Switched to Gradio; much more convenient AND much better results.

Dockerfile

@@ -1,17 +0,0 @@
-FROM python:3.11.7
-
-# Set up a new user named "user" with user ID 1000
-RUN useradd -m -u 1000 user
-
-# Switch to the "user" user
-USER user
-
-COPY . .
-
-WORKDIR /
-
-RUN pip install --no-cache-dir --upgrade -r /requirements.txt
-
-CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]
-
-

README.md

@@ -1,15 +1,18 @@
 ---
 title: Star-Wars-Expert
 colorFrom: yellow
-colorTo: yellow
-sdk: docker
+colorTo: blue
+sdk: gradio
 app_file: app.py
 pinned: false
+license: mit
 ---
 
 # Star-Wars-Expert
-A LLM with RAG making it knowledgeable about Star Wars plot and data.
 
+A LLM with RAG making it knowledgeable about Star Wars lore (built with LangChain, GPT-3.5 and Gradio).
+
+Chat with it [on its Huggingface space](https://huggingface.co/spaces/T-Flet/Star-Wars-Expert) (or clone the repo and `python app.py`).
 
 
 ## Data

app.py

@@ -1,71 +1,34 @@
-# Chains
-from langchain_core.pydantic_v1 import BaseModel, Field
+from langchain_core.messages import HumanMessage, AIMessage
 
-# To serve the app
-from fastapi import FastAPI
-from langchain_core.messages import BaseMessage
-from langserve import add_routes, CustomUserType
+import gradio as gr
 
+# For local testing; not used in the Huggingface space
 import dotenv
 dotenv.load_dotenv()
 
+# The available backends to use in the app
 from ingredients import script_db, woo_db, full_chain, compound_chain, agent_executor
 
 
+def chat(message, history):
+    formatted_history = []
+    for human, ai in history:
+        formatted_history.append(HumanMessage(content = human))
+        formatted_history.append(AIMessage(content = ai))
+    
+    # Yes, the context chat entries are not fed back to the system, but that is probably for the best due to input size limit
+    response = compound_chain.invoke(dict(input = HumanMessage(content = message), chat_history = formatted_history))
 
-## Type specifications (with unusual class-scope fields)
+    return response['answer']
 
-class StrInput(BaseModel):
-    input: str
 
-class Input(BaseModel):
-    input: str
-    chat_history: list[BaseMessage] = Field(
-        ...,
-        extra = dict(widget = dict(type = 'chat', input = 'location')),
-    )
-
-class Output(BaseModel):
-    output: str
-
-
-
-## App definition
-# NOTE: The chat playground type has a web page issue (flashes and becomes white, hence non-interactable; this was supposedly solved in an issue late last year)
-
-app = FastAPI(
-    title = 'Star Wars Expert',
-    version = '1.0',
-    description = 'A Star Wars expert chatbot',
-)
-
-
-# Basic retriever versions
-
-# add_routes(app, script_db.as_retriever())
-# add_routes(app, woo_db.as_retriever())
-
-
-# History-aware retriever version
-# add_routes(app, full_chain.with_types(input_type = StrInput, output_type = Output), playground_type = 'default')
-
-
-# Agent version
-
-# add_routes(app, agent_executor, playground_type = 'chat')
-# add_routes(app, agent_executor.with_types(input_type = StrInput, output_type = Output))
-
-
-# Non-agent chain-logic version
-
-add_routes(app, compound_chain.with_types(input_type = StrInput))
-# add_routes(app, compound_chain.with_types(input_type = Input), playground_type = 'chat')
-
-
-
-if __name__ == '__main__':
-    import uvicorn
-
-    uvicorn.run(app, host = 'localhost', port = 8000)
+gr.ChatInterface(
+    chat,
+    textbox = gr.Textbox(placeholder = 'Ask something about Star Wars', container = False, scale = 7),
+    title = 'Star Wars Expert', description = 'I am knowledgeable about Star Wars; ask me about it',
+    examples = ['Do you know the tragedy of Darth Plagueis the Wise?', 'What power source did the Death Star use?', "Who participates in Han's rescue from Jabba? And where is the palace?"],
+    cache_examples = False, # This would avoid invoking the chatbot for the example queries (it would invokes it on them on startup instead)
+    theme = 'soft', retry_btn = None, undo_btn = 'Delete Previous', clear_btn = 'Clear'
+).launch()
 
