From 4b3a3d79e03c5203f61de2534a7010955c7dd01f Mon Sep 17 00:00:00 2001
From: Anas Khan <83116240+anxkhn@users.noreply.github.com>
Date: Sun, 5 Jul 2026 08:10:05 +0530
Subject: [PATCH] client: widen sync generate think type to accept string
 levels

The `think` parameter accepts either a bool or the 'low'/'medium'/'high'
string levels, and the request model (GenerateRequest.think) serializes all
of them. PR #553 widened this annotation across Client.chat,
AsyncClient.generate, and AsyncClient.chat, but missed the sync
Client.generate overloads and implementation, which stayed at
Optional[bool].

As a result, ollama.generate(..., think='high') (bound to the sync
Client.generate) was flagged as a false type error by static type checkers,
even though it works at runtime and the equivalent chat() and async
generate() calls type-check cleanly.

Widen the three sync Client.generate occurrences to
Optional[Union[bool, Literal['low', 'medium', 'high']]] to match the other
methods, and add a signature-parity test guarding against future drift.

Signed-off-by: Anas Khan <83116240+anxkhn@users.noreply.github.com>
---
 ollama/_client.py    |  6 +++---
 tests/test_client.py | 11 +++++++++++
 2 files changed, 14 insertions(+), 3 deletions(-)

diff --git a/ollama/_client.py b/ollama/_client.py
index 18cb0fb4..8dfce824 100644
--- a/ollama/_client.py
+++ b/ollama/_client.py
@@ -209,7 +209,7 @@ def generate(
     template: str = '',
     context: Optional[Sequence[int]] = None,
     stream: Literal[False] = False,
-    think: Optional[bool] = None,
+    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
     logprobs: Optional[bool] = None,
     top_logprobs: Optional[int] = None,
     raw: bool = False,
@@ -233,7 +233,7 @@ def generate(
     template: str = '',
     context: Optional[Sequence[int]] = None,
     stream: Literal[True] = True,
-    think: Optional[bool] = None,
+    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
     logprobs: Optional[bool] = None,
     top_logprobs: Optional[int] = None,
     raw: bool = False,
@@ -256,7 +256,7 @@ def generate(
     template: Optional[str] = None,
     context: Optional[Sequence[int]] = None,
     stream: bool = False,
-    think: Optional[bool] = None,
+    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
     logprobs: Optional[bool] = None,
     top_logprobs: Optional[int] = None,
     raw: Optional[bool] = None,
diff --git a/tests/test_client.py b/tests/test_client.py
index 34657513..7b7ab38e 100644
--- a/tests/test_client.py
+++ b/tests/test_client.py
@@ -1,4 +1,5 @@
 import base64
+import inspect
 import json
 import os
 import re
@@ -1486,3 +1487,13 @@ async def test_async_client_context_manager():
     assert not client._client.is_closed
 
   assert client._client.is_closed
+
+
+def test_generate_think_annotation_matches_chat():
+  # The `think` parameter accepts bool or the 'low'/'medium'/'high' string levels.
+  # Client.generate must keep the same annotation as Client.chat and
+  # AsyncClient.generate so passing a string level does not raise a false type
+  # error (regression guard for the sync generate overloads/implementation).
+  expected = inspect.signature(Client.chat).parameters['think'].annotation
+  assert inspect.signature(Client.generate).parameters['think'].annotation == expected
+  assert inspect.signature(AsyncClient.generate).parameters['think'].annotation == expected