Commit Graph

2201 Commits

Author SHA1 Message Date
00c4ed3bc2 milli: refactor getting embedder and embedder name 2024-04-04 15:57:29 +02:00
928e6e4c05 Breaking change: remove vector for score details 2024-04-04 15:57:29 +02:00
339a5e3431 Merge #4549
4549: Hugging Face embedder improvements r=dureuill a=dureuill

Architectural changes/Internal improvements

### 1. Prefer safetensors weights over pytorch weights when available

safetensors weights are memory mapped, which reduces memory usage of supported models.

### 2. Update candle

Updates candle to `0.4.1`, now targeting crates.io and the tokenizers to `v0.15.2` (still on github).

This might fix https://github.com/meilisearch/meilisearch/issues/4399 thanks to the now included https://github.com/huggingface/candle/issues/1454

Co-authored-by: Louis Dureuil <louis@meilisearch.com>
2024-04-04 13:47:18 +00:00
5509bafff8 Merge #4535
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>
2024-04-04 13:10:27 +00:00
56bf8503db Merge #4537
4537: Expose distribution shift in settings r=ManyTheFish a=dureuill

See [usage page](https://meilisearch.notion.site/v1-8-AI-search-API-usage-135552d6e85a4a52bc7109be82aeca42#d652adc0890445658aaf36352dbc8802)

# Changes

- Distribution shift added to all embedders.
- Exposed in settings
- Changed the reindexing logic to not trigger a reindex operation when only the distribution shift or API key change

Co-authored-by: Louis Dureuil <louis@meilisearch.com>
2024-04-03 09:08:58 +00:00
a1eccc762a Prefer safetensors to pytorch when both are available 2024-04-03 11:05:59 +02:00
182cb42953 chore: fix some typos in conments
Signed-off-by: redistay <wujunjing@outlook.com>
2024-04-02 19:37:55 +08:00
92a049c2dd Merge #4543
4543: Bring back changes from v1.7.4 into main r=Kerollmops a=dureuill



Co-authored-by: Louis Dureuil <louis@meilisearch.com>
Co-authored-by: meili-bors[bot] <89034592+meili-bors[bot]@users.noreply.github.com>
Co-authored-by: dureuill <dureuill@users.noreply.github.com>
2024-03-28 16:53:51 +00:00
877f4b1045 Support negative phrases 2024-03-28 15:51:43 +01:00
796213af9a Merge branch 'main' into tmp-release-v1.7.4 2024-03-28 10:51:49 +01:00
69f8b2730d Fix the tests 2024-03-28 10:47:04 +01:00
ee8cbea810 Don't optimize reindexing when fields contain dots 2024-03-27 17:04:45 +01:00
572fb3a51d Finer granularity for embedder needs reindex 2024-03-27 12:01:34 +01:00
4ff0255783 remove unused function 2024-03-27 11:51:14 +01:00
a25456120d Expose distribution in settings 2024-03-27 11:51:04 +01:00
168ded3b9d Deserr for distribution 2024-03-27 11:50:33 +01:00
afd1da5642 Add distribution to all embedders 2024-03-27 11:50:22 +01:00
34262c7a0d Add analytics for the negative operator 2024-03-26 18:01:27 +01:00
1da9e0f246 Better support space around the negative operator (-) 2024-03-26 17:47:13 +01:00
e4a3e603b3 Expose a first working version of the negative keyword 2024-03-26 17:47:13 +01:00
817ccc089a also allow api_key 2024-03-25 11:50:00 +01:00
4136630ea5 Use constants instead of raw strings in set_*set() 2024-03-25 11:39:33 +01:00
58972f35cb Allow url parameter for ollama embedder 2024-03-25 11:32:55 +01:00
dfa5e41ea6 Check validity of the URL setting 2024-03-25 11:23:16 +01:00
a1db342f01 Expose REST embedder to the API 2024-03-25 11:23:15 +01:00
f87747f4d3 Remove unwraps 2024-03-25 11:23:04 +01:00
ac52c857e8 Update ollama and openai impls to use the rest embedder internally 2024-03-25 11:23:03 +01:00
8708cbef25 Add RestEmbedder 2024-03-25 11:23:03 +01:00
c3d02f092d OpenAI sync 2024-03-25 11:23:03 +01:00
bc58e8a310 Documentation for the vector module 2024-03-25 11:23:03 +01:00
fc1c3f4a29 Merge #4466
4466: Implements the search cutoff r=irevoire a=irevoire

