From 3d1e7401876cdb568d32283324ef32ef3a036b22 Mon Sep 17 00:00:00 2001
From: "Jonathan C. McKinney" <pseudotensor@gmail.com>
Date: Mon, 23 Oct 2023 01:36:11 -0700
Subject: [PATCH] Work-around langchain's SERP non-thread-safe code for
 sys.stdout =.  Note others

---
 src/gpt4all_llm.py   | 11 +++++++++++
 src/gpt_langchain.py | 14 ++++++--------
 src/serpapi.py       |  8 ++++++++
 3 files changed, 25 insertions(+), 8 deletions(-)

diff --git a/src/gpt4all_llm.py b/src/gpt4all_llm.py
index e214fc55e..dacb36496 100644
--- a/src/gpt4all_llm.py
+++ b/src/gpt4all_llm.py
@@ -132,6 +132,9 @@ def get_llm_gpt4all(model_name,
                                    max_seq_len=max_seq_len,
                                    )
     if model_name == 'llama':
+        # FIXME: streaming not thread safe due to:
+        # llama_cpp/utils.py:        sys.stdout = self.outnull_file
+        # llama_cpp/utils.py:        sys.stdout = self.old_stdout
         cls = H2OLlamaCpp
         if model is None:
             llamacpp_dict = llamacpp_dict.copy()
@@ -159,6 +162,10 @@ def get_llm_gpt4all(model_name,
         llm.client.verbose = verbose
         inner_model = llm.client
     elif model_name == 'gpt4all_llama':
+        # FIXME: streaming not thread safe due to:
+        # gpt4all/pyllmodel.py:        sys.stdout = stream_processor
+        # gpt4all/pyllmodel.py:        sys.stdout = old_stdout
+
         cls = H2OGPT4All
         if model is None:
             llamacpp_dict = llamacpp_dict.copy()
@@ -177,6 +184,10 @@ def get_llm_gpt4all(model_name,
         llm = cls(**model_kwargs)
         inner_model = llm.client
     elif model_name == 'gptj':
+        # FIXME: streaming not thread safe due to:
+        # gpt4all/pyllmodel.py:        sys.stdout = stream_processor
+        # gpt4all/pyllmodel.py:        sys.stdout = old_stdout
+
         cls = H2OGPT4All
         if model is None:
             llamacpp_dict = llamacpp_dict.copy()
diff --git a/src/gpt_langchain.py b/src/gpt_langchain.py
index 54dfde50e..b8759f406 100644
--- a/src/gpt_langchain.py
+++ b/src/gpt_langchain.py
@@ -3675,13 +3675,7 @@ def run_qa_db(**kwargs):
     # only keep actual used
     kwargs = {k: v for k, v in kwargs.items() if k in func_names}
     try:
-        if kwargs.get('verbose', False):
-            # maybe helps avoid sys.stdout getting closed
-            from contextlib import redirect_stdout
-            with redirect_stdout(None):
-                return _run_qa_db(**kwargs)
-        else:
-            return _run_qa_db(**kwargs)
+        return _run_qa_db(**kwargs)
     finally:
         clear_torch_cache()
 
@@ -4518,7 +4512,11 @@ def get_chain(query=None,
             llm, model_name, streamer, prompt_type_out, async_output, only_new_text
 
     if LangChainAgent.PYTHON.value in langchain_agents:
-        if does_support_functiontools(inference_server, model_name):
+        # FIXME: not thread-safe due to sys.stdout = assignments in worker
+        # langchain/utilities/python.py:        sys.stdout = mystdout = StringIO()
+        # langchain/utilities/python.py:            sys.stdout = old_stdout
+        # langchain/utilities/python.py:            sys.stdout = old_stdout
+        if does_support_functiontools(inference_server, model_name) and False:
             chain = create_python_agent(
                 llm=llm,
                 tool=PythonREPLTool(),
diff --git a/src/serpapi.py b/src/serpapi.py
index f7ed7f066..8d5ae49a3 100644
--- a/src/serpapi.py
+++ b/src/serpapi.py
@@ -165,3 +165,11 @@ def __process_response(res: dict, query: str, headsize: int) -> list:
         add_meta(docs, query)
 
         return docs
+
+    def results(self, query: str) -> dict:
+        # Fix non-thread-safe langchain swapping out sys directly.
+        """Run query through SerpAPI and return the raw result."""
+        params = self.get_params(query)
+        search = self.search_engine(params)
+        res = search.get_dict()
+        return res