Skip to content

The agent engineering platform.

fix(langchain): preserve supplied llm in `FlareChain.from_llm` (#32847)

Gal Bloch428c2ee6c56d
diff --git a/libs/langchain/langchain/chains/flare/base.py b/libs/langchain/langchain/chains/flare/base.py
index 207b58b664..d2621b3bfc 100644
--- a/libs/langchain/langchain/chains/flare/base.py
+++ b/libs/langchain/langchain/chains/flare/base.py
@@ -249,7 +249,7 @@ class FlareChain(Chain):
@classmethod
def from_llm(
cls,
- llm: BaseLanguageModel,
+ llm: Optional[BaseLanguageModel],
max_generation_len: int = 32,
**kwargs: Any,
) -> FlareChain:
@@ -272,11 +272,36 @@ class FlareChain(Chain):
"pip install langchain-openai"
)
raise ImportError(msg) from e
- llm = ChatOpenAI(
- max_completion_tokens=max_generation_len,
- logprobs=True,
- temperature=0,
- )
+ # Preserve supplied llm instead of always creating a new ChatOpenAI.
+ # Enforce ChatOpenAI requirement (token logprobs needed for FLARE).
+ if llm is None:
+ llm = ChatOpenAI(
+ max_completion_tokens=max_generation_len,
+ logprobs=True,
+ temperature=0,
+ )
+ else:
+ if not isinstance(llm, ChatOpenAI):
+ msg = (
+ f"FlareChain.from_llm requires ChatOpenAI; got "
+ f"{type(llm).__name__}."
+ )
+ raise TypeError(msg)
+ if not getattr(llm, "logprobs", False): # attribute presence may vary
+ msg = (
+ "Provided ChatOpenAI instance must be constructed with "
+ "logprobs=True for FlareChain."
+ )
+ raise ValueError(msg)
+ current_max = getattr(llm, "max_completion_tokens", None)
+ if current_max is not None and current_max != max_generation_len:
+ logger.debug(
+ "FlareChain.from_llm: supplied llm max_completion_tokens=%s "
+ "differs from requested max_generation_len=%s; "
+ "leaving model unchanged.",
+ current_max,
+ max_generation_len,
+ )
response_chain = PROMPT | llm
question_gen_chain = QUESTION_GENERATOR_PROMPT | llm | StrOutputParser()
return cls(
diff --git a/libs/langchain/tests/unit_tests/chains/test_flare.py b/libs/langchain/tests/unit_tests/chains/test_flare.py
new file mode 100644
index 0000000000..5d7c679de2
--- /dev/null
+++ b/libs/langchain/tests/unit_tests/chains/test_flare.py
@@ -0,0 +1,51 @@
+"""Tests for FlareChain.from_llm preserving supplied ChatOpenAI instance."""
+
+from typing import cast
+
+import pytest
+from langchain_core.documents import Document
+from langchain_core.retrievers import BaseRetriever
+from langchain_core.runnables import RunnableSequence
+
+from langchain.chains.flare.base import FlareChain
+
+
+class _EmptyRetriever(BaseRetriever):
+ """Minimal no-op retriever used only for constructing FlareChain in tests."""
+
+ def _get_relevant_documents(self, query: str) -> list[Document]: # type: ignore[override]
+ del query # mark used
+ return []
+
+ async def _aget_relevant_documents(self, query: str) -> list[Document]: # type: ignore[override]
+ del query # mark used
+ return []
+
+
+def test_from_llm_rejects_non_chatopenai() -> None:
+ class Dummy:
+ pass
+
+ with pytest.raises(TypeError):
+ FlareChain.from_llm(Dummy()) # type: ignore[arg-type]
+
+
[email protected]("langchain_openai")
+def test_from_llm_uses_supplied_chatopenai(monkeypatch: pytest.MonkeyPatch) -> None:
+ try:
+ from langchain_openai import ChatOpenAI
+ except ImportError: # pragma: no cover
+ pytest.skip("langchain-openai not installed")
+
+ # Provide dummy API key to satisfy constructor env validation.
+ monkeypatch.setenv("OPENAI_API_KEY", "TEST")
+
+ supplied = ChatOpenAI(temperature=0.51, logprobs=True, max_completion_tokens=21)
+ chain = FlareChain.from_llm(
+ supplied,
+ max_generation_len=32,
+ retriever=_EmptyRetriever(),
+ )
+
+ llm_in_chain = cast("RunnableSequence", chain.question_generator_chain).steps[1]
+ assert llm_in_chain is supplied