Merge pull request #877 from Undertone0809/v1.18.1/docs-add-streamlit…

…-widget

add e2b code-interpreter

.gitignore

@@ -111,6 +111,3 @@ venv.bak/
 .mypy_cache/
 _trial_temp
 .idea
-
-# vscode
-.vscode

.vscode/extensions.json

@@ -0,0 +1,8 @@
+{
+  "recommendations": [
+    "ms-python.python",
+    "ms-python.vscode-pylance",
+    "ms-python.black-formatter",
+    "charliermarsh.ruff"
+  ]
+}

Makefile

@@ -34,13 +34,10 @@ install-docs:
 pre-commit-install:
 	poetry run pre-commit install
 
-polish-codestyle:
+format:
 	poetry run ruff format --config pyproject.toml promptulate tests example
 	poetry run ruff check --fix --config pyproject.toml promptulate tests example
 
-formatting: polish-codestyle
-format: polish-codestyle
-
 test:
 	$(TEST_COMMAND)
 
@@ -52,6 +49,8 @@ check-codestyle:
 	poetry run ruff format --check --config pyproject.toml promptulate tests example
 	poetry run ruff check --config pyproject.toml promptulate tests example
 
+check-format: check-codestyle
+
 lint: check-codestyle test
 
 # https://github.com/Maxlinn/linn-jupyter-site-template/blob/main/.github/workflows/linn-jupyter-site-template-deploy.yml
@@ -95,10 +94,11 @@ help:
 	@echo "install-docs: Install the dependencies for building docs"
 	@echo "pre-commit-install: Install the pre-commit hooks"
 	@echo "polish-codestyle: Format the code"
-	@echo "formatting: Format the code"
+	@echo "format: Format the code"
+	@echo "check-format: Check the code format"
+	@echo "check-codestyle: Check the code style, the same as make check-format"
 	@echo "test: Run the tests"
 	@echo "test-prod: Run the tests for production"
-	@echo "check-codestyle: Check the code style"
 	@echo "lint: Run the tests and check the code style"
 	@echo "build-docs: Build the docs"
 	@echo "start-docs: Start the docs server"
@@ -109,4 +109,4 @@ help:
 	@echo "build-remove: Remove the build directory"
 	@echo "cleanup: Remove all the cache files"
 
-.PHONY: lock install install-integration install-docs pre-commit-install polish-codestyle formatting format test test-prod check-codestyle lint build-docs start-docs pycache-remove dsstore-remove ipynbcheckpoints-remove pytestcache-remove build-remove cleanup help
+.PHONY: lock install install-integration install-docs pre-commit-install polish-codestyle formatting format test test-prod check-codestyle lint build-docs start-docs pycache-remove dsstore-remove ipynbcheckpoints-remove pytestcache-remove build-remove cleanup help check-format

example/agent/tool_agent_terminal.py

@@ -1,4 +1,5 @@
 """A Agent with duckduckgo search"""
+
 from promptulate.agents import ToolAgent
 from promptulate.tools import DuckDuckGoTool
 

example/ai-code-interpreter/README.md

@@ -0,0 +1,31 @@
+# AI-Code-Interpreter
+
+This example is an example of building ai code interpreter using promptulate.
+
+You see try the live demo [here](https://pne-chatbot.streamlit.app/).
+
+# Quick Start
+
+1. Clone the repository and install the dependencies
+
+```shell
+git clone https://www.github.com/Undertone0809/promptulate
+```
+
+2. Switch the current directory to the example
+
+```shell
+cd ./example/ai-code-interpreter
+```
+
+3. Install the dependencies
+
+```shell
+pip install -r requirements.txt
+```
+
+4. Run the application
+
+```shell
+streamlit run app.py
+```

