litellm-mirror/litellm/router_utils/pattern_match_deployments.py
Krish Dholakia 136693cac4
LiteLLM Minor Fixes & Improvements (11/05/2024) (#6590)
* fix(pattern_matching_router.py): update model name using correct function

* fix(langfuse.py): metadata deepcopy can cause unhandled error (#6563)

Co-authored-by: seva <seva@inita.com>

* fix(stream_chunk_builder_utils.py): correctly set prompt tokens + log correct streaming usage

Closes https://github.com/BerriAI/litellm/issues/6488

* build(deps): bump cookie and express in /docs/my-website (#6566)

Bumps [cookie](https://github.com/jshttp/cookie) and [express](https://github.com/expressjs/express). These dependencies needed to be updated together.

Updates `cookie` from 0.6.0 to 0.7.1
- [Release notes](https://github.com/jshttp/cookie/releases)
- [Commits](https://github.com/jshttp/cookie/compare/v0.6.0...v0.7.1)

Updates `express` from 4.20.0 to 4.21.1
- [Release notes](https://github.com/expressjs/express/releases)
- [Changelog](https://github.com/expressjs/express/blob/4.21.1/History.md)
- [Commits](https://github.com/expressjs/express/compare/4.20.0...4.21.1)

---
updated-dependencies:
- dependency-name: cookie
  dependency-type: indirect
- dependency-name: express
  dependency-type: indirect
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* docs(virtual_keys.md): update Dockerfile reference (#6554)

Signed-off-by: Emmanuel Ferdman <emmanuelferdman@gmail.com>

* (proxy fix) - call connect on prisma client when running setup (#6534)

* critical fix - call connect on prisma client when running setup

* fix test_proxy_server_prisma_setup

* fix test_proxy_server_prisma_setup

* Add 3.5 haiku (#6588)

* feat: add claude-3-5-haiku-20241022 entries

* feat: add claude-3-5-haiku-20241022 and vertex_ai/claude-3-5-haiku@20241022 models

* add missing entries, remove vision

* remove image token costs

* Litellm perf improvements 3 (#6573)

* perf: move writing key to cache, to background task

* perf(litellm_pre_call_utils.py): add otel tracing for pre-call utils

adds 200ms on calls with pgdb connected

* fix(litellm_pre_call_utils.py'): rename call_type to actual call used

* perf(proxy_server.py): remove db logic from _get_config_from_file

was causing db calls to occur on every llm request, if team_id was set on key

* fix(auth_checks.py): add check for reducing db calls if user/team id does not exist in db

reduces latency/call by ~100ms

* fix(proxy_server.py): minor fix on existing_settings not incl alerting

* fix(exception_mapping_utils.py): map databricks exception string

* fix(auth_checks.py): fix auth check logic

* test: correctly mark flaky test

* fix(utils.py): handle auth token error for tokenizers.from_pretrained

* build: fix map

* build: fix map

* build: fix json for model map

* fix ImageObject conversion (#6584)

* (fix) litellm.text_completion raises a non-blocking error on simple usage (#6546)

* unit test test_huggingface_text_completion_logprobs

* fix return TextCompletionHandler convert_chat_to_text_completion

* fix hf rest api

* fix test_huggingface_text_completion_logprobs

* fix linting errors

* fix importLiteLLMResponseObjectHandler

* fix test for LiteLLMResponseObjectHandler

* fix test text completion

* fix allow using 15 seconds for premium license check

* testing fix bedrock deprecated cohere.command-text-v14

* (feat) add `Predicted Outputs` for OpenAI  (#6594)

* bump openai to openai==1.54.0

* add 'prediction' param

* testing fix bedrock deprecated cohere.command-text-v14

* test test_openai_prediction_param.py

* test_openai_prediction_param_with_caching

* doc Predicted Outputs

* doc Predicted Output

* (fix) Vertex Improve Performance when using `image_url`  (#6593)

* fix transformation vertex

* test test_process_gemini_image

* test_image_completion_request

* testing fix - bedrock has deprecated cohere.command-text-v14

* fix vertex pdf

* bump: version 1.51.5 → 1.52.0

* fix(lowest_tpm_rpm_routing.py): fix parallel rate limit check (#6577)

* fix(lowest_tpm_rpm_routing.py): fix parallel rate limit check

* fix(lowest_tpm_rpm_v2.py): return headers in correct format

* test: update test

* build(deps): bump cookie and express in /docs/my-website (#6566)

Bumps [cookie](https://github.com/jshttp/cookie) and [express](https://github.com/expressjs/express). These dependencies needed to be updated together.

