diff --git a/README.md b/README.md
index 6ea69a9..7c1fdca 100644
--- a/README.md
+++ b/README.md
@@ -13,7 +13,8 @@ This package(>=1.0.0) will focus on wrapper of official chat elements to make ch
A Streamlit component to show chat messages.
It's basiclly a wrapper of streamlit officeial elements including the chat elemnts.
-
+
+
## Features
@@ -83,10 +84,10 @@ if query := st.chat_input('input your question here'):
text = ""
for x, docs in generator:
text += x
- chat_box.update_msg(text, 0, streaming=True)
- chat_box.update_msg("\n\n".join(docs), 1, streaming=False)
+ chat_box.update_msg(text, element_index=0, streaming=True)
# update the element without focus
- chat_box.update_msg(text, 0, streaming=False)
+ chat_box.update_msg(text, element_index=0, streaming=False, state="complete")
+ chat_box.update_msg("\n\n".join(docs), element_index=1, streaming=False, state="complete")
else:
text, docs = llm.chat(query)
chat_box.ai_say(
@@ -98,7 +99,8 @@ if query := st.chat_input('input your question here'):
]
)
-if st.button('show me the multimedia'):
+cols = st.columns(2)
+if cols[0].button('show me the multimedia'):
chat_box.ai_say(Image(
'https://tse4-mm.cn.bing.net/th/id/OIP-C.cy76ifbr2oQPMEs2H82D-QHaEv?w=284&h=181&c=7&r=0&o=5&dpr=1.5&pid=1.7'))
time.sleep(0.5)
@@ -108,6 +110,29 @@ if st.button('show me the multimedia'):
chat_box.ai_say(
Audio('https://sample-videos.com/video123/mp4/720/big_buck_bunny_720p_1mb.mp4'))
+if cols[1].button('run agent'):
+ chat_box.user_say('run agent')
+ agent = FakeAgent()
+ text = ""
+
+ if streaming:
+ # streaming:
+ chat_box.ai_say() # generate a blank placeholder to render messages
+ for d in agent.run_stream():
+ if d["type"] == "complete":
+ chat_box.update_msg(expanded=False, state="complete")
+ chat_box.insert_msg(d["llm_output"])
+ break
+
+ if d["status"] == 1:
+ chat_box.update_msg(expanded=False, state="complete")
+ text = ""
+ chat_box.insert_msg(Markdown(text, title=d["text"], in_expander=True, expanded=True))
+ elif d["status"] == 2:
+ text += d["llm_output"]
+ chat_box.update_msg(text, streaming=True)
+ else:
+ chat_box.update_msg(text, streaming=False)
btns.download_button(
"Export Markdown",
@@ -130,7 +155,6 @@ if btns.button("clear history"):
if show_history:
st.write(chat_box.history)
-
```
## Todos
@@ -160,3 +184,5 @@ if show_history:
- [x] export to markdown
- [x] export to json
- [x] import json
+
+- [x] support output of langchain' Agent.
diff --git a/demo_agent.gif b/demo_agent.gif
new file mode 100644
index 0000000..19844aa
Binary files /dev/null and b/demo_agent.gif differ
diff --git a/example.py b/example.py
index caa57e5..7ffe8a4 100644
--- a/example.py
+++ b/example.py
@@ -47,10 +47,10 @@
text = ""
for x, docs in generator:
text += x
- chat_box.update_msg(text, 0, streaming=True)
- chat_box.update_msg("\n\n".join(docs), 1, streaming=False)
+ chat_box.update_msg(text, element_index=0, streaming=True)
# update the element without focus
- chat_box.update_msg(text, 0, streaming=False)
+ chat_box.update_msg(text, element_index=0, streaming=False, state="complete")
+ chat_box.update_msg("\n\n".join(docs), element_index=1, streaming=False, state="complete")
