Ishaan Jaff
|
41ab5f2f56
|
fix - updating router settings from DB
|
2024-04-23 12:07:58 -07:00 |
|
Krrish Dholakia
|
9d2726c2ac
|
fix(proxy_server.py): handle router being initialized without a model list
|
2024-04-23 10:52:28 -07:00 |
|
Krrish Dholakia
|
a520e1bd6f
|
fix(router.py): add random shuffle and tpm-based shuffle for async shuffle logic
|
2024-04-22 12:58:59 -07:00 |
|
Krrish Dholakia
|
b96741e4f4
|
fix(router.py): async simple-shuffle support
|
2024-04-20 15:01:12 -07:00 |
|
Krrish Dholakia
|
c96ca1f85e
|
fix(router.py): improve debug logsd
|
2024-04-20 13:12:12 -07:00 |
|
Krrish Dholakia
|
0f69f0b44e
|
test(test_router_max_parallel_requests.py): more extensive testing for setting max parallel requests
|
2024-04-20 12:56:54 -07:00 |
|
Krrish Dholakia
|
7aa737cf10
|
fix(router.py): add if router caching setup on info logs
|
2024-04-20 12:34:09 -07:00 |
|
Krrish Dholakia
|
47e9d5f2ec
|
fix(router.py): fix init line for self.default_max_parallel_requests
|
2024-04-20 12:08:21 -07:00 |
|
Krrish Dholakia
|
4c78f8f309
|
fix(router.py): calculate max_parallel_requests from given tpm limits
use the azure formula to calculate rpm -> max_parallel_requests based on a deployment's tpm limits
|
2024-04-20 10:43:18 -07:00 |
|
Krish Dholakia
|
f1340b52dc
|
Merge pull request #3153 from BerriAI/litellm_usage_based_routing_v2_improvements
usage based routing v2 improvements - unit testing + *NEW* async + sync 'pre_call_checks'
|
2024-04-18 22:16:16 -07:00 |
|
Krrish Dholakia
|
9c42c847a5
|
fix(router.py): instrument pre-call-checks for all openai endpoints
|
2024-04-18 21:54:25 -07:00 |
|
Krrish Dholakia
|
81573b2dd9
|
fix(test_lowest_tpm_rpm_routing_v2.py): unit testing for usage-based-routing-v2
|
2024-04-18 21:38:00 -07:00 |
|
Ishaan Jaff
|
67d356b933
|
fix - show api base on hanging requests
|
2024-04-18 20:58:02 -07:00 |
|
Krrish Dholakia
|
2ffd057042
|
test(test_models.py): ensure only admin can call /health
|
2024-04-16 18:13:40 -07:00 |
|
Ishaan Jaff
|
caec0a3938
|
fix - make router set_settings non blocking
|
2024-04-16 18:04:21 -07:00 |
|
Ishaan Jaff
|
7e99854d05
|
Merge pull request #3079 from BerriAI/litellm_router_save_settings_ui
UI - Save / Edit router settings UI
|
2024-04-16 16:57:42 -07:00 |
|
Ishaan Jaff
|
59b154f152
|
feat - update router settings on Admin UI
|
2024-04-16 15:36:26 -07:00 |
|
Krrish Dholakia
|
13cd252f3e
|
fix(proxy_server.py): ensure id used in delete deployment matches id used in litellm Router
|
2024-04-16 15:17:18 -07:00 |
|
Ishaan Jaff
|
e271ce8030
|
router - get settings
|
2024-04-16 14:22:54 -07:00 |
|
Krrish Dholakia
|
2d4fe072ad
|
fix(proxy_server.py): fix delete models endpoint
https://github.com/BerriAI/litellm/issues/2951
|
2024-04-15 18:34:58 -07:00 |
|
Krrish Dholakia
|
e4bcc51e44
|
build(ui): add vertex ai models via ui
|
2024-04-15 15:59:36 -07:00 |
|
Krish Dholakia
|
0d2a75d301
|
Merge pull request #2981 from grav/grav/default_model_name_to_none
Default model_name to None in _aembedding
|
2024-04-15 14:45:01 -07:00 |
|
Krrish Dholakia
|
43c37c31ea
|
fix(proxy_server.py): return none if no model list set in router
https://github.com/BerriAI/litellm/issues/2979
|
2024-04-15 09:02:18 -07:00 |
|
Krrish Dholakia
|
ea1574c160
|
test(test_openai_endpoints.py): add concurrency testing for user defined rate limits on proxy
|
2024-04-12 18:56:13 -07:00 |
|
Krrish Dholakia
|
c03b0bbb24
|
fix(router.py): support pre_call_rpm_check for lowest_tpm_rpm_v2 routing
have routing strategies expose an ‘update rpm’ function; for checking + updating rpm pre call
|
2024-04-12 18:25:14 -07:00 |
|
Krrish Dholakia
|