Updates `cookie` from 0.6.0 to 0.7.1
- [Release notes](https://github.com/jshttp/cookie/releases)
- [Commits](https://github.com/jshttp/cookie/compare/v0.6.0...v0.7.1)

Updates `express` from 4.20.0 to 4.21.1
- [Release notes](https://github.com/expressjs/express/releases)
- [Changelog](https://github.com/expressjs/express/blob/4.21.1/History.md)
- [Commits](https://github.com/expressjs/express/compare/4.20.0...4.21.1)

---
updated-dependencies:
- dependency-name: cookie
  dependency-type: indirect
- dependency-name: express
  dependency-type: indirect
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* docs(virtual_keys.md): update Dockerfile reference (#6554)

Signed-off-by: Emmanuel Ferdman <emmanuelferdman@gmail.com>

* (proxy fix) - call connect on prisma client when running setup (#6534)

* critical fix - call connect on prisma client when running setup

* fix test_proxy_server_prisma_setup

* fix test_proxy_server_prisma_setup

* Add 3.5 haiku (#6588)

* feat: add claude-3-5-haiku-20241022 entries

* feat: add claude-3-5-haiku-20241022 and vertex_ai/claude-3-5-haiku@20241022 models

* add missing entries, remove vision

* remove image token costs

* Litellm perf improvements 3 (#6573)

* perf: move writing key to cache, to background task

* perf(litellm_pre_call_utils.py): add otel tracing for pre-call utils

adds 200ms on calls with pgdb connected

* fix(litellm_pre_call_utils.py'): rename call_type to actual call used

* perf(proxy_server.py): remove db logic from _get_config_from_file

was causing db calls to occur on every llm request, if team_id was set on key

* fix(auth_checks.py): add check for reducing db calls if user/team id does not exist in db

reduces latency/call by ~100ms

* fix(proxy_server.py): minor fix on existing_settings not incl alerting

* fix(exception_mapping_utils.py): map databricks exception string

* fix(auth_checks.py): fix auth check logic

* test: correctly mark flaky test

* fix(utils.py): handle auth token error for tokenizers.from_pretrained

* build: fix map

* build: fix map

* build: fix json for model map

* test: remove eol model

* fix(proxy_server.py): fix db config loading logic

* fix(proxy_server.py): fix order of config / db updates, to ensure fields not overwritten

* test: skip test if required env var is missing

* test: fix test

---------

Signed-off-by: dependabot[bot] <support@github.com>
Signed-off-by: Emmanuel Ferdman <emmanuelferdman@gmail.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Emmanuel Ferdman <emmanuelferdman@gmail.com>
Co-authored-by: Ishaan Jaff <ishaanjaffer0324@gmail.com>
Co-authored-by: paul-gauthier <69695708+paul-gauthier@users.noreply.github.com>

* test: mark flaky test

* test: handle anthropic api instability

* test(test_proxy_utils.py): add testing for db config update logic

* Update setuptools in docker and fastapi to latest verison, in order to upgrade starlette version (#6597)

* build(deps): bump cookie and express in /docs/my-website (#6566)

Bumps [cookie](https://github.com/jshttp/cookie) and [express](https://github.com/expressjs/express). These dependencies needed to be updated together.

Updates `cookie` from 0.6.0 to 0.7.1
- [Release notes](https://github.com/jshttp/cookie/releases)
- [Commits](https://github.com/jshttp/cookie/compare/v0.6.0...v0.7.1)

Updates `express` from 4.20.0 to 4.21.1
- [Release notes](https://github.com/expressjs/express/releases)
- [Changelog](https://github.com/expressjs/express/blob/4.21.1/History.md)
- [Commits](https://github.com/expressjs/express/compare/4.20.0...4.21.1)

---
updated-dependencies:
- dependency-name: cookie
  dependency-type: indirect
- dependency-name: express
  dependency-type: indirect
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* docs(virtual_keys.md): update Dockerfile reference (#6554)

Signed-off-by: Emmanuel Ferdman <emmanuelferdman@gmail.com>

* (proxy fix) - call connect on prisma client when running setup (#6534)

* critical fix - call connect on prisma client when running setup

* fix test_proxy_server_prisma_setup

* fix test_proxy_server_prisma_setup

* Add 3.5 haiku (#6588)

* feat: add claude-3-5-haiku-20241022 entries

* feat: add claude-3-5-haiku-20241022 and vertex_ai/claude-3-5-haiku@20241022 models

* add missing entries, remove vision

* remove image token costs

* Litellm perf improvements 3 (#6573)

