
Self::DOCUMENT_TEMPLATE => {
&[EmbedderSource::HuggingFace, EmbedderSource::OpenAi, EmbedderSource::Ollama]
}
Self::DOCUMENT_TEMPLATE => &[EmbedderSource::HuggingFace, EmbedderSource::OpenAi],
_other => unreachable!("unknown field"),
}
}
@ -105,7 +101,6 @@ impl EmbeddingSettings {
EmbedderSource::HuggingFace => {
&[Self::SOURCE, Self::MODEL, Self::REVISION, Self::DOCUMENT_TEMPLATE]
}
EmbedderSource::Ollama => &[Self::SOURCE, Self::MODEL, Self::DOCUMENT_TEMPLATE],
EmbedderSource::UserProvided => &[Self::SOURCE, Self::DIMENSIONS],
}
}
@ -139,7 +134,6 @@ pub enum EmbedderSource {
#[default]
OpenAi,
HuggingFace,
Ollama,
UserProvided,
}
@ -149,7 +143,6 @@ impl std::fmt::Display for EmbedderSource {
EmbedderSource::OpenAi => "openAi",
EmbedderSource::HuggingFace => "huggingFace",
EmbedderSource::UserProvided => "userProvided",
EmbedderSource::Ollama => "ollama",
};
f.write_str(s)
}
@ -202,14 +195,6 @@ impl From<EmbeddingConfig> for EmbeddingSettings {
dimensions: options.dimensions.map(Setting::Set).unwrap_or_default(),
document_template: Setting::Set(prompt.template),
},
super::EmbedderOptions::Ollama(options) => Self {
source: Setting::Set(EmbedderSource::Ollama),
model: Setting::Set(options.embedding_model.name().to_owned()),
revision: Setting::NotSet,
api_key: Setting::NotSet,
dimensions: Setting::NotSet,
document_template: Setting::Set(prompt.template),
},
super::EmbedderOptions::UserProvided(options) => Self {
source: Setting::Set(EmbedderSource::UserProvided),
model: Setting::NotSet,
@ -244,14 +229,6 @@ impl From<EmbeddingSettings> for EmbeddingConfig {
}
this.embedder_options = super::EmbedderOptions::OpenAi(options);
}
EmbedderSource::Ollama => {
let mut options: ollama::EmbedderOptions =
super::ollama::EmbedderOptions::with_default_model();
if let Some(model) = model.set() {
options.embedding_model = super::ollama::EmbeddingModel::from_name(&model);
}
this.embedder_options = super::EmbedderOptions::Ollama(options);
}
EmbedderSource::HuggingFace => {
let mut options = super::hf::EmbedderOptions::default();
if let Some(model) = model.set() {