src/reviewllama/cli.py

@@ -69,24 +69,27 @@ Examples:
         default=(
             "You are a PR review assistant in charge of software quality control. "
             "You analyze code changes in the context of the full code base to verify style, "
-            "syntax, and functionality. You respond with suggestions to improve code quality "
-            "You only provide sugestions when you find a flaw in the code otherwise you say that "
-            "no issues were found. Each suggestion should reference the old code and the new "
-            "suggested code."
-            "Do not provide an analysis of the code and do not summarize suggestions. "
-            "Answer as briefly as possible and return only the suggestions in the requested format "
-            "with bullet points and no extra text. Provide examples when appropriate."
+            "syntax, and functionality. Your response should contain exactly 3 suggestions. "
+            "Each suggestion must be in the following format:\n"
+            "```diff\n"
+            "-<old code>\n"
+            "+<new code>\n"
+            "```\n"
+            "Reason: <explanation>\n\n"
+            "Here are two examples of the required format:\n"
+            "```diff\n"
+            "-somvr = 2 + 2\n"
+            "+somevar = 2 + 2\n"
+            "```\n"
+            "Reason: somvr is likely a typo, try replacing with somevar\n\n"
+            "```diff\n"
+            "-add_two_numbers(\"1\", \"2\")\n"
+            "+add_two_numbers(1,2)\n"
+            "```\n"
+            "Reason: add_two_numbers requires numeric values and does not accept strings"
         ),
         help="Base branch to compare against (default: %(default)s)",
     )
-
-    parser.add_argument(
-        "-v",
-        "--verbose",
-        action="store_true",
-        default=False,
-        help="Enable verbose output, including messages sent to the LLM.",
-    )
     return parser
 
 

src/reviewllama/configs.py

@@ -13,7 +13,7 @@ class OllamaConfig:
     base_url: str
     system_prompt: str
     # TODO: Update this to be a passed in value
-    temperature: float = field(default=0.0)
+    temperature: float = field(default=0.7)
 
 
 @dataclass(frozen=True)
@@ -23,14 +23,13 @@ class ReviewConfig:
     paths: List[Path]
     ollama: OllamaConfig
     base_branch: str
-    verbose: bool
 
 
 def create_ollama_config(
     model: str,
     server_url: str,
     system_prompt: str,
-    temperature=0.0,
+    temperature=0.7,
     embedding_model="nomic-embed-text",
 ) -> OllamaConfig:
     """Create OllamaConfig with validated parameters."""
@@ -44,24 +43,19 @@ def create_ollama_config(
 
 
 def create_review_config(
-    paths: List[Path], ollama_config: OllamaConfig, base_branch: str, verbose
+    paths: List[Path], ollama_config: OllamaConfig, base_branch: str
 ) -> ReviewConfig:
     """Create complete ReviewConfig from validated components."""
-    return ReviewConfig(
-        paths=paths, ollama=ollama_config, base_branch=base_branch, verbose=verbose
-    )
+    return ReviewConfig(paths=paths, ollama=ollama_config, base_branch=base_branch)
 
 
-def namespace_to_config(namespace: argparse.Namespace):
+def namespace_to_config(
+        namespace: argparse.Namespace
+):
     """Transform argparse namespace into ReviewConfig."""
     paths = [Path(path_str) for path_str in namespace.paths]
     ollama_config = OllamaConfig(
-        chat_model=namespace.model,
-        base_url=namespace.server_url,
-        system_prompt=namespace.system_prompt,
-        embedding_model=namespace.embedding_model,
+        chat_model=namespace.model, base_url=namespace.server_url, system_prompt=namespace.system_prompt, embedding_model=namespace.embedding_model
     )
 
-    return create_review_config(
-        paths, ollama_config, namespace.base_branch, namespace.verbose
-    )
+    return create_review_config(paths, ollama_config, namespace.base_branch)

src/reviewllama/llm.py

@@ -1,8 +1,9 @@
 from dataclasses import dataclass
 from typing import Any
 
-from langchain_core.messages import BaseMessage
-from langchain_core.prompts import ChatPromptTemplate
+from langchain.memory import ConversationBufferMemory
+from langchain.prompts import ChatPromptTemplate, MessagesPlaceholder
+from langchain.schema import BaseMessage
 from langchain_core.runnables import RunnableLambda
 from langchain_core.runnables.base import RunnableSerializable
 from langchain_core.runnables.passthrough import RunnablePassthrough
@@ -16,13 +17,20 @@ from .configs import OllamaConfig
 
