Meng Zhang
0e4a2d2a12
feat: simplify download management, model file should be able to indi… ( #690 )
...
* feat: simplify download management, model file should be able to individually introduced
* fix typo
* update local model support
* update spec back
* update spec
* update
* update
2023-11-02 16:01:04 -07:00
Meng Zhang
36ffeb63f1
refactor: remove useless rust-cxx-cmake-bridge
2023-10-31 17:58:21 -07:00
Meng Zhang
296342efd8
refactor: use llama.cpp tokenizer ( #683 )
...
* refactor: switch to llama.cpp tokenizer to simplify implementation
* refactor: remove tokenizer dependency from tabby
* refactor: renaming decoding to stop condition
* refactor: remove tokenizer dependency
* refactor: remove submodule
* chore: update formatting
* move tokenization to c++
2023-10-31 22:16:09 +00:00
Meng Zhang
f15926f233
chore(ui): update crates/tabby/ui
2023-10-31 08:51:22 -07:00
Meng Zhang
3f1f8bfd30
fix: add /swagger to tabby ui handler
2023-10-30 17:36:22 -07:00
Meng Zhang
73758e207d
feat: improve dashboard UI ( #677 )
2023-10-30 21:47:38 +00:00
Meng Zhang
b4772fbcd0
feat(ui): add dashboard ( #674 )
...
* feat(ui): add dashboard
* handle path
2023-10-30 07:29:50 +00:00
Meng Zhang
89a63dbf33
fix: when send failed, treat the request as stopped ( #673 )
2023-10-30 06:27:09 +00:00
Meng Zhang
de827b1e74
Revert "feat: make --model optional ( #668 )" ( #672 )
...
This reverts commit c55e4481ba .
2023-10-29 21:44:11 -07:00
leiwen83
b47bdd5d77
fix: align with fastchat mainstream ( #670 )
...
Fastchat mainstream change its return format, and text now is only
string in choices structure.
So make this change, to work with mainstream fastchat.
Signed-off-by: Lei Wen <wenlei03@qiyi.com>
Co-authored-by: Lei Wen <wenlei03@qiyi.com>
2023-10-29 19:31:46 -07:00
Meng Zhang
88d2617a34
fix: move events writer to individual thread ( #669 )
2023-10-30 01:31:41 +00:00
Meng Zhang
c55e4481ba
feat: make --model optional ( #668 )
2023-10-30 00:04:42 +00:00
Meng Zhang
7330d75de6
chore: clear cache when there's no active requests
2023-10-29 16:30:30 -07:00
Meng Zhang
2ee5dbfd4f
chore: move tabby-ui under ee license. ( #667 )
...
* chore: introduce tabby-ui EE license.
* update
2023-10-29 15:56:57 -07:00
Meng Zhang
8c680a73fb
feat(ui): add /api page ( #665 )
...
* refactor(tabby-ui): extract tabby-fetcher
* feat(tabby-ui): add /api page
* feat(tabby-ui): add chat model badge
* fix: add components.json for shadcn
* chore: release tabby-ui
2023-10-29 14:55:50 -07:00
Meng Zhang
7bd99d14c0
feat: support continuous batching in llama.cpp backend ( #659 )
...
* refactor: switch back to llama batch interface
* feat: support cont batching
2023-10-28 23:37:05 -07:00
Meng Zhang
14d03b6826
fix(ui): handle invalid semver error
2023-10-28 23:30:35 -07:00
Meng Zhang
8dc5526091
Revert "feat: supports PHP ( #634 )"
...
This reverts commit 688e7d75b5 .
2023-10-28 23:02:10 -07:00
Meng Zhang
43cc5f38cc
feat: do not download ctranslate2 files in downloader
2023-10-28 02:27:09 -07:00
Meng Zhang
444222683a
fix(llama.cpp): bump upstream fix for starcoder model on cuda
2023-10-28 02:03:34 -07:00
Meng Zhang
9309e0314f
fix: fix docker build
2023-10-27 21:25:45 -07:00
Meng Zhang
6dd12ce1ec
fix: adding cuda search path to docker build.
