From fc40bd4cdb53d5fca19ef2a27d615e422d71bb57 Mon Sep 17 00:00:00 2001
From: Johannes Foulds <hfoulds@gmail.com>
Date: Tue, 28 Nov 2023 23:22:55 +0200
Subject: [PATCH] AnthropicFunctions  function_call compatibility (#13901)

- **Description:** Updates to `AnthropicFunctions` to be compatible with
the OpenAI `function_call` functionality.
- **Issue:** The functionality to indicate `auto`, `none` and a forced
function_call was not completely implemented in the existing code.
  - **Dependencies:** None
- **Tag maintainer:** @baskaryan , and any of the other maintainers if
needed.
  - **Twitter handle:** None

I have specifically tested this functionality via AWS Bedrock with the
Claude-2 and Claude-Instant models.
---
 .../llms/anthropic_functions.py               | 45 +++++++++++++------
 1 file changed, 32 insertions(+), 13 deletions(-)

diff --git a/libs/experimental/langchain_experimental/llms/anthropic_functions.py b/libs/experimental/langchain_experimental/llms/anthropic_functions.py
index 48e38690dea..0df09ae06d9 100644
--- a/libs/experimental/langchain_experimental/llms/anthropic_functions.py
+++ b/libs/experimental/langchain_experimental/llms/anthropic_functions.py
@@ -144,19 +144,30 @@ class AnthropicFunctions(BaseChatModel):
         forced = False
         function_call = ""
         if "functions" in kwargs:
-            content = prompt.format(tools=json.dumps(kwargs["functions"], indent=2))
-            system = SystemMessage(content=content)
-            messages = [system] + messages
+            # get the function call method
+            if "function_call" in kwargs:
+                function_call = kwargs["function_call"]
+                del kwargs["function_call"]
+            else:
+                function_call = "auto"
+
+            # should function calling be used
+            if function_call != "none":
+                content = prompt.format(tools=json.dumps(kwargs["functions"], indent=2))
+                system = SystemMessage(content=content)
+                messages = [system] + messages
+
+            # is the function call a dictionary (forced function calling)
+            if isinstance(function_call, dict):
+                forced = True
+                function_call_name = function_call["name"]
+                messages.append(AIMessage(content=f"<tool>{function_call_name}</tool>"))
+
             del kwargs["functions"]
             if stop is None:
                 stop = ["</tool_input>"]
             else:
                 stop.append("</tool_input>")
-            if "function_call" in kwargs:
-                forced = True
-                function_call = kwargs["function_call"]["name"]
-                AIMessage(content=f"<tool>{function_call}</tool>")
-                del kwargs["function_call"]
         else:
             if "function_call" in kwargs:
                 raise ValueError(
@@ -168,12 +179,19 @@ class AnthropicFunctions(BaseChatModel):
         completion = response.content
         if forced:
             tag_parser = TagParser()
-            tag_parser.feed(completion.strip() + "</tool_input>")
-            v1 = tag_parser.parse_data["tool_input"][0]
+
+            if "<tool_input>" in completion:
+                tag_parser.feed(completion.strip() + "</tool_input>")
+                v1 = tag_parser.parse_data["tool_input"][0]
+                arguments = json.dumps(_destrip(v1))
+            else:
+                v1 = completion
+                arguments = ""
+
             kwargs = {
                 "function_call": {
-                    "name": function_call,
-                    "arguments": json.dumps(_destrip(v1)),
+                    "name": function_call_name,
+                    "arguments": arguments,
                 }
             }
             message = AIMessage(content="", additional_kwargs=kwargs)
@@ -181,7 +199,7 @@ class AnthropicFunctions(BaseChatModel):
         elif "<tool>" in completion:
             tag_parser = TagParser()
             tag_parser.feed(completion.strip() + "</tool_input>")
-            msg = completion.split("<tool>")[0]
+            msg = completion.split("<tool>")[0].strip()
             v1 = tag_parser.parse_data["tool_input"][0]
             kwargs = {
                 "function_call": {
@@ -192,6 +210,7 @@ class AnthropicFunctions(BaseChatModel):
             message = AIMessage(content=msg, additional_kwargs=kwargs)
             return ChatResult(generations=[ChatGeneration(message=message)])
         else:
+            response.content = response.content.strip()
             return ChatResult(generations=[ChatGeneration(message=response)])
 
     @property