5299: Remote federated search r=dureuill a=dureuill
Fixes#4980
- Usage: https://www.notion.so/meilisearch/API-usage-Remote-search-request-f64fae093abf409e9434c9b9c8fab6f3?pvs=25#1894b06b651f809a9f3dcc6b7189646e
- Changes database format:
- Adds a new database key: the code is resilient to the case where the key is missing
- Adds a new experimental feature: the code for experimental features is resilient to this case
Changes:
- Add experimental feature `proxySearch`
- Add network routes
- Dump support for network
- Add proxy search
- Add various tests
Co-authored-by: Louis Dureuil <louis@meilisearch.com>
5314: Activate used database size r=irevoire a=ManyTheFish
# Pull Request
make the `/stats` route return the `usedDatabaseSize` corresponding to the size used to store the "real" data in the database and not the disk size used by LMDB
Co-authored-by: ManyTheFish <many@meilisearch.com>
5312: Send the OSS analytics once per day instead of once per hour r=ManyTheFish a=irevoire
# Pull Request
## Related issue
Fixes https://github.com/meilisearch/meilisearch/issues/5311
## What does this PR do?
- If the instance is OSS => we send the analytics once every day
- If the instance is on the meilisearch cloud => we send the analytics every hour
Co-authored-by: Tamo <tamo@meilisearch.com>
5288: Improve AI logging r=dureuill a=Kerollmops
This PR fixes#5285 and brings the changes from #5233 to simplify debugging indexation and search performance issues related to AI. The following texts can be found in the logs to debug and understand performance issues:
- `embed_one: search` represents the time we spent waiting for the embedding generation, i.e., OpenAI, local HuggingFace, Ollama.
- `filtered_universe: search::universe` the time spent filtering the documents.
- ~`next_bucket: search::vector_sort` is the time spent finding the nearest neighbors (ANNs) in the vector store (arroy), locally~ was being triggered too many times.
- `indexing::vectors` is the time arroy spends indexing the new vectors for a batch.
- `documents::extract vectors` and `documents::merge vectors` to see the time spent generating and writing the embeddings.
Co-authored-by: Kerollmops <clement@meilisearch.com>
5306: Fix internal error when passing `documentTemplateMaxBytes` to a source that doesn't support it r=ManyTheFish a=dureuill
# Pull Request
## Related issue
Fixes#5305
## What does this PR do?
- add `DOCUMENT_TEMPLATE_MAX_BYTES` to `allowed_sources_for_field` and `allowed_fields_for_source` to prevent a panic
Co-authored-by: Louis Dureuil <louis@meilisearch.com>
5303: Bring back changes from v1.12.8 into v1.13.0 r=Kerollmops a=Kerollmops
Fixes#5087 and other problems that you can find in the original PR #5294.
Co-authored-by: Kerollmops <clement@meilisearch.com>
5235: Introduce a compaction subcommand in meilitool r=dureuill a=Kerollmops
This PR proposes a change to the meilitool helper, introducing the `compact-index` subcommand to reduce the size of the indexes.
While working on this tool, I discovered that the current heed `Env::copy_to_file` API is not very temp file friendly and [could be improved](https://github.com/meilisearch/heed/issues/306).
Co-authored-by: Kerollmops <clement@meilisearch.com>
Co-authored-by: Clément Renault <clement@meilisearch.com>
5140: Fix workload inversion r=dureuill a=ManyTheFish
The used assets were inverted between `workloads/hackernews-modify-facet-numbers.json`
and `workloads/hackernews-modify-facet-strings.json`, now fixed.
Co-authored-by: ManyTheFish <many@meilisearch.com>
5300: Improve unexpected panic message r=irevoire a=irevoire
# Pull Request
## Related issue
Fixes https://github.com/meilisearch/meilisearch/issues/5273
## What does this PR do?
- When an unexpected panic happens in the index-scheduler we catch it and rebuild an error message from the join_error
- Same when the upgrade index-scheduler fails
Co-authored-by: Tamo <tamo@meilisearch.com>
5177: Debug log the channel congestion r=Kerollmops a=Kerollmops
This PR displays the congestion of the BBQueue channel and the allocated memory for the channel and the extraction. This information can be beneficial for debugging and noticing slow disks. We show three pieces of information in debug:
- The direct attempts: the number of tries to send something in the BBQueue channel,
- The blocked attempts: the number of unsuccessful attempts that must be retried,
- The congestion: The percentage of blocking attempts. The higher, the slower the receiver and, therefore, the disk.
Co-authored-by: Kerollmops <clement@meilisearch.com>
Co-authored-by: Clément Renault <clement@meilisearch.com>