2023-10-27 19:40:35 -07:00
Meng Zhang
2d948639be
fix: docker build for llama cuda backend
2023-10-27 16:36:54 -07:00
Meng Zhang
361091385d
Update CMakeLists.txt
2023-10-27 15:40:52 -07:00
Meng Zhang
23bd542cec
feat: switch cuda backend to llama.cpp ( #656 )
...
* feat: switch cuda backend to llama.cpp
* fix
* fix
2023-10-27 13:41:22 -07:00
Meng Zhang
f37840566b
feat: upgrade llama.cpp ( #645 )
...
* feat: upgrade llama.cpp
* update download files
* update changelog
* Update CHANGELOG.md
* Update CHANGELOG.md
2023-10-27 12:18:46 -07:00
Oleksandr
688e7d75b5
feat: supports PHP ( #634 )
...
* feat: add php language configuration
* feat: add php repository context support
* docs: update documentation to indicate php support
* Update programming-languages.md
* added rev to tree-sitter-php
---------
Co-authored-by: Aleksandr Tarabaka <aleksandr.tarabaka@privatbank.ua>
Co-authored-by: Meng Zhang <meng@tabbyml.com>
2023-10-26 19:27:57 +00:00
Meng Zhang
62054cb4f1
feat: add new version notice in tabby-ui ( #640 )
...
* refactor: extract useHealth hook
* feat: add new version notice
2023-10-26 00:48:18 +00:00
Meng Zhang
9ecbf9031f
refactor: rename tabby-playground to tabby-ui ( #639 )
2023-10-25 23:05:11 +00:00
Meng Zhang
1a4c2aa71f
feat: swtich cpu backend to llama.cpp ( #638 )
...
* feat: swtich Cpu backend to llama.cpp
* feat: switch cpu serving to ggml
* fix cargo.toml
* use optional dependency
* fix compliation
* update ci target
2023-10-25 15:40:11 -07:00
Meng Zhang
21ec60eddf
feat: add server.completion_timeout to control timeout of /v1/completion ( #637 )
...
* feat: add server.completion_timeout to control timeout of /v1/completion
* Update config.rs
2023-10-25 22:05:23 +00:00
Meng Zhang
d6296bb121
refactor: remove useless SwaggerConfig
2023-10-25 13:46:46 -07:00
Meng Zhang
e171776774
Release 0.5.0-dev
...
ctranslate2-bindings@0.5.0-dev
http-api-bindings@0.5.0-dev
llama-cpp-bindings@0.5.0-dev
rust-cxx-cmake-bridge@0.5.0-dev
tabby@0.5.0-dev
tabby-common@0.5.0-dev
tabby-download@0.5.0-dev
tabby-inference@0.5.0-dev
tabby-scheduler@0.5.0-dev
Generated by cargo-workspaces
2023-10-24 13:05:33 -07:00
Meng Zhang
887e50000a
docs: add telemetry information when creating usage id for tracking ( #627 )
2023-10-24 13:02:37 -07:00
Meng Zhang
b4842619a3
feat: encrich home screen info ( #628 )
2023-10-24 13:01:14 -07:00
Meng Zhang
4c6f4b159a
fix: chat completions should use empty stop words ( #626 )
2023-10-24 15:49:11 +00:00
Meng Zhang
2a40d36e20
feat: add main screen to localhost:8080 ( #625 )
...
* feat: add main screen to localhost:8080
* update playground
* feat: use / in playground handler
* update
* fix slack link
* fix next routing
* fix lint
2023-10-24 15:36:08 +00:00
Siwen Yu
44cac0b1c0
chore: add support for building on jetson device ( #620 )
...
* add support for building on jetson devices
* fix wrong file to check
* refactor the environment check of CTranslate2 building
* chore: remove unused code
* chore: format code
2023-10-24 07:31:14 +00:00
Meng Zhang
d3331c591e
fix(TAB-251): local url should be in format of file:///path/to/repo
2023-10-23 18:41:52 -07:00
Meng Zhang
422b27d2ed
feat: support local directory for Repository.git_url ( #622 )
2023-10-24 01:29:38 +00:00
Meng Zhang
8fca850037
feat(eval): add debug_options.raw_prompt to tabby api for evaluation purpose ( #605 )
...