 @dataclass(frozen=True)
 class ChatClient:
-    chain: RunnableSerializable
+    chain: RunnableSerializable[dict[str, Any], BaseMessage]
+    memory: ConversationBufferMemory
+
+    def get_last_response_or_none(self):
+        try:
+            return self.memory.chat_memory.messages[-1]
+        except IndexError:
+            return None
 
 
 def create_chat_chain(
     config: OllamaConfig,
 ) -> RunnableSerializable[dict[str, Any], BaseMessage]:
-    """Create the base chat chain for use by the code."""
+    """Create the chat chain for use by the code."""
     llm = ChatOllama(
         model=config.chat_model,
         base_url=config.base_url,
@@ -32,16 +40,28 @@ def create_chat_chain(
     prompt = ChatPromptTemplate.from_messages(
         [
             ("system", config.system_prompt),
+            MessagesPlaceholder("chat_history"),
             ("human", "Context\n{context}\n\nQuestion:\n{input}"),
         ]
     )
 
+    def get_chat_history(inputs: dict) -> list:
+        """Extract chat history from memory object."""
+        try:
+            return inputs["memory"].chat_memory.messages
+        except AttributeError:
+            return []
+
     def get_context(inputs: dict) -> str:
         """Extract the RAG context from the input object"""
-        return inputs.get("context", "")
+        try:
+            return inputs["context"]
+        except AttributeError:
+            return ""
 
     return (
         RunnablePassthrough.assign(
+            chat_history=RunnableLambda(get_chat_history),
             context=RunnableLambda(get_context),
         )
         | prompt
@@ -49,28 +69,32 @@ def create_chat_chain(
     )
 
 
-def create_chat_client(config: OllamaConfig):
-    base_chain = create_chat_chain(config)
+def create_memory():
+    return ConversationBufferMemory(memory_key="chat_history", return_messages=True)
 
+
+def create_chat_client(config: OllamaConfig):
     return ChatClient(
-        chain=base_chain,
+        chain=create_chat_chain(config),
+        memory=create_memory(),
     )
 
 
 def chat_with_client(
-    client: ChatClient,
-    message: str,
-    retriever: VectorStoreRetriever | None = None,
-    verbose: bool = False,
-) -> str:
-    """Chat with the client and return the response content."""
+    client: ChatClient, message: str, retriever: VectorStoreRetriever | None = None
+) -> ChatClient:
+
     if retriever:
         context = get_context_from_store(message, retriever)
     else:
         context = ""
 
     response = client.chain.invoke(
-        {"input": message, "context": context},
+        {"input": message, "memory": client.memory, "context": context}
     )
 
-    return response.content
+    memory = client.memory
+    memory.chat_memory.add_user_message(message)
+    memory.chat_memory.add_ai_message(response.content)
+
+    return ChatClient(chain=client.chain, memory=memory)

src/reviewllama/reviewllama.py

@@ -19,7 +19,7 @@ def run_reviewllama(config: ReviewConfig):
         retriever = create_and_log_vector_store_retriever(analysis.repo, config.ollama)
 
         for diff in analysis.diffs:
-            get_suggestions(diff, retriever, chat_client, config.verbose)
+            chat_client = get_suggestions(diff, retriever, chat_client)
 
 
 def create_and_log_chat_client(config: OllamaConfig) -> ChatClient:
@@ -44,16 +44,20 @@ def create_and_log_vector_store_retriever(
 
 
 def get_suggestions(
-    diff: GitDiff,
-    retriever: VectorStoreRetriever,
-    chat_client: ChatClient,
-    verbose: bool,
-):
-    response = chat_with_client(chat_client, craft_message(diff), retriever, verbose)
-    log_info(response)
+    diff: GitDiff, retriever: VectorStoreRetriever, chat_client: ChatClient
+) -> ChatClient:
+    new_client = chat_with_client(chat_client, craft_message(diff), retriever)
+    log_info(str(new_client.get_last_response_or_none().content))
+    return new_client
 
 
 def craft_message(diff) -> str:
     return (
+        "Review the following code changes and make up to three suggestions on "
+        "how to improve it. If the code is sufficiently simple or accurate then say "
+        "no suggestions can be found. Important issues you should consider are consistent "
+        "style, introduction of syntax errors, and potentially breaking changes in "
+        "interfaces/APIs that aren't properly handled.\n\n"
+        f"The original code:\n```\n{diff.old_content}\n```\n"
         f"The new code:\n```\n{diff.new_content}```"
     )

tests/test_llm.py

@@ -17,10 +17,11 @@ def test_chat_client(ollama_config, chat_client):
     if not is_ollama_available(ollama_config):
         pytest.skip("Local Ollama server is not available")
 
-    response = chat_with_client(
+    chat_client = chat_with_client(
         chat_client, "Tell me your name and introduce yourself briefly"
     )
+    response = chat_client.get_last_response_or_none()
 
     assert response is not None
-    assert len(response) > 0
-    assert "gemma" in response.lower()
+    assert len(response.content) > 0
+    assert "gemma" in response.content.lower()