example/ai-code-interpreter/app.py

@@ -0,0 +1,171 @@
+import base64
+import os
+import re
+from io import BytesIO
+from typing import Optional
+
+import pne
+import streamlit as st
+from e2b_code_interpreter import CodeInterpreter
+from prompt import system_prompt
+
+
+def extract_code(text) -> list:
+    pattern = r"```python\n(.*?)\n```"
+    matches = re.findall(pattern, text, re.DOTALL)
+    return matches
+
+
+def render_model() -> dict:
+    e2b_key = st.text_input("E2B Key")
+    return {"e2b_key": e2b_key}
+
+
+def handle_code_block(
+    code: str, config, code_interpreter: Optional[CodeInterpreter] = None
+):
+    code_interpreter = code_interpreter or CodeInterpreter()
+    max_retries = 3
+    retry_count = 0
+
+    st.write("Running code:")
+    st.code(code, language="python")
+
+    while retry_count < max_retries:
+        exec = code_interpreter.notebook.exec_cell(code)
+
+        if exec.logs:
+            st.write("Run Logs:")
+            st.text(exec.logs)
+
+        if exec.error:
+            error_message = f"An error occurred: {exec.error}"
+            st.error(error_message)
+
+            error_response = pne.chat(
+                model=config.model_name,
+                messages=[
+                    *st.session_state.messages,
+                    {
+                        "role": "assistant",
+                        "content": "I encountered an error while executing the code.",
+                    },
+                    {
+                        "role": "user",
+                        "content": f"The code execution resulted in an error: {error_message}. Can you explain what went wrong and provide a corrected version of the code?",  # noqa
+                    },
+                ],
+                model_config={
+                    "api_base": config.api_base,
+                    "api_key": config.api_key,
+                },
+            )
+
+            # st.write("Let me help you fix that eror:")
+            st.write(error_response)
+
+            corrected_code_blocks = extract_code(error_response)
+            if corrected_code_blocks:
+                code = corrected_code_blocks[0]
+                st.write("Corrected code:")
+                st.code(code, language="python")
+
+            retry_count += 1
+        else:
+            st.success("Code executed successfully!")
+            for output in exec.results:
+                if output.png:
+                    image = BytesIO(base64.b64decode(output.png))
+                    st.image(image)
+                elif output.text:
+                    st.text(output.text)
+            break
+
+    if retry_count == max_retries:
+        st.error(
+            f"Failed to execute the code after {max_retries} attempts. Please review the code and try again."  # noqa
+        )
+
+
+def additional_config() -> dict:
+    e2b_api_key = st.text_input("E2B API Key")
+    return {"e2b_api_key": e2b_api_key}
+
+
+def main():
+    st.set_page_config(layout="wide")
+    config = pne.beta.st.model_sidebar(additional_sidebar_fn=additional_config)
+    os.environ["E2B_API_KEY"] = config.e2b_api_key
+
+    st.title("💬 Code-Interpreter")
+    st.caption(
+        "🚀 Code-Interpreter is a tool that helps you run Python code snippets in your chat messages."  # noqa
+    )
+
+    # Initialize session state
+    if "show_code_results" not in st.session_state:
+        st.session_state.show_code_results = False
+    if "code_blocks" not in st.session_state:
+        st.session_state.code_blocks = []
+
+    # Create a layout with a dynamic right column
+    left_col, right_col = st.columns([2, 1])
+
+    with left_col:
+        if "messages" not in st.session_state:
+            st.session_state["messages"] = [
+                {
+                    "role": "assistant",
+                    "content": system_prompt,
+                }
+            ]
+
+        for msg in st.session_state.messages:
+            st.chat_message(msg["role"]).write(msg["content"])
+
+        if prompt := st.chat_input("How can I help you?"):
+            if not config.api_key:
+                st.info("Please add your API key to continue.")
+                st.stop()
+
+            st.session_state.messages.append({"role": "user", "content": prompt})
+            st.chat_message("user").write(prompt)
+
+            stream = pne.chat(
+                model=config.model_name,
+                stream=True,
+                messages=st.session_state.messages,
+                model_config={"api_base": config.api_base, "api_key": config.api_key},
+            )
+            response = st.chat_message("assistant").write_stream(stream)
+
+            st.session_state.code_blocks = extract_code(response)
+
+            if st.session_state.code_blocks:
+                st.session_state.show_code_results = True
+                st.chat_message("assistant").write(
+                    f"Found {len(st.session_state.code_blocks)} code blocks."
+                )
+
+            st.session_state.messages.append({"role": "assistant", "content": response})
+
+    if st.session_state.code_blocks:
+        st.button(
+            "Toggle Code Results",
+            on_click=lambda: setattr(
+                st.session_state,
+                "show_code_results",
+                not st.session_state.show_code_results,
+            ),
+        )
+
+    if st.session_state.show_code_results:
+        with right_col:
+            st.header("Code Execution Results")
+            with CodeInterpreter() as interpreter:
+                for code in st.session_state.code_blocks:
+                    handle_code_block(code, config, interpreter)
+
+
+if __name__ == "__main__":
+    main()

example/ai-code-interpreter/prompt.py

@@ -0,0 +1,7 @@
+system_prompt = """
+You are a helpful assistant. When you output markdown python code, code-interpreter interprets the code blocks and executes the code. so You don't need to output result.
+If you are using some third-party library, you need to first output as follows, then generate the code.
+```python
+!pip install {package}
+```
+"""  # noqa

example/ai-code-interpreter/requirements.txt

@@ -0,0 +1,3 @@
+streamlit>=1.28
+pne
+e2b_code_interpreter