2267aeb803
|
fix(router.py): create a semaphore for each deployment with rpm
run semaphore logic for each deployment with rpm
|
2024-04-12 18:03:23 -07:00 |
|
Krrish Dholakia
|
a4e415b23c
|
fix(router.py): initial commit for semaphores on router
|
2024-04-12 17:59:05 -07:00 |
|
Mikkel Gravgaard
|
759414004a
|
Default model_name to None in _aembedding
|
2024-04-12 11:33:03 +02:00 |
|
Ishaan Jaff
|
921f6d307e
|
fix - stop printing api_key in debug mode
|
2024-04-11 15:05:22 -07:00 |
|
Krrish Dholakia
|
84d43484c6
|
fix(router.py): make sure pre call rpm check runs even when model not in model cost map
|
2024-04-11 09:27:46 -07:00 |
|
Krrish Dholakia
|
266dba65e7
|
fix(router.py): handle 1 deployment being picked
|
2024-04-10 18:32:54 -07:00 |
|
Krrish Dholakia
|
52462e8bac
|
fix(router.py): move specific deployment check outside common functions
|
2024-04-10 18:06:31 -07:00 |
|
Krrish Dholakia
|
37ac17aebd
|
fix(router.py): fix datetime object
|
2024-04-10 17:55:24 -07:00 |
|
Krrish Dholakia
|
2531701a2a
|
fix(router.py): make get_cooldown_deployment logic async
|
2024-04-10 16:57:01 -07:00 |
|
Krrish Dholakia
|
a47a719caa
|
fix(router.py): generate consistent model id's
having the same id for a deployment, lets redis usage caching work across multiple instances
|
2024-04-10 15:23:57 -07:00 |
|
Krrish Dholakia
|
180cf9bd5c
|
feat(lowest_tpm_rpm_v2.py): move to using redis.incr and redis.mget for getting model usage from redis
makes routing work across multiple instances
|
2024-04-10 14:56:23 -07:00 |
|
Krrish Dholakia
|
b2741933dc
|
fix(proxy_cli.py): don't double load the router config
was causing callbacks to be instantiated twice - double couting usage in cache
|
2024-04-10 13:23:56 -07:00 |
|
Krish Dholakia
|
83f608dc5d
|
Merge pull request #2880 from BerriAI/litellm_api_base_alerting
feat(proxy/utils.py): return api base for request hanging alerts
|
2024-04-06 19:17:18 -07:00 |
|
Krrish Dholakia
|
460546956d
|
fix(utils.py): fix import
|
2024-04-06 18:37:38 -07:00 |
|
Krrish Dholakia
|
6f94f3d127
|
fix(router.py): improve pre-call check -> get model group cache one-time
|
2024-04-06 18:24:51 -07:00 |
|
Krrish Dholakia
|
7ae6432f94
|
fix(router.py): check usage based routing cache in pre-call check
allows pre-call rpm check to work across instances
|
2024-04-06 18:19:02 -07:00 |
|
Krrish Dholakia
|
205ac1496a
|
fix(router.py): store in-memory deployment request count for 60s only
|
2024-04-06 17:53:39 -07:00 |
|
Krrish Dholakia
|
0d1cca9aa0
|
fix(router.py): make router async calls coroutine safe
uses pre-call checks to check if a call is below it's rpm limit, works even if multiple async calls are
made simultaneously
|
2024-04-06 17:31:26 -07:00 |
|
Krrish Dholakia
|
6110d32b1c
|
feat(proxy/utils.py): return api base for request hanging alerts
|
2024-04-06 15:58:53 -07:00 |
|
Ishaan Jaff
|
a6bc673ffa
|
feat - re-use OpenAI client for azure text
|
2024-04-06 12:23:58 -07:00 |
|
Ishaan Jaff
|
01fef1a9f8
|
feat - re-use openai client for text completion
|
2024-04-06 11:28:25 -07:00 |
|
Ishaan Jaff
|
faa0d38087
|
Merge pull request #2868 from BerriAI/litellm_add_command_r_on_proxy
Add Azure Command-r-plus on litellm proxy
|
2024-04-05 15:13:47 -07:00 |
|
Ishaan Jaff
|
9055a071e6
|
proxy - add azure/command r
|
2024-04-05 14:35:31 -07:00 |
|
Krrish Dholakia
|
e3c2bdef4d
|
feat(ui): add models via ui
adds ability to add models via ui to the proxy. also fixes additional bugs around new /model/new endpoint
|
2024-04-04 18:56:20 -07:00 |
|
Krrish Dholakia
|
2236f283fe
|
fix(router.py): handle id being passed in as int
|
2024-04-04 14:23:10 -07:00 |
|