Latest LiteLLM for development and using pytest-recording

.pre-commit-config.yaml

@@ -36,6 +36,7 @@ repos:
         args:
           - --word-list=.secrets.allowlist
           - --exclude-files=.secrets.baseline$
+        exclude: tests/cassettes
   - repo: https://github.com/jumanjihouse/pre-commit-hooks
     rev: 3.0.0
     hooks:
@@ -62,7 +63,7 @@ repos:
         additional_dependencies:
           - "validate-pyproject-schema-store[all]>=2024.08.19" # For Ruff renaming RUF025 to C420
   - repo: https://github.com/astral-sh/uv-pre-commit
-    rev: 0.4.6
+    rev: 0.4.9
     hooks:
       - id: uv-lock
   - repo: https://github.com/pre-commit/mirrors-mypy
@@ -73,7 +74,7 @@ repos:
           - fastapi>=0.109 # Match pyproject.toml
           - fhaviary>=0.6 # Match pyproject.toml
           - httpx
-          - litellm>=1.40.9,<=1.40.12 # Match pyproject.toml
+          - litellm>=1.42.1 # Match pyproject.toml
           - numpy
           - pydantic~=2.0 # Match pyproject.toml
           - tenacity

.secrets.allowlist

@@ -0,0 +1,2 @@
+authorization
+x-api-key

ldp/llms/__init__.py

@@ -3,7 +3,6 @@
     LLMModel,
     LLMResult,
     MultipleCompletionLLMModel,
-    process_llm_config,
     sum_logprobs,
     validate_json_completion,
 )
@@ -35,7 +34,6 @@
     "append_to_sys",
     "prepend_sys",
     "prepend_sys_and_append_sys",
-    "process_llm_config",
     "sum_logprobs",
     "validate_json_completion",
 ]

ldp/llms/chat.py

@@ -59,24 +59,6 @@ def get_supported_openai_params(self) -> list[str] | None:
         return litellm.get_supported_openai_params(self.model)
 
 
-def process_llm_config(llm_config: dict) -> dict:
-    """Remove model_type and try to set max_tokens."""
-    result = llm_config.copy()
-    result.pop("model_type", None)
-
-    if result.get("max_tokens", -1) == -1:  # Either max_tokens is missing or it's -1
-        model = llm_config["model"]
-        # these are estimates - should probably do something better in the future.
-        if model.startswith("gpt-4") or (
-            model.startswith("gpt-3.5") and "0125" in model
-        ):
-            result["max_tokens"] = 4000
-        elif "rrr" not in model:
-            result["max_tokens"] = 2500
-
-    return result
-
-
 def sum_logprobs(choice: litellm.utils.Choices) -> float | None:
     """Calculate the sum of the log probabilities of an LLM completion (a Choices object).
 
@@ -154,22 +136,20 @@ async def achat(
         self, messages: Iterable[Message], **kwargs
     ) -> litellm.ModelResponse:
         return await litellm.acompletion(
-            messages=[m.model_dump(exclude_none=True, by_alias=True) for m in messages],
-            **(process_llm_config(self.config) | kwargs),
+            messages=[m.model_dump(by_alias=True) for m in messages],
+            **(self.config | kwargs),
         )
 
     async def achat_iter(self, messages: Iterable[Message], **kwargs) -> AsyncGenerator:
         return cast(
             AsyncGenerator,
             await litellm.acompletion(
-                messages=[
-                    m.model_dump(exclude_none=True, by_alias=True) for m in messages
-                ],
-                **(process_llm_config(self.config) | kwargs),
+                messages=[m.model_dump(by_alias=True) for m in messages],
                 stream=True,
                 stream_options={
                     "include_usage": True,  # Included to get prompt token counts
                 },
+                **(self.config | kwargs),
             ),
         )
 
@@ -225,7 +205,7 @@ async def call(  # noqa: C901, PLR0915
             chat_kwargs["response_format"] = {"type": "json_object"}
 
         # add static configuration to kwargs
-        chat_kwargs = process_llm_config(self.config) | chat_kwargs
+        chat_kwargs = self.config | chat_kwargs
         n = chat_kwargs.get("n", 1)  # number of completions
         if n < 1:
             raise ValueError("Number of completions (n) must be >= 1.")

pyproject.toml

@@ -74,6 +74,7 @@ tests-root = "tests"
 check-filenames = true
 check-hidden = true
 ignore-words-list = "astroid,ser"
+skip = "tests/cassettes/*"
 
 [tool.mypy]
 # Type-checks the interior of functions without type annotations.
@@ -199,6 +200,8 @@ disable = [
     "too-many-return-statements",  # Rely on ruff PLR0911 for this
     "too-many-statements",  # Rely on ruff PLR0915 for this
     "ungrouped-imports",  # Rely on ruff I001 for this
+    "unidiomatic-typecheck",  # Rely on ruff E721 for this
+    "unreachable",  # Rely on mypy unreachable for this
     "unspecified-encoding",  # Don't care to enforce this
     "unsubscriptable-object",  # Buggy, SEE: https://github.com/PyCQA/pylint/issues/3637
     "unsupported-membership-test",  # Buggy, SEE: https://github.com/pylint-dev/pylint/issues/3045
@@ -400,13 +403,14 @@ dev-dependencies = [
     "fhaviary[xml]",
     "ipython>=8",  # Pin to keep recent
     "ldp[monitor,nn,server,typing,visualization]",
-    "litellm>=1.40.9,<=1.40.12",  # Pin lower for get_supported_openai_params not requiring custom LLM, upper for https://github.com/BerriAI/litellm/issues/4032
+    "litellm>=1.42.1",  # Pin lower for UnsupportedParamsError fix
     "mypy>=1.8",  # Pin for mutable-override
     "pre-commit~=3.4",  # Pin to keep recent
     "pydantic~=2.9",  # Pydantic 2.9 changed JSON schema exports 'allOf', so ensure tests match
     "pylint-pydantic",
     "pylint>=3.2",  # Pin to keep recent
     "pytest-asyncio",
+    "pytest-recording",
     "pytest-rerunfailures",
     "pytest-subtests",
     "pytest-sugar",

tests/__init__.py

@@ -1,3 +1,4 @@
+import pathlib
 from enum import StrEnum
 
 
@@ -6,3 +7,7 @@ class CILLMModelNames(StrEnum):
 
     ANTHROPIC = "claude-3-haiku-20240307"  # Cheap and not Anthropic's cutting edge
     OPENAI = "gpt-4o-mini-2024-07-18"  # Cheap and not OpenAI's cutting edge
+
+
+TESTS_DIR = pathlib.Path(__file__).parent
+CASSETTES_DIR = TESTS_DIR / "cassettes"