* update eval
* feat: re-purpose as raw input to LLM
* move prompt to Debug options
* Update crates/tabby/src/serve/completions.rs
2023-10-21 20:47:44 +00:00
Meng Zhang
049ebdf9a9
fix: use ruby tree sitter tag query
2023-10-21 13:01:18 -07:00
Chromium
da6109c5af
feat: Add Ruby Language Support ( #597 )
...
* feat: add Ruby Support
* fix: address the comments
* Update dataset.rs
---------
Co-authored-by: Meng Zhang <meng@tabbyml.com>
2023-10-20 18:48:37 +00:00
Meng Zhang
5a822c03b6
fix: cap parallelisim to 4 for cuda to avoid oom ( #601 )
2023-10-20 07:54:53 +00:00
Meng Zhang
fc63dd7458
chore: set health heartbit to every 30 minutes ( #591 )
2023-10-18 15:59:16 -07:00
Meng Zhang
f9eb052659
feat: dedup against existing snippets for retrieval augmented code ( #582 )
...
completion.
2023-10-17 23:10:01 +00:00
Meng Zhang
99d1bf34bb
fix: add flag to disable retrieval augmented code completion and set … ( #580 )
...
* fix: add flag to disable retrieval augmented code completion and set it in golden test
* update
2023-10-17 22:26:23 +00:00
Meng Zhang
981133d6c8
fix: remove empty tokens for similarity calculation ( #581 )
2023-10-17 22:25:38 +00:00
Meng Zhang
2060d47a95
fix: trim the sanitized text to skip empty queries. ( #573 )
2023-10-16 17:45:49 -07:00
Meng Zhang
1503ef6aba
test: add golden test for tabby server ( #566 )
...
* test: add golden test for tabby server
* fix
2023-10-16 08:06:37 +00:00
Meng Zhang
99a7053b6f
refactor: extract language configuration into individual toml file ( #564 )
...
* refactor: extract language configuration into individual toml file
* feat: add golang language configuration (#565 )
2023-10-16 00:24:44 +00:00
Meng Zhang
2e8b12951d
fix(prompt): revert back the behavior, enabling fill-in-the-middle if suffix is empty (use an linebreak as suffix) ( #560 )
2023-10-14 11:42:18 -07:00
Meng Zhang
d27c09d75d
fix: ignore NotReady error for IndexServer ( #559 )
2023-10-14 02:21:17 -07:00
Meng Zhang
b9df0eb260
feat: add golang repository context support ( #555 )
2023-10-14 07:57:54 +00:00
Meng Zhang
9dc5acb62e
feat: add golang's top-level keywords to golang's stop words. ( #554 )
2023-10-14 07:42:26 +00:00
Meng Zhang
82e893d569
Release 0.4.0-dev
...
ctranslate2-bindings@0.4.0-dev
http-api-bindings@0.4.0-dev
llama-cpp-bindings@0.4.0-dev
rust-cxx-cmake-bridge@0.4.0-dev
tabby@0.4.0-dev
tabby-common@0.4.0-dev
tabby-download@0.4.0-dev
tabby-inference@0.4.0-dev
tabby-scheduler@0.4.0-dev
Generated by cargo-workspaces
2023-10-13 17:54:14 -07:00
Meng Zhang
4dbaf4f312
Release 0.3.0
...
ctranslate2-bindings@0.3.0
http-api-bindings@0.3.0
llama-cpp-bindings@0.3.0
rust-cxx-cmake-bridge@0.3.0
tabby@0.3.0
tabby-common@0.3.0
tabby-download@0.3.0
tabby-inference@0.3.0
tabby-scheduler@0.3.0
Generated by cargo-workspaces
2023-10-13 17:45:07 -07:00
Meng Zhang
ac73cbb6f9
fix: revisit language stop words support
2023-10-13 13:40:55 -07:00
Meng Zhang
eb463ba496
Release 0.3.0-rc.1
...
ctranslate2-bindings@0.3.0-rc.1
http-api-bindings@0.3.0-rc.1
llama-cpp-bindings@0.3.0-rc.1
rust-cxx-cmake-bridge@0.3.0-rc.1
tabby@0.3.0-rc.1
tabby-common@0.3.0-rc.1
tabby-download@0.3.0-rc.1
tabby-inference@0.3.0-rc.1
tabby-scheduler@0.3.0-rc.1
Generated by cargo-workspaces
2023-10-13 11:43:34 -07:00
Meng Zhang
f9b9a1a174
Merge remote-tracking branch 'origin/main' into r0.3
2023-10-13 11:39:51 -07:00
Meng Zhang
f8be71d27d
refactor: improve swagger docs ( #549 )
...