else:
text, docs = llm.chat(query)
chat_box.ai_say(
@@ -62,7 +62,8 @@
]
)
-if st.button('show me the multimedia'):
+cols = st.columns(2)
+if cols[0].button('show me the multimedia'):
chat_box.ai_say(Image(
'https://tse4-mm.cn.bing.net/th/id/OIP-C.cy76ifbr2oQPMEs2H82D-QHaEv?w=284&h=181&c=7&r=0&o=5&dpr=1.5&pid=1.7'))
time.sleep(0.5)
@@ -72,6 +73,28 @@
chat_box.ai_say(
Audio('https://sample-videos.com/video123/mp4/720/big_buck_bunny_720p_1mb.mp4'))
+if cols[1].button('run agent'):
+ chat_box.user_say('run agent')
+ agent = FakeAgent()
+ text = ""
+
+ # streaming:
+ chat_box.ai_say() # generate a blank placeholder to render messages
+ for d in agent.run_stream():
+ if d["type"] == "complete":
+ chat_box.update_msg(expanded=False, state="complete")
+ chat_box.insert_msg(d["llm_output"])
+ break
+
+ if d["status"] == 1:
+ chat_box.update_msg(expanded=False, state="complete")
+ text = ""
+ chat_box.insert_msg(Markdown(text, title=d["text"], in_expander=True, expanded=True))
+ elif d["status"] == 2:
+ text += d["llm_output"]
+ chat_box.update_msg(text, streaming=True)
+ else:
+ chat_box.update_msg(text, streaming=False)
btns.download_button(
"Export Markdown",
diff --git a/setup.py b/setup.py
index a0532b2..c44755a 100644
--- a/setup.py
+++ b/setup.py
@@ -9,7 +9,7 @@ def readme():
setuptools.setup(
name='streamlit-chatbox',
- version='1.1.7',
+ version='1.1.8',
author='liunux',
author_email='liunux@qq.com',
description='A chat box and some helpful tools used to build chatbot app with streamlit',
@@ -21,7 +21,7 @@ def readme():
classifiers=[],
python_requires='>=3.8',
install_requires=[
- 'streamlit>=1.24.0',
+ 'streamlit>=1.26.0',
'simplejson',
]
)
diff --git a/streamlit_chatbox/__init__.py b/streamlit_chatbox/__init__.py
index 069ac14..1c26c67 100644
--- a/streamlit_chatbox/__init__.py
+++ b/streamlit_chatbox/__init__.py
@@ -14,6 +14,7 @@
"Video",
"OutputElement",
"FakeLLM",
+ "FakeAgent",
]
diff --git a/streamlit_chatbox/elements.py b/streamlit_chatbox/elements.py
index 3e781c2..213ac81 100644
--- a/streamlit_chatbox/elements.py
+++ b/streamlit_chatbox/elements.py
@@ -1,6 +1,6 @@
from typing import *
import streamlit as st
-from pydantic import BaseModel, Field
+# from pydantic import BaseModel, Field
class Element:
@@ -9,12 +9,13 @@ class Element:
'''
def __init__(self,
+ *,
output_method: str = "markdown",
- *args: Any,
+ metadata: Dict = {},
**kwargs: Any,
) -> None:
self._output_method = output_method
- self._args = args
+ self._metadata = metadata
self._kwargs = kwargs
self._defualt_kwargs = {
"markdown": {
@@ -39,15 +40,28 @@ def _set_default_kwargs(self) -> None:
for k, v in default.items():
self._kwargs.setdefault(k, v)
- def __call__(self) -> st._DeltaGenerator:
+ def __call__(self, render_to: st._DeltaGenerator=None) -> st._DeltaGenerator:
# assert self._dg is None, "Every element can be rendered once only."
- self._place_holder = st.empty()
+ render_to = render_to or st
+ self._place_holder = render_to.empty()
output_method = getattr(self._place_holder, self._output_method)
assert callable(
output_method), f"The attribute st.{self._output_mehtod} is not callable."
