Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

feat:add system prompt support--backend #44

Merged
merged 1 commit into from
Sep 24, 2024
Merged

feat:add system prompt support--backend #44

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions backend/app/core/workflow/init_graph.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -140,6 +140,7 @@ def initialize_graph(
break
if model_info is None:
raise ValueError(f"Model {model_name} not supported now.")

# in the future wo can use more langchain templates here apply to different node type TODO
if is_sequential:
# node_class = SequentialWorkerNode
Expand Down Expand Up @@ -179,6 +180,7 @@ def initialize_graph(
openai_api_key=model_info["api_key"],
openai_api_base=model_info["base_url"],
temperature=node_data["temperature"],
system_prompt=node_data["systemMessage"],
).work
),
)
Expand Down
97 changes: 58 additions & 39 deletions backend/app/core/workflow/node.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -162,8 +162,9 @@ def __init__(
openai_api_key: str,
openai_api_base: str,
temperature: float,
system_prompt: str,
):

self.system_prompt = system_prompt
if provider in ["zhipuai", "Siliconflow"]:
self.model = ChatOpenAI(
model=model,
Expand Down Expand Up @@ -236,6 +237,62 @@ def get_team_members_name(
return ",".join(list(team_members))


class LLMNode(BaseNode):
"""Perform LLM Node actions"""

async def work(self, state: TeamState, config: RunnableConfig) -> ReturnTeamState:
if self.system_prompt:
llm_node_prompts = ChatPromptTemplate.from_messages(
[
(
"system",
"Perform the task given to you.\n"
"If you are unable to perform the task, that's OK, you can ask human for help, or just say that you are unable to perform the task."
"Execute what you can to make progress. "
"And your role is:" + self.system_prompt + "\n"
"Stay true to your role and use your tools if necessary.\n\n",
),
(
"human",
"Here is the previous conversation: \n\n {history_string} \n\n Provide your response.",
),
MessagesPlaceholder(variable_name="messages"),
]
)

else:
llm_node_prompts = ChatPromptTemplate.from_messages(
[
(
"system",
(
"Perform the task given to you.\n"
"If you are unable to perform the task, that's OK, you can ask human for help, or just say that you are unable to perform the task."
"Execute what you can to make progress. "
"Stay true to your role and use your tools if necessary.\n\n"
),
),
(
"human",
"Here is the previous conversation: \n\n {history_string} \n\n Provide your response.",
),
MessagesPlaceholder(variable_name="messages"),
]
)
history = state.get("history", [])
messages = state.get("messages", [])
prompt = llm_node_prompts.partial(history_string=format_messages(history))
chain: RunnableSerializable[dict[str, Any], AnyMessage] = prompt | self.model
result: AIMessage = await chain.ainvoke(state, config)

return_state: ReturnTeamState = {
"history": history + [result],
"messages": [result] if result.tool_calls else [],
"all_messages": messages + [result],
}
return return_state


class WorkerNode(BaseNode):
worker_prompt = ChatPromptTemplate.from_messages(
[
Expand Down Expand Up @@ -334,44 +391,6 @@ async def work(self, state: TeamState, config: RunnableConfig) -> ReturnTeamStat
return return_state


class LLMNode(BaseNode):
"""Perform Sequential Worker actions"""

worker_prompt = ChatPromptTemplate.from_messages(
[
(
"system",
(
"Perform the task given to you.\n"
"If you are unable to perform the task, that's OK, you can ask human for help, or just say that you are unable to perform the task."
"Execute what you can to make progress. "
"Stay true to your role and use your tools if necessary.\n\n"
),
),
(
"human",
"Here is the previous conversation: \n\n {history_string} \n\n Provide your response.",
),
MessagesPlaceholder(variable_name="messages"),
]
)

async def work(self, state: TeamState, config: RunnableConfig) -> ReturnTeamState:
history = state.get("history", [])
messages = state.get("messages", [])
prompt = self.worker_prompt.partial(history_string=format_messages(history))
chain: RunnableSerializable[dict[str, Any], AnyMessage] = prompt | self.model
work_chain = chain
result: AIMessage = await work_chain.ainvoke(state, config)

return_state: ReturnTeamState = {
"history": history + [result],
"messages": [result] if result.tool_calls else [],
"all_messages": messages + [result],
}
return return_state


class LeaderNode(BaseNode):
leader_prompt = ChatPromptTemplate.from_messages(
[
Expand Down
Loading