* perf: move writing key to cache, to background task

* perf(litellm_pre_call_utils.py): add otel tracing for pre-call utils

adds 200ms on calls with pgdb connected

* fix(litellm_pre_call_utils.py'): rename call_type to actual call used

* perf(proxy_server.py): remove db logic from _get_config_from_file

was causing db calls to occur on every llm request, if team_id was set on key

* fix(auth_checks.py): add check for reducing db calls if user/team id does not exist in db

reduces latency/call by ~100ms

* fix(proxy_server.py): minor fix on existing_settings not incl alerting

* fix(exception_mapping_utils.py): map databricks exception string

* fix(auth_checks.py): fix auth check logic

* test: correctly mark flaky test

* fix(utils.py): handle auth token error for tokenizers.from_pretrained

* build: fix map

* build: fix map

* build: fix json for model map

* fix ImageObject conversion (#6584)

* (fix) litellm.text_completion raises a non-blocking error on simple usage (#6546)

* unit test test_huggingface_text_completion_logprobs

* fix return TextCompletionHandler convert_chat_to_text_completion

* fix hf rest api

* fix test_huggingface_text_completion_logprobs

* fix linting errors

* fix importLiteLLMResponseObjectHandler

* fix test for LiteLLMResponseObjectHandler

* fix test text completion

* fix allow using 15 seconds for premium license check

* testing fix bedrock deprecated cohere.command-text-v14

* (feat) add `Predicted Outputs` for OpenAI  (#6594)

* bump openai to openai==1.54.0

* add 'prediction' param

* testing fix bedrock deprecated cohere.command-text-v14

* test test_openai_prediction_param.py

* test_openai_prediction_param_with_caching

* doc Predicted Outputs

* doc Predicted Output

* (fix) Vertex Improve Performance when using `image_url`  (#6593)

* fix transformation vertex

* test test_process_gemini_image

* test_image_completion_request

* testing fix - bedrock has deprecated cohere.command-text-v14

* fix vertex pdf

* bump: version 1.51.5 → 1.52.0

* Update setuptools in docker and fastapi to latest verison, in order to upgrade starlette version

---------

Signed-off-by: dependabot[bot] <support@github.com>
Signed-off-by: Emmanuel Ferdman <emmanuelferdman@gmail.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Emmanuel Ferdman <emmanuelferdman@gmail.com>
Co-authored-by: Ishaan Jaff <ishaanjaffer0324@gmail.com>
Co-authored-by: paul-gauthier <69695708+paul-gauthier@users.noreply.github.com>
Co-authored-by: Krish Dholakia <krrishdholakia@gmail.com>
Co-authored-by: Jacob Hagstedt <wcgs@novonordisk.com>

* fix(langfuse.py): fix linting errors

* fix: fix linting errors

* fix: fix casting error

* fix: fix typing error

* fix: add more tests

* fix(utils.py): fix return_processed_chunk_logic

* Revert "Update setuptools in docker and fastapi to latest verison, in order t…" (#6615)

This reverts commit 1a7f7bdfb7.

* docs fix clarify team_id on team based logging

* doc fix team based logging with langfuse

* fix flake8 checks

* test: bump sleep time

* refactor: replace claude-instant-1.2 with haiku in testing

* fix(proxy_server.py): move to using sl payload in track_cost_callback

* fix(proxy_server.py): fix linting errors

* fix(proxy_server.py): fallback to kwargs(response_cost) if given

* test: remove claude-instant-1 from tests

* test: fix claude test

* docs fix clarify team_id on team based logging

* doc fix team based logging with langfuse

* build: remove lint.yml

---------

Signed-off-by: dependabot[bot] <support@github.com>
Signed-off-by: Emmanuel Ferdman <emmanuelferdman@gmail.com>
Co-authored-by: Vsevolod Karvetskiy <56288164+karvetskiy@users.noreply.github.com>
Co-authored-by: seva <seva@inita.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Emmanuel Ferdman <emmanuelferdman@gmail.com>
Co-authored-by: Ishaan Jaff <ishaanjaffer0324@gmail.com>
Co-authored-by: paul-gauthier <69695708+paul-gauthier@users.noreply.github.com>
Co-authored-by: Jacob Hagstedt P Suorra <Jacobh2@users.noreply.github.com>
Co-authored-by: Jacob Hagstedt <wcgs@novonordisk.com>
2024-11-07 04:17:05 +05:30