* refactor: remove server url in swagger config (as we set "/" for default
server path)
* refactor: polish field naming
* refactor: remove useless modifier for ApiDoc
2023-10-13 18:38:05 +00:00
Meng Zhang
182aceed41
Release 0.3.0-rc.0
...
ctranslate2-bindings@0.3.0-rc.0
http-api-bindings@0.3.0-rc.0
llama-cpp-bindings@0.3.0-rc.0
tabby@0.3.0-rc.0
tabby-common@0.3.0-rc.0
tabby-download@0.3.0-rc.0
tabby-inference@0.3.0-rc.0
tabby-scheduler@0.3.0-rc.0
Generated by cargo-workspaces
2023-10-13 11:24:36 -07:00
Meng Zhang
1a87c99488
feat: add debug flag disable_prompt_rewrite ( #545 )
2023-10-13 06:55:41 +00:00
Meng Zhang
1ad871e1ff
feat: add debug request / response to visualize prompting with source code index ( #544 )
...
* feat: logs segments in completion log
* feat: tune prompt format and improve testing
* add debug options for easier of visualizing the prompt
* update
2023-10-12 19:27:52 -07:00
Meng Zhang
888fbca7e1
feat: support indexing typescript / javascript / tsx / jsx ( #539 )
2023-10-12 00:42:58 +00:00
Meng Zhang
92c1f5a8c0
feat: turn on index server if avaiable ( #536 )
2023-10-11 23:27:41 +00:00
Meng Zhang
1f9e248dfa
refactor: support multiple page in playground ( #537 )
2023-10-11 23:27:20 +00:00
Meng Zhang
99c39375fd
feat: set default empty screen questions for tabby-playground ( #535 )
2023-10-11 22:31:35 +00:00
Meng Zhang
75d2944fb6
feat: support loading the source code index whenever it's ready in file system ( #530 )
...
* feat: support loading index whenever it's ready
* fix test
2023-10-10 21:35:20 -07:00
Meng Zhang
6dbb712918
Release 0.3.0-dev
...
ctranslate2-bindings@0.3.0-dev
http-api-bindings@0.3.0-dev
llama-cpp-bindings@0.3.0-dev
tabby@0.3.0-dev
tabby-common@0.3.0-dev
tabby-download@0.3.0-dev
tabby-inference@0.3.0-dev
tabby-scheduler@0.3.0-dev
Generated by cargo-workspaces
2023-10-09 19:39:27 -07:00
Meng Zhang
d7a8f50de8
fix: use spawn_blocking for sub task without await
2023-10-09 19:38:28 -07:00
Meng Zhang
d21a4de79c
chore: set max timeout for /v1/completions handler ( #526 )
...
* chore: set max timeout for /v1/completions handler
* refactor: extract sub routers
* fix
2023-10-09 18:44:55 -07:00
Meng Zhang
24eadf0de8
refactor: make /v1/health accept GET requests ( #527 )
2023-10-09 18:34:56 +00:00
Meng Zhang
3eb5f4132c
chore: add 'v*' match pattern to restrict git describe only compare against a actual release
2023-10-09 11:13:37 -07:00
Meng Zhang
0f8ee7f589
refactor: move http engine creation to its sub crates ( #524 )
2023-10-09 17:37:04 +00:00
Meng Zhang
8c09f75360
refactor: extract language related data into languages.rs ( #518 )
...
* refactor: extract language related data into languages.rs
* fix
* cleanup index
* fix
* further sanitize
* add a score threshold
2023-10-07 01:40:21 +00:00
Meng Zhang
d85a7892d1
feat: connect prompt rewriting part ( #517 )
...
* feat: enable /v1beta/search if index is available
* make prompt rewriting work
* update
* fix test
* fix api doc
2023-10-07 00:29:24 +00:00
Meng Zhang
8497fb1372
feat: implement /v1beta/search interface ( #516 )
...
