ef4c87accf
Merge pull request #5732 from meilisearch/chat-route-support-metrics
...
Add chat-related metrics on the prometheus route
2025-07-07 08:33:31 +00:00
ced7ea4a5c
Merge pull request #5731 from meilisearch/chat-route-support-dumps
...
Export and import chat completions workspace settings in dumps
2025-07-07 08:31:41 +00:00
fa3990daf9
Format
2025-07-04 13:33:49 +02:00
c5993196b3
Add test
2025-07-04 13:32:55 +02:00
16234e1313
Add fragment swapping test
2025-07-04 13:25:42 +02:00
be9f4f96df
Add experimental feature test
2025-07-04 13:15:15 +02:00
b274106ad3
Add test
2025-07-04 13:05:52 +02:00
48527761e7
Add test
2025-07-04 12:01:15 +02:00
6792d048b8
Test both fragments and document template
2025-07-04 11:47:38 +02:00
07bfed99e6
Expose the host in the analytics
2025-07-04 11:08:02 +02:00
8dfded2993
Update tests
2025-07-04 10:49:03 +02:00
3714f16696
Fix bug
2025-07-04 10:40:50 +02:00
d0cd3cacec
Add a way to reproduce the bug
2025-07-03 18:18:04 +02:00
fef089c7b6
Merge pull request #5596 from meilisearch/request-fragments
...
Request fragments
2025-07-03 15:01:44 +00:00
caccb51814
Add a complex value test
2025-07-03 16:10:23 +02:00
a76a3e8f11
Change the metric name for the search to use a label
2025-07-03 16:01:31 +02:00
32dede35c7
Update snapshots
2025-07-03 15:59:14 +02:00
6397ef12a0
Use three metrics for the three different tokens
2025-07-03 15:56:56 +02:00
cf9b311f71
Format
2025-07-03 15:53:09 +02:00
7423243be0
Add test with multiple embedders
2025-07-03 15:52:18 +02:00
b5e41f0e46
Fix the Mistral uncompatibility with the usage of OpenAI
2025-07-03 15:21:40 +02:00
5690700601
Add fragment addition test
2025-07-03 15:19:31 +02:00
2faad504c6
Add test
2025-07-03 15:12:47 +02:00
2bcd69750f
Add fragment modification test
2025-07-03 15:08:27 +02:00
9f0d33ec99
Expose the number of tokens on the chat completions routes
2025-07-03 15:05:15 +02:00
de24e75be8
Update test
2025-07-03 15:00:11 +02:00
a3af9fe057
new extractor bugfixes:
...
- fix old_has_fragments
- new_is_user_provided is always false when generating fragments,
even if no fragment ever matches
2025-07-03 14:44:34 +02:00
90683d0e4e
add snapshot of get settings
2025-07-03 14:43:06 +02:00
5c79273748
Add TODOs
2025-07-03 14:42:49 +02:00
90e6b6416f
new extractor bugfixes:
...
- fix old_has_fragments
- new_is_user_provided is always false when generating fragments,
even if no fragment ever matches
2025-07-03 14:35:02 +02:00
2b75072b09
Expose the number of internal chat searches on the /metrics route
2025-07-03 14:04:27 +02:00
6e6fd077d4
Ignore unexisting chat completions settings folder
2025-07-03 13:37:38 +02:00
b45eea0d3e
Add test for fragment deletion
2025-07-03 13:26:44 +02:00
a051ab3d9a
Support importing chat completions settings
2025-07-03 12:04:40 +02:00
0b89ef1fd7
Make tests use a shared index
2025-07-03 11:32:49 +02:00
65ba7b47af
Test search fragments
2025-07-03 11:32:49 +02:00
8af76a65bf
Add test_fragment_indexing
2025-07-03 11:32:49 +02:00
6b94033c97
Correctly export the chat completions settings in dumps
2025-07-03 11:30:24 +02:00
dfe0c8664e
Add a version of prompt::Context that has no fields
2025-07-03 11:08:31 +02:00
0ca652de28
Extract vector points: remove the {
2025-07-03 10:52:30 +02:00
87f105747f
Add documentation to Extractor trait
2025-07-03 10:41:20 +02:00
735634e998
Send owned metadata and clear inputs in case of error
2025-07-03 10:32:57 +02:00
3740755d9c
Compare to RawValue::NULL constant rather than explicit "null"
2025-07-03 10:11:07 +02:00
a06cb1bfd6
Remove Embed::process_embeddings and have it be an inherent function of the type that uses it
2025-07-03 10:02:16 +02:00
549dc985b8
Old dump import indexer: fix the case where going from Generated to Generated
2025-07-03 09:58:41 +02:00
428463e45c
Check indexing fragments as well as search fragments
2025-07-02 16:17:22 +02:00
7113fcf63a
New error
2025-07-02 16:17:12 +02:00
aa6855cd4f
Vector settings: don't assume which kind of request is asked when looking at a settings update without fragments
2025-07-02 16:12:23 +02:00
895db76a51
Fix snaps
2025-07-02 16:10:05 +02:00
45da2257ec
Take into account the allowed max memory of the container
...
When Meilisearch runs inside a container (e.g. Docker or Kubernetes) it
may run with less max memory than the available on the host, e.g.
`docker run --memory 1G ...`
Fixes #5616
Signed-off-by: Martin Tzvetanov Grigorov <mgrigorov@apache.org >
2025-07-02 14:23:11 +03:00