200 lines
6.5 KiB
Python

"""
Class to handle llm wildcard routing and regex pattern matching
"""
import copy
import re
from re import Match
from typing import Dict, List, Optional
from litellm import get_llm_provider
from litellm._logging import verbose_router_logger
class PatternMatchRouter:
"""
Class to handle llm wildcard routing and regex pattern matching
doc: https://docs.litellm.ai/docs/proxy/configs#provider-specific-wildcard-routing
This class will store a mapping for regex pattern: List[Deployments]
"""
def __init__(self):
self.patterns: Dict[str, List] = {}
def add_pattern(self, pattern: str, llm_deployment: Dict):
"""
Add a regex pattern and the corresponding llm deployments to the patterns
Args:
pattern: str
llm_deployment: str or List[str]
"""
# Convert the pattern to a regex
regex = self._pattern_to_regex(pattern)
if regex not in self.patterns:
self.patterns[regex] = []
self.patterns[regex].append(llm_deployment)
def _pattern_to_regex(self, pattern: str) -> str:
"""
Convert a wildcard pattern to a regex pattern
example:
pattern: openai/*
regex: openai/.*
pattern: openai/fo::*::static::*
regex: openai/fo::.*::static::.*
Args:
pattern: str
Returns:
str: regex pattern
"""
# # Replace '*' with '.*' for regex matching
# regex = pattern.replace("*", ".*")
# # Escape other special characters
# regex = re.escape(regex).replace(r"\.\*", ".*")
# return f"^{regex}$"
return re.escape(pattern).replace(r"\*", "(.*)")
def _return_pattern_matched_deployments(
self, matched_pattern: Match, deployments: List[Dict]
) -> List[Dict]:
new_deployments = []
for deployment in deployments:
new_deployment = copy.deepcopy(deployment)
new_deployment["litellm_params"]["model"] = (
PatternMatchRouter.set_deployment_model_name(
matched_pattern=matched_pattern,
litellm_deployment_litellm_model=deployment["litellm_params"][
"model"
],
)
)
new_deployments.append(new_deployment)
return new_deployments
def route(self, request: Optional[str]) -> Optional[List[Dict]]:
"""
Route a requested model to the corresponding llm deployments based on the regex pattern
loop through all the patterns and find the matching pattern
if a pattern is found, return the corresponding llm deployments
if no pattern is found, return None
Args:
request: Optional[str]
Returns:
Optional[List[Deployment]]: llm deployments
"""
try:
if request is None:
return None
for pattern, llm_deployments in self.patterns.items():
pattern_match = re.match(pattern, request)
if pattern_match:
return self._return_pattern_matched_deployments(
matched_pattern=pattern_match, deployments=llm_deployments
)
except Exception as e:
verbose_router_logger.debug(f"Error in PatternMatchRouter.route: {str(e)}")
return None # No matching pattern found
@staticmethod
def set_deployment_model_name(
matched_pattern: Match,
litellm_deployment_litellm_model: str,
) -> str:
"""
Set the model name for the matched pattern llm deployment
E.g.:
model_name: llmengine/* (can be any regex pattern or wildcard pattern)
litellm_params:
model: openai/*
if model_name = "llmengine/foo" -> model = "openai/foo"
"""
## BASE CASE: if the deployment model name does not contain a wildcard, return the deployment model name
if "*" not in litellm_deployment_litellm_model:
return litellm_deployment_litellm_model
wildcard_count = litellm_deployment_litellm_model.count("*")
# Extract all dynamic segments from the request
dynamic_segments = matched_pattern.groups()
if len(dynamic_segments) > wildcard_count:
raise ValueError(
f"More wildcards in the deployment model name than the pattern. Wildcard count: {wildcard_count}, dynamic segments count: {len(dynamic_segments)}"
)
# Replace the corresponding wildcards in the litellm model pattern with extracted segments
for segment in dynamic_segments:
litellm_deployment_litellm_model = litellm_deployment_litellm_model.replace(
"*", segment, 1
)
return litellm_deployment_litellm_model
def get_pattern(
self, model: str, custom_llm_provider: Optional[str] = None
) -> Optional[List[Dict]]:
"""
Check if a pattern exists for the given model and custom llm provider
Args:
model: str
custom_llm_provider: Optional[str]
Returns:
bool: True if pattern exists, False otherwise
"""
if custom_llm_provider is None:
try:
(
_,
custom_llm_provider,
_,
_,
) = get_llm_provider(model=model)
except Exception:
# get_llm_provider raises exception when provider is unknown
pass
return self.route(model) or self.route(f"{custom_llm_provider}/{model}")
def get_deployments_by_pattern(
self, model: str, custom_llm_provider: Optional[str] = None
) -> List[Dict]:
"""
Get the deployments by pattern
Args:
model: str
custom_llm_provider: Optional[str]
Returns:
List[Dict]: llm deployments matching the pattern
"""
pattern_match = self.get_pattern(model, custom_llm_provider)
if pattern_match:
return pattern_match
return []
# Example usage:
# router = PatternRouter()
# router.add_pattern('openai/*', [Deployment(), Deployment()])
# router.add_pattern('openai/fo::*::static::*', Deployment())
# print(router.route('openai/gpt-4')) # Output: [Deployment(), Deployment()]
# print(router.route('openai/fo::hi::static::hi')) # Output: [Deployment()]
# print(router.route('something/else')) # Output: None