* feat: implement /v1beta/search interface
* update
* update
* improve debugger
2023-10-06 18:54:12 +00:00
Meng Zhang
fd2a1ab865
fix: switch back to regex based implementation for stop words ( #513 )
2023-10-06 09:04:37 +00:00
Meng Zhang
4c00ac06fb
fix(download): mark ggml model downloading should be optional, as ggml is only used for metal backend for now ( #512 )
2023-10-05 16:54:56 +00:00
Meng Zhang
9cd2accbaa
feat: adjust code indexing logic ( #510 )
2023-10-05 05:29:41 +00:00
Meng Zhang
1babc38902
feat: add datset viewer, remove treesitter languages that hasn't been verified ( #509 )
...
* refactor: remove not verified tree sitter queries
* feat(experimental): add dataset viewer
update
2023-10-04 21:49:42 -07:00
Meng Zhang
55f68d4224
test: unit test for indexing job ( #508 )
...
* test: unit test for indexing job
* update
* reduce test fixture length
2023-10-04 23:27:19 +00:00
Meng Zhang
f7ebce2514
refactor: deprecate --compute-type ( #505 )
2023-10-04 18:45:34 +00:00
Meng Zhang
8a03c9bf17
refactor: use / as server url
...
Swagger access to the page, regardless of host / port, will just work.
2023-10-03 18:33:56 -07:00
Meng Zhang
2d5b3e4ff5
chore: release v0.2.1
2023-10-03 17:13:39 -07:00
Meng Zhang
503c44e7c5
fix: playground environment misconfig
2023-10-03 17:10:02 -07:00
Meng Zhang
b3b498624c
feat: deprecate num_replicas_per_thread, generate default value for it
2023-10-03 17:02:37 -07:00
Meng Zhang
1afba47059
feat: allow set num_replicas_per_device for CUDA to increase throughput
2023-10-03 15:52:25 -07:00
Meng Zhang
ceaa7ab012
chore: update main branch to v0.3.0-dev
2023-10-03 13:38:27 -07:00
Meng Zhang
1731c3075e
chore: Update version to 0.2.0
2023-10-03 13:32:21 -07:00
Meng Zhang
0e5128e8fb
feat: add chat_template field in tabby.json
2023-10-03 11:46:05 -07:00
Meng Zhang
7fc76228f7
chore: add debug log for /chat interface
2023-10-03 11:38:58 -07:00
Meng Zhang
692c2fe0fd
Release 0.2.0-rc.0
...
ctranslate2-bindings@0.2.0-rc.0
http-api-bindings@0.2.0-rc.0
llama-cpp-bindings@0.2.0-rc.0
tabby@0.2.0-rc.0
tabby-common@0.2.0-rc.0
tabby-download@0.2.0-rc.0
tabby-inference@0.2.0-rc.0
tabby-scheduler@0.2.0-rc.0
Generated by cargo-workspaces
2023-10-02 19:14:12 -07:00
Meng Zhang
6306bb3f01
fix: if local file doens't exist, local_cache_key should be cleared ( #501 )
...
* fix: if local file doens't exist, local_cache_key should be cleared
* fix
2023-10-02 23:48:35 +00:00
Meng Zhang
ce20bd6154
refactor: use RegexSet for cleaer stop regex construction ( #499 )
...
* fix: add a regression test cased for stop words regex matching
* refactor: use RegexSet for cleaer stop regex construction
2023-10-02 23:21:51 +00:00
Meng Zhang
63612d5a67
fix(tabby-download): even when prefer_local_file is set to true, we should still check for remote (if network is avaialble), to see if a file should be upgraded ( #500 )
2023-10-02 23:09:57 +00:00
Meng Zhang
80a17aea37
feat: only show /v1/chat api if --chat-model is set ( #498 )
2023-10-02 17:17:27 +00:00
Meng Zhang
f05dd3a2f6
refactor: cleanup chat api make it message oriented ( #497 )
...
* refactor: refactor into /chat/completions api
* Revert "feat: support request level stop words (#492 )"
This reverts commit 0d6840e372 .
* feat: adjust interface
* switch interface in tabby-playground
* move to chat/prompt, add unit test
* update interface
2023-10-02 15:39:15 +00:00
Meng Zhang
dfdd0373a6
fix: when llama model loads failed, panic in rust stack
2023-10-01 22:25:25 -07:00