4646: Reduce `Transform`'s disk usage r=Kerollmops a=Kerollmops
This PR implements what is described in #4485. It reduces the number of disk writes and disk usage.
Co-authored-by: Clément Renault <clement@meilisearch.com>
4633: Allow to mark vectors as "userProvided" r=Kerollmops a=dureuill
# Pull Request
## Related issue
Fixes#4606
## What does this PR do?
[See usage in PRD](https://meilisearch.notion.site/v1-9-AI-search-changes-e90d6803eca8417aa70a1ac5d0225697#deb96fb0595947bda7d4a371100326eb)
- Extends the shape of the special `_vectors` field in documents.
- previously, the `_vectors` field had to be an object, with each field the name of a configured embedder, and each value either `null`, an embedding (array of numbers), or an array of embeddings.
- In this PR, the value of an embedder in the `_vectors` field can additionally be an object. The object has two fields:
1. `embeddings`: `null`, an embedding (array of numbers), or an array of embeddings.
2. `userProvided`: a boolean indicating if the vector was provided by the user.
- The previous form `embedder_or_array_of_embedders` is semantically equivalent to:
```json
{
"embeddings": embedder_or_array_of_embedders,
"userProvided": true
}
```
- During the indexing step, the subfields and values of the `_vectors` field that have `userProvided` set to **false** are added in the vector DB, but not in the documents DB: that means that future modifications of the documents will trigger a regeneration of that particular vector using the document template.
- This allows **importing** embeddings as a one-shot process, while still retaining the ability to regenerate embeddings on document change.
- The dump process now uses this ability: it enriches the `_vectors` fields of documents with the embeddings that were autogenerated, marking them as not `userProvided`. This allows importing the vectors from a dump without regenerating them.
### Tests
This PR adds the following tests
- Long-needed hybrid search tests of a simple hf embedder
- Dump test that imports vectors. Due to the difficulty of actually importing a dump in tests, we just read the dump and check it contains the expected content.
- Tests in the index-scheduler: this tests that documents containing the same kind of instructions as in the dump indexes as expected
Co-authored-by: Louis Dureuil <louis@meilisearch.com>
4621: Bring back changes from v1.8.0 into main r=curquiza a=curquiza
Co-authored-by: ManyTheFish <many@meilisearch.com>
Co-authored-by: Tamo <tamo@meilisearch.com>
Co-authored-by: meili-bors[bot] <89034592+meili-bors[bot]@users.noreply.github.com>
Co-authored-by: Clément Renault <clement@meilisearch.com>
4580: Update the search logs r=Kerollmops a=irevoire
# Pull Request
## Related issue
Fixes https://github.com/meilisearch/meilisearch/issues/4579
## What does this PR do?
- Update the debug implementation of the search query and search results so it’s way smaller and doesn’t display useless information
Co-authored-by: Tamo <tamo@meilisearch.com>
4535: Support Negative Keywords r=ManyTheFish a=Kerollmops
This PR fixes#4422 by supporting `-` before any word in the query.
The minus symbol `-`, from the ASCII table, is not the only character that can be considered the negative operator. You can see the two other matching characters under the `Based on "-" (U+002D)` section on [this unicode reference website](https://www.compart.com/en/unicode/U+002D).
It's important to notice the strange behavior when a query includes and excludes the same word; only the derivative ( synonyms and split) will be kept:
- If you input `progamer -progamer`, the engine will still search for `pro gamer`.
- If you have the synonym `like = love` and you input `like -like`, it will still search for `love`.
## TODO
- [x] Add analytics
- [x] Add support to the `-` operator
- [x] Make sure to support spaces around `-` well
- [x] Support phrase negation
- [x] Add tests
Co-authored-by: Clément Renault <clement@meilisearch.com>
4536: Limit concurrent search requests r=ManyTheFish a=irevoire
# Pull Request
## Related issue
Fixes https://github.com/meilisearch/meilisearch/issues/4489
## What does this PR do?
- Adds a « search queue » that limits the number of search requests we can process at the same time and stores search requests to be processed
- Process only one search request per core/thread (we use available_parallelism)
- When the search queue is full, new search requests replace old ones **randomly**. The reason is that:
- If we serve the oldest one first, like Typesense, we give the worst performances to everyone
- If we serve the latest one, it gets too easy to DoS us (you just need to fill the queue with as many search requests as we can process simultaneously to ensure no other request will ever be processed)
- By picking the search request randomly, we give a chance to recent search requests to be processed while ensuring that we can't be owned unless they fill our queue entirely and we start returning errors 5xx
- Adds an experimental parameter to control the size of the queue
- Adds a bunch of tests to ensure the search queue works correctly
- Ensure the loop consuming the search queue is running in the health route and crashes if it’s not the case
Co-authored-by: Tamo <tamo@meilisearch.com>
4509: Rest embedder r=ManyTheFish a=dureuill
Fixes#4531
See [Usage page](https://meilisearch.notion.site/v1-8-AI-search-API-usage-135552d6e85a4a52bc7109be82aeca42?pvs=25#e6f58c3b742c4effb4ddc625ce12ee16)
### Implementation changes
- Remove tokio, futures, reqwests
- Add a new `milli::vector::rest::Embedder` embedder
- Update OpenAI and Ollama embedders to use the REST embedder internally
- Make Embedder::embed a sync method
- Add the new embedder source as described in the usage
Co-authored-by: Louis Dureuil <louis@meilisearch.com>
