Enable Langserve chat server and playground (#17)

* Enable Langserve

* Use relative path

* update readme

* Use direct imports

.github/workflows/CI.yml

@@ -39,6 +39,7 @@ jobs:
         # exit-zero treats all errors as warnings. The GitHub editor is 127 chars wide
         flake8 . --count --exit-zero --max-complexity=10 --max-line-length=127 --statistics
     - name: Build and Run
+      working-directory: ./app
       run: |
         pip install -e .
         llama4u -q "Which is the best Llama model to use?"

README.md

@@ -8,10 +8,18 @@ Llama4U is a privacy-focused AI assistant developed using [Ollama][1], [LangChai
 ## Steps to run
 1. Host `llama3` model from [Ollama][1] on your computer.
 2. Clone this repository.
-3. `pip install -e .`
-4. `llama4u`
 
-`llama4u --help` for full CLI.
+There are 2 usage modes:
+
+### LangServe
+3. `pip install -U langchain-cli && langchain serve`
+4. The default server is hosted on `127.0.0.1` or `localhost` and port `8000`.
+5. Playground can be accessed at `<host_ip>:<port>/llama4u/playground`.
+
+### CLI
+3. `cd app/ && pip install -e .`
+4. `llama4u`
+5. `llama4u --help` for full CLI.
 
 ## List of chat commands
 

app/pyproject.toml

@@ -0,0 +1,34 @@
+[build-system]
+requires = ["setuptools>=61.0"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "llama4u"
+version = "0.1"
+authors = [
+    {name = "Viraj Malia", email="[email protected]"}
+]
+description = '''
+    Llama4U is a privacy-focused AI assistant developed using Ollama, LangChain and Llama3.
+'''
+readme = "README.md"
+requires-python = ">=3.7"
+classifiers = [
+    "Programming Language :: Python :: 3",
+    "License :: MIT License",
+    "Operating System :: OS Independent",
+]
+dependencies = [
+    "langchain",
+    "langchain-core",
+    "langchain-community",
+    "langchain-chroma",
+    "termcolor"
+]
+
+[project.scripts]
+llama4u = "llama4u:main"
+
+[project.urls]
+"Homepage" = "https://github.com/virajmalia/llama4u"
+"Bug Tracker" = "https://github.com/virajmalia/llama4u/issues"

app/server.py

@@ -0,0 +1,66 @@
+#!/usr/bin/env python
+""" Llama4U Server """
+from typing import List, Union
+from fastapi import FastAPI
+from fastapi.middleware.cors import CORSMiddleware
+from langchain.pydantic_v1 import Field
+from langchain_core.runnables import RunnableLambda
+from langchain_core.messages import HumanMessage, AIMessage
+from langserve import CustomUserType
+from langserve.server import add_routes
+from app.src.llama4u import Llama4U
+
+app = FastAPI(
+    title="LangChain Server",
+    version="1.0",
+    description="Spin up a simple api server using Langchain's Runnable interfaces",
+)
+
+# Set all CORS enabled origins
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+    expose_headers=["*"],
+)
+
+class ChatHistory(CustomUserType):
+    """ Playground Chat Widget """
+    chat: List[Union[HumanMessage, AIMessage]] = \
+        Field(..., extra={"widget": {"type": "chat", "input": "chat"}})
+
+def format_input(input_data: ChatHistory):
+    """ Format input from the langserve UI """
+    chat_history = input_data.chat
+
+    msg_history = []
+    for message in chat_history:
+        if isinstance(message, HumanMessage):
+            msg_history.append({"role": "user", "content": message.content})
+        elif isinstance(message, AIMessage):
+            msg_history.append({"role": "assistant", "content": message.content})
+
+    input_data = {"input": msg_history}
+
+    return input_data
+
+def format_output(response_data):
+    """ Format output from the Chain for the langserve UI """
+    return response_data.content
+
+llama4u = Llama4U()
+input_formatter = RunnableLambda(format_input)
+chat_model = input_formatter | llama4u.with_msg_history
+add_routes(
+    app,
+    chat_model.with_types(input_type=ChatHistory, output_type=ChatHistory),
+    config_keys=["configurable"],
+    path="/llama4u",
+)
+
+if __name__ == "__main__":
+    import uvicorn
+
+    uvicorn.run(app, host="localhost")

pyproject.toml

@@ -1,34 +1,23 @@
-[build-system]
-requires = ["setuptools>=61.0"]
-build-backend = "setuptools.build_meta"
-
-[project]
-name = "llama4u"
-version = "0.1"
-authors = [
-    {name = "Viraj Malia", email="[email protected]"}
-]
-description = '''
-    Llama4U is a free AI solution that can be hosted locally, while providing online capabilities in a responsible and user-controllable way.
-'''
+[tool.poetry]
+name = "llama4u-server"
+version = "0.1.0"
+description = ""
+authors = ["Viraj Malia [email protected]"]
 readme = "README.md"
-requires-python = ">=3.7"
-classifiers = [
-    "Programming Language :: Python :: 3",
-    "License :: MIT License",
-    "Operating System :: OS Independent",
-]
-dependencies = [
-    "langchain",
-    "langchain-core",
-    "langchain-community",
-    "langchain-chroma",
-    "termcolor"
+packages = [
+    { include = "app" },
 ]
 
-[project.scripts]
-llama4u = "llama4u:main"
+[tool.poetry.dependencies]
+python = "^3.11"
+uvicorn = "^0.23.2"
+langserve = {extras = ["server"], version = ">=0.0.30"}
+pydantic = "<2"
 
-[project.urls]
-"Homepage" = "https://github.com/virajmalia/llama4u"
-"Bug Tracker" = "https://github.com/virajmalia/llama4u/issues"
+
+[tool.poetry.group.dev.dependencies]
+langchain-cli = ">=0.0.15"
+
+[build-system]
+requires = ["poetry-core"]
+build-backend = "poetry.core.masonry.api"