# Pull Request

## Related issue
Fixes https://github.com/meilisearch/meilisearch/issues/4488

## What does this PR do?
- Adds a cutoff to the bucket sort after 150ms has been spent
- Adds a new setting to customize the default value of 150ms
- When the time is exceeded, we exit early with what we had the time to sort
- If the cutoff has been reached, the search details are updated with a new `Skip` ranking details for the ranking rules that were skipped
- Adds analytics to measure the total number of degraded search requests
- Adds the number of degraded search requests to the Prometheus metrics and Grafana dashboard
- The cutoff **must not** skip the filters; otherwise, we would leak documents to people who don’t have the right to see them


Co-authored-by: Tamo <tamo@meilisearch.com>
Co-authored-by: Louis Dureuil <louis@meilisearch.com>
2024-03-20 13:06:53 +00:00
c5322df519 Revert "Revert "Merge remote-tracking branch 'origin/main' into release-v1.7.1"" 2024-03-20 10:08:28 +01:00
6079141ea6 snapshot the scores side by side with the score details 2024-03-19 18:30:14 +01:00
2c3af8e513 query the detailed score detail in the test 2024-03-19 18:09:02 +01:00
098ab594eb A score of 0.0 is now lesser than a sort result
handles the niche case 🐩 in the hybrid search where:
1. a sort ranking rule is the first rule.
2. the keyword search is skipped at the first rule.
3. the semantic search is not skipped at the first rule.

Previously, we would have the skipped search winning, whereas we want the non skipped one winning.
2024-03-19 17:32:32 +01:00
567194b925 Revert "Merge remote-tracking branch 'origin/main' into release-v1.7.1"
This reverts commit bd74cce86a, reversing
changes made to d2f77e88bd.
2024-03-19 16:56:21 +01:00
d8fe4fe49d return the order in the score details 2024-03-19 15:45:04 +01:00
7b9e0d2944 forward the degraded parameter to the hybrid search 2024-03-19 15:11:21 +01:00
bfec9468d4 Update milli/src/search/mod.rs
Co-authored-by: Louis Dureuil <louis@meilisearch.com>
2024-03-19 14:49:15 +01:00
bd74cce86a Merge remote-tracking branch 'origin/main' into release-v1.7.1 2024-03-19 13:39:17 +01:00
b8cda6c300 fix the search cutoff and add a test 2024-03-19 10:35:47 +01:00
d1db495119 add a settings for the search cutoff 2024-03-19 10:28:23 +01:00
4a467739cd implements a first version of the cutoff without settings 2024-03-19 10:28:21 +01:00
a302e258bd Don't display dimensions as 0 when it is not set 2024-03-18 16:10:12 +01:00
5c95b5c933 chore: remove repetitive words
Signed-off-by: shuangcui <fliter@qq.com>
2024-03-14 21:28:55 +08:00
abd954755d Merge #4476
4476: Make the `/facet-search` route use the `sortFacetValuesBy` setting r=irevoire a=Kerollmops

This PR fixes #4423 by ensuring that the `/facet-search` route uses the `sortFacetValuesBy` setting.

Note for the documentation team (to be moved in the tracking issue): Using the new `sortFacetValuesBy` setting can slow down the facet-search requests as Meilisearch iterates over the whole list of facet values and computes the count of documents on every entry. That is hardly or even impossible to optimize correctly.

### TODO
 - [x] Create a custom HashMap wrapper for the facet `OrderBy` settings.
         This wrapper will return the `OrderBy` setting of the facet, if not defined will use the default `*` one, and if not there either (strange) will fall back on the lexicographic one.
- [x] Create a `ValuesCollection` wrapper that implements the logic for the lexicographic and count order by.
  - [x] Use it when there is no search query.
  - [x] Use it when there is a search query with and without allowed typos.
  - [x] Do not change the original logic, only use a wrapper.
- [x] Add tests

Co-authored-by: Clément Renault <clement@meilisearch.com>
2024-03-13 14:36:14 +00:00
f3fc2bd01f Address some issues with preallocations 2024-03-13 15:22:14 +01:00
e0dac5a22f Simplify the algorithm by using the new facet values collection wrapper 2024-03-13 11:31:34 +01:00
b918b55c6b Introduce a new facet value collection wrapper to simply the usage 2024-03-13 11:31:34 +01:00
306b25ad3a Move the searchForFacetValues struct into a dedicated module 2024-03-13 10:24:21 +01:00