self._dg = output_method(*self._args, **self._kwargs)
return self._dg
+ @property
+ def dg(self) -> st._DeltaGenerator:
+ return self._dg
+
+ @property
+ def place_holder(self) -> st._DeltaGenerator:
+ return self._place_holder
+
+ @property
+ def metadata(self) -> Dict:
+ return self._metadata
+
class OutputElement(Element):
def __init__(self,
@@ -56,6 +70,7 @@ def __init__(self,
title: str = "",
in_expander: bool = False,
expanded: bool = False,
+ state: Literal["running", "complete", "error"] = "running",
**kwargs: Any,
) -> None:
super().__init__(output_method=output_method, **kwargs)
@@ -63,18 +78,23 @@ def __init__(self,
self._title = title
self._in_expander = in_expander
self._expanded = expanded
+ self._state = state
+ self._attrs = ["_content", "_output_method", "_kwargs", "_metadata",
+ "_title", "_in_expander", "_expanded", "_state",]
- def __call__(self) -> st._DeltaGenerator:
- self._args = (self._content,)
- if self._in_expander:
- with st.expander(self._title, self._expanded):
- return super().__call__()
- else:
- return super().__call__()
+ def clone(self) -> "OutputElement":
+ obj = type(self)()
+ for n in self._attrs:
+ setattr(obj, n, getattr(self, n))
+ return obj
+
+ @property
+ def content(self) -> Union[str, bytes]:
+ return self._content
def __repr__(self) -> str:
method = self._output_method.capitalize()
- return f"{method} Element:\n{self._content}"
+ return f"{method} Element:\n{self.content}"
def to_dict(self) -> Dict:
return {
@@ -83,6 +103,8 @@ def to_dict(self) -> Dict:
"title": self._title,
"in_expander": self._in_expander,
"expanded": self._expanded,
+ "state": self._state,
+ "metadata": self._metadata,
"kwargs": self._kwargs,
}
@@ -101,6 +123,8 @@ def from_dict(cls, d: Dict) -> "OutputElement":
title=d.get("title"),
in_expander=d.get("in_expander"),
expanded=d.get("expanded"),
+ state=d.get("state"),
+ metadata=d.get("metadata", {}),
**d.get("kwargs", {}),
)
if factory_cls is cls:
@@ -108,18 +132,52 @@ def from_dict(cls, d: Dict) -> "OutputElement":
return factory_cls(**kwargs)
+ def __call__(self, render_to: st._DeltaGenerator=None, direct: bool=False) -> st._DeltaGenerator:
+ if render_to is None:
+ if self._place_holder is None:
+ self._place_holder = st.empty()
+ else:
+ if direct:
+ self._place_holder = render_to
+ else:
+ self._place_holder = render_to.empty()
+ temp_dg = self._place_holder
+
+ if self._in_expander:
+ temp_dg = self._place_holder.status(self._title, expanded=self._expanded, state=self._state)
+ output_method = getattr(temp_dg, self._output_method)
+ assert callable(
+ output_method), f"The attribute st.{self._output_mehtod} is not callable."
+ self._dg = output_method(self._content, **self._kwargs)
+ return self._dg
+
def update_element(
self,
- element: "OutputElement",
- streaming: Optional[bool] = None,
+ element: "OutputElement" = None,
+ *,
+ title: str = None,
+ expanded: bool = None,
+ state: bool = None,
) -> st._DeltaGenerator:
- assert self._place_holder is not None, "You must render the element before setting new element."
- with self._place_holder:
- self._dg = element()
+ assert self.place_holder is not None, f"You must render the element {self} before setting new element."
+ attrs = {}
+ if title is not None:
+ attrs["_title"] = title
+ if expanded is not None:
+ attrs["_expanded"] = expanded
+ if state is not None:
+ attrs["_state"] = state
+
+ if element is None:
+ element = self
+ for k, v in attrs.items():
+ setattr(element, k, v)
+
+ element(self.place_holder, direct=True)
return self._dg
- def attrs_from(self, target):
- for attr in ["_in_expander", "_expanded", "_title"]:
+ def status_from(self, target):
+ for attr in ["_in_expander", "_expanded", "_title", "_state"]:
setattr(self, attr, getattr(target, attr))
@@ -128,24 +186,60 @@ class InputElement(Element):
class Markdown(OutputElement):
- def __init__(self, content: Union[str, bytes] = "", title: str = "", in_expander: bool = False, expanded: bool = False, **kwargs: Any) -> None:
+ def __init__(
+ self,
+ content: Union[str, bytes] = "",
+ title: str = "",
+ in_expander: bool = False,
+ expanded: bool = False,
+ state: Literal["running", "complete", "error"] = "running",
+ **kwargs: Any,
+ ) -> None:
super().__init__(content, output_method="markdown", title=title,
- in_expander=in_expander, expanded=expanded, **kwargs)
+ in_expander=in_expander, expanded=expanded,
+ state=state, **kwargs)
class Image(OutputElement):
- def __init__(self, content: Union[str, bytes] = "", title: str = "", in_expander: bool = False, expanded: bool = False, **kwargs: Any) -> None:
+ def __init__(
+ self,
+ content: Union[str, bytes] = "",
+ title: str = "",
+ in_expander: bool = False,
+ expanded: bool = False,
+ state: Literal["running", "complete", "error"] = "running",
+ **kwargs: Any,
+ ) -> None:
super().__init__(content, output_method="image", title=title,
- in_expander=in_expander, expanded=expanded, **kwargs)
+ in_expander=in_expander, expanded=expanded,
+ state=state, **kwargs)
class Audio(OutputElement):
- def __init__(self, content: Union[str, bytes] = "", title: str = "", in_expander: bool = False, expanded: bool = False, **kwargs: Any) -> None:
+ def __init__(
+ self,
+ content: Union[str, bytes] = "",
+ title: str = "",
+ in_expander: bool = False,
+ expanded: bool = False,
+ state: Literal["running", "complete", "error"] = "running",
+ **kwargs: Any,
+ ) -> None:
super().__init__(content, output_method="audio", title=title,
- in_expander=in_expander, expanded=expanded, **kwargs)
+ in_expander=in_expander, expanded=expanded,
+ state=state, **kwargs)
class Video(OutputElement):
- def __init__(self, content: Union[str, bytes] = "", title: str = "", in_expander: bool = False, expanded: bool = False, **kwargs: Any) -> None:
+ def __init__(
+ self,
+ content: Union[str, bytes] = "",
+ title: str = "",
+ in_expander: bool = False,
+ expanded: bool = False,
+ state: Literal["running", "complete", "error"] = "running",
+ **kwargs: Any,
+ ) -> None:
super().__init__(content, output_method="video", title=title,
- in_expander=in_expander, expanded=expanded, **kwargs)
+ in_expander=in_expander, expanded=expanded,
+ state=state, **kwargs)
diff --git a/streamlit_chatbox/messages.py b/streamlit_chatbox/messages.py
index a696b77..0d35e16 100644
--- a/streamlit_chatbox/messages.py
+++ b/streamlit_chatbox/messages.py
@@ -4,16 +4,28 @@
import simplejson as json
+# main concept：
+# ChatBox is the top level object, repsents all history messages。
+# every message is feed to st.chat_message，including two objects：
+# - role
+# - elements: list of OutputElement to render. every element includes：
+# - content and output_method, ie. st.output_method(content, **kwargs)
+# - in_exapander decides the element is rendered directly or in st.status
+# - if directly: element is rendered in a st.empty in the st.container
+# - if in expander: element is rendered in a st.empty in the st.status
+
+
class ChatBox:
def __init__(
self,
chat_name: str = "default",
- session_key: str = "messages",
+ session_key: str = "chat_history",
user_avatar: str = "user",
assistant_avatar: str = "assistant",
greetings: Union[str, OutputElement, List[Union[str, OutputElement]]] = [],
) -> None:
self._chat_name = chat_name
+ self._chat_containers = []
self._session_key = session_key
self._user_avatar = user_avatar
self._assistant_avatar = assistant_avatar
@@ -27,38 +39,42 @@ def __init__(
@property
def chat_inited(self):
- return self._session_key in st.session_state
+ return self._session_key in st.session_state.keys()
def init_session(self, clear: bool =False):
if not self.chat_inited or clear:
st.session_state[self._session_key] = {}
+ time.sleep(0.1)
self.reset_history(self._chat_name)
def reset_history(self, name=None):
- assert self.chat_inited, "please call init_session first"
+ if not self.chat_inited:
+ st.session_state[self._session_key] = {}
+
name = name or self._chat_name
- st.session_state[self._session_key].update({name: []})
+ st.session_state[self._session_key][name] = []
if self._greetings:
st.session_state[self._session_key][name] = [{
"role": "assistant",
"elements": self._greetings,
+ "metadata": {},
}]
def use_chat_name(self, name: str ="default") -> None:
- assert self.chat_inited, "please call init_session first"
+ self.init_session()
self._chat_name = name
if name not in st.session_state[self._session_key]:
self.reset_history(name)
def del_chat_name(self, name: str):
- assert self.chat_inited, "please call init_session first"
+ self.init_session()
if name in st.session_state[self._session_key]:
msgs = st.session_state[self._session_key].pop(name)
self._chat_name=self.get_chat_names()[0]
return msgs
def get_chat_names(self):
- assert self.chat_inited, "please call init_session first"
+ self.init_session()
return list(st.session_state[self._session_key].keys())
@property
@@ -67,11 +83,11 @@ def cur_chat_name(self):
@property
def history(self) -> List:
- assert self.chat_inited, "please call init_session first"
+ self.init_session()
return st.session_state[self._session_key].get(self._chat_name, [])
def other_history(self, chat_name: str, default: List=[]) -> Optional[List]:
- assert self.chat_inited, "please call init_session first"
+ self.init_session()
chat_name = chat_name or self.cur_chat_name
return st.session_state[self._session_key].get(chat_name, default)
@@ -87,13 +103,13 @@ def filter_history(
filter: custom filter fucntion with arguments (msg,) or (msg, index), return None if skipping msg. default filter returns all text/markdown content.
stop: custom function to stop filtering with arguments (history,) history is already filtered messages, return True if stop. default stop on history_len
'''
- assert self.chat_inited, "please call init_session first"
+ self.init_session()
def default_filter(msg, index=None):
'''
filter text messages only with the format {"role":role, "content":content}
'''
- content = [x._content for x in msg["elements"] if x._output_method in ["markdown", "text"]]
+ content = [x.content for x in msg["elements"] if x._output_method in ["markdown", "text"]]
return {
"role": msg["role"],
"content": "\n\n".join(content),
@@ -101,8 +117,8 @@ def default_filter(msg, index=None):
def default_stop(history):
if isinstance(history_len, int):
- ai_count = len(x for x in history if x["role"] == "user")
- return ai_count >= history_len
+ user_count = len(x for x in history if x["role"] == "user")
+ return user_count >= history_len
else:
return False
@@ -122,15 +138,15 @@ def default_stop(history):
filtered = filter(msg, i)
if filtered is not None:
result.insert(0, filtered)
- if isinstance(history_len, int) and len(result) >= history_len:
- break
+
+ if stop(history):
+ break
return result
def export2md(
self,
chat_name: str = None,
- filter: Callable = None,
user_avatar: str = "User",
ai_avatar: str = "AI",
user_bg_color: str = "#DCFDC8",
@@ -141,7 +157,7 @@ def export2md(
default export messages as table of text.
use callback(msg) to custom exported content.
'''
- assert self.chat_inited, "please call init_session first"
+ self.init_session()
lines = [
"\n"
"| | |\n",
@@ -156,7 +172,7 @@ def set_bg_color(text, bg_color):
if callable(callback):
line = callback(msg)
else:
- contents = [e._content for e in msg["elements"]]
+ contents = [e.content for e in msg["elements"]]
if msg["role"] == "user":
content = "
".join(set_bg_color(c, user_bg_color) for c in contents)
avatar = set_bg_color(user_avatar, user_bg_color)
@@ -173,7 +189,7 @@ def to_dict(
'''
export current state to dict
'''
- assert self.chat_inited, "please call init_session first"
+ self.init_session()
def p(val):
if isinstance(val, (list, tuple)):
@@ -242,56 +258,67 @@ def _prepare_elements(
elif isinstance(elements, list):
elements = [Markdown(e) if isinstance(
e, str) else e for e in elements]
- return elements
+ return elements or []
def user_say(
self,
- elements: Union[OutputElement, str, List[Union[OutputElement, str]]],
- to_history: bool = True,
- not_render: bool = False,
+ elements: Union[OutputElement, str, List[Union[OutputElement, str]]] = None,
+ metadata: Dict = {},
) -> List[OutputElement]:
- assert self.chat_inited, "please call init_session first"
+ self.init_session()
elements = self._prepare_elements(elements)
- if not not_render:
- with st.chat_message("user", avatar=self._user_avatar):
- for element in elements:
- element()
- if to_history:
- self.history.append({"role": "user", "elements": elements})
+
+ chat_ele = st.chat_message("user", avatar=self._user_avatar)
+ self._chat_containers.append(chat_ele)
+ for element in elements:
+ element(render_to=chat_ele)
+
+ self.history.append({"role": "user", "elements": elements, "metadata": metadata})
return elements
def ai_say(
self,
- elements: Union[OutputElement, str, List[Union[OutputElement, str]]],
- to_history: bool = True,
- not_render: bool = False,
+ elements: Union[OutputElement, str, List[Union[OutputElement, str]]] = None,
+ metadata: Dict = {},
) -> List[OutputElement]:
- assert self.chat_inited, "please call init_session first"
+ self.init_session()
elements = self._prepare_elements(elements)
- if not not_render:
- with st.chat_message("assistant", avatar=self._assistant_avatar):
- for element in elements:
- element()
- if to_history:
- self.history.append({"role": "assistant", "elements": elements})
+
+ chat_ele = st.chat_message("assistant", avatar=self._assistant_avatar)
+ container = chat_ele.container()
+ self._chat_containers.append(container)
+ for element in elements:
+ element(render_to=container)
+
+ self.history.append({"role": "assistant", "elements": elements, "metadata": metadata})
return elements
def output_messages(self):
- assert self.chat_inited, "please call init_session first"
+ self.init_session()
+ self._chat_containers = []
for msg in self.history:
avatar = self._user_avatar if msg["role"] == "user" else self._assistant_avatar
- with st.chat_message(msg["role"], avatar=avatar):
- for element in msg["elements"]:
- element()
+ chat_ele = st.chat_message(msg["role"], avatar=avatar)
+ container = chat_ele.container()
+ self._chat_containers.append(container)
+ for element in msg["elements"]:
+ element(render_to=container)
def update_msg(
self,
- element: Union["OutputElement", str],
+ element: Union["OutputElement", str] = None,
+ *,
element_index: int = -1,
history_index: int = -1,
streaming: Optional[bool] = None,
+ title: str = None,
+ expanded: bool = None,
+ state: bool = None,
) -> st._DeltaGenerator:
- assert self.chat_inited, "please call init_session first"
+ self.init_session()
+ if not self.history or not self.history[history_index]["elements"]:
+ return
+
if isinstance(element, str):
element = Markdown(element)
if streaming is None:
@@ -299,18 +326,41 @@ def update_msg(
if streaming and isinstance(element, Markdown):
element._content += " ▌"
- old_element = self.history[history_index]["elements"][element_index]
+ old_element: OutputElement = self.history[history_index]["elements"][element_index]
+ if element is not None:
+ element.status_from(old_element)
+ self.history[history_index]["elements"][element_index] = element
+
dg = old_element.update_element(
- element, streaming
+ element,
+ title=title,
+ expanded=expanded,
+ state=state,
)
- element.attrs_from(old_element)
- self.history[history_index]["elements"][element_index] = element
return dg
+ def insert_msg(
+ self,
+ element: Union["OutputElement", str],
+ *,
+ history_index: int = -1,
+ pos: int = -1,
+ ) -> OutputElement:
+ self.init_session()
+ if isinstance(element, str):
+ element = Markdown(element)
+ elements = self.history[history_index]["elements"]
+ if pos < 0:
+ pos += len(elements) + 1
+ elements.insert(pos, element)
+
+ element(render_to=self._chat_containers[history_index])
+ return element
+
class FakeLLM:
def _answer(self, query: str) -> str:
- answer = f"this is my answer for your question:\n\n{query}"
+ answer = f"this is llm answer for your question:\n\n{query}"
docs = ["reference 1", "reference 2", "reference 3"]
return answer, docs
@@ -322,3 +372,104 @@ def chat_stream(self, query: str):
for t in text:
yield t, docs
time.sleep(0.1)
+
+
+class FakeAgent:
+ llm = FakeLLM()
+ tools = ["search", "math"]
+
+ def thought(self, msg):
+ return f"thought {msg}"
+
+ def action(self, msg):
+ return f"action {msg}"
+
+ def run(self, query: str = "", steps: int = 2):
+ result = []
+ for i in range(1, steps + 1):
+ thought = self.thought(i)
+ result.append({
+ "type": "thought",
+ "id": i,
+ "text": thought,
+ "llm_output": self.llm.chat(thought)[0]
+ })
+
+ action = self.action(i)
+ result.append({
+ "type": "action",
+ "id": i,
+ "text": action,
+ "llm_output": self.llm.chat(action)[0]
+ })
+
+ result.append({
+ "type": "complete",
+ "llm_output": "final answer"
+ })
+
+ return result
+
+ def run_stream(self, query: str = "", steps: int = 2):
+ for i in range(1, steps + 1):
+ thought = self.thought(i)
+
+ yield {
+ "type": "thought",
+ "id": i,
+ "text": thought,
+ "status": 1,
+ "llm_output": "",
+ }
+
+ for chunk, _ in self.llm.chat_stream(thought):
+ d = {
+ "type": "thought",
+ "id": i,
+ "text": thought,
+ "status": 2,
+ "llm_output": chunk,
+ }
+ print(d)
+ yield d
+
+ yield {
+ "type": "thought",
+ "id": i,
+ "text": thought,
+ "status": 3,
+ "llm_output": "",
+ }
+
+ action = self.action(i)
+ yield {
+ "type": "action",
+ "id": i,
+ "text": action,
+ "status": 1,
+ "llm_output": "",
+ }
+
+ for chunk, _ in self.llm.chat_stream(action):
+ d = {
+ "type": "action",
+ "id": i,
+ "text": action,
+ "status": 2,
+ "llm_output": chunk,
+ }
+ print(d)
+ yield d
+
+ yield {
+ "type": "action",
+ "id": i,
+ "text": action,
+ "status": 3,
+ "llm_output": "",
+ }
+
+ yield {
+ "type": "complete",
+ "llm_output": "final answer"
+ }