 

app_notebook.ipynb

@@ -1,69 +0,0 @@
-{
- "cells": [
-  {
-   "cell_type": "code",
-   "execution_count": 5,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "from langserve import RemoteRunnable\n",
-    "\n",
-    "remote_runnable = RemoteRunnable('http://localhost:8000/')"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 7,
-   "metadata": {},
-   "outputs": [
-    {
-     "ename": "HTTPStatusError",
-     "evalue": "Server error '500 Internal Server Error' for url 'http://localhost:8000/invoke'\nFor more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/500 for Internal Server Error",
-     "output_type": "error",
-     "traceback": [
-      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
-      "\u001b[1;31mHTTPStatusError\u001b[0m                           Traceback (most recent call last)",
-      "File \u001b[1;32mc:\\Users\\Dr-Lo\\miniconda3\\envs\\ML11\\Lib\\site-packages\\langserve\\client.py:157\u001b[0m, in \u001b[0;36m_raise_for_status\u001b[1;34m(response)\u001b[0m\n\u001b[0;32m    156\u001b[0m \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[1;32m--> 157\u001b[0m     \u001b[43mresponse\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mraise_for_status\u001b[49m\u001b[43m(\u001b[49m\u001b[43m)\u001b[49m\n\u001b[0;32m    158\u001b[0m \u001b[38;5;28;01mexcept\u001b[39;00m httpx\u001b[38;5;241m.\u001b[39mHTTPStatusError \u001b[38;5;28;01mas\u001b[39;00m e:\n",
-      "File \u001b[1;32mc:\\Users\\Dr-Lo\\miniconda3\\envs\\ML11\\Lib\\site-packages\\httpx\\_models.py:761\u001b[0m, in \u001b[0;36mResponse.raise_for_status\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m    760\u001b[0m message \u001b[38;5;241m=\u001b[39m message\u001b[38;5;241m.\u001b[39mformat(\u001b[38;5;28mself\u001b[39m, error_type\u001b[38;5;241m=\u001b[39merror_type)\n\u001b[1;32m--> 761\u001b[0m \u001b[38;5;28;01mraise\u001b[39;00m HTTPStatusError(message, request\u001b[38;5;241m=\u001b[39mrequest, response\u001b[38;5;241m=\u001b[39m\u001b[38;5;28mself\u001b[39m)\n",
-      "\u001b[1;31mHTTPStatusError\u001b[0m: Server error '500 Internal Server Error' for url 'http://localhost:8000/invoke'\nFor more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/500",
-      "\nDuring handling of the above exception, another exception occurred:\n",
-      "\u001b[1;31mHTTPStatusError\u001b[0m                           Traceback (most recent call last)",
-      "Cell \u001b[1;32mIn[7], line 1\u001b[0m\n\u001b[1;32m----> 1\u001b[0m \u001b[43mremote_runnable\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43minvoke\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;124;43mDo you know the tragedy of Darth Plagueis the Wise?\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[43m)\u001b[49m\n\u001b[0;32m      2\u001b[0m \u001b[38;5;66;03m# remote_runnable.invoke(dict(input = 'Do you know the tragedy of Darth Plagueis the Wise?', chat_history = []))\u001b[39;00m\n",
-      "File \u001b[1;32mc:\\Users\\Dr-Lo\\miniconda3\\envs\\ML11\\Lib\\site-packages\\langserve\\client.py:356\u001b[0m, in \u001b[0;36mRemoteRunnable.invoke\u001b[1;34m(self, input, config, **kwargs)\u001b[0m\n\u001b[0;32m    354\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m kwargs:\n\u001b[0;32m    355\u001b[0m     \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mNotImplementedError\u001b[39;00m(\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mkwargs not implemented yet.\u001b[39m\u001b[38;5;124m\"\u001b[39m)\n\u001b[1;32m--> 356\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_call_with_config\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_invoke\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;28;43minput\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mconfig\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mconfig\u001b[49m\u001b[43m)\u001b[49m\n",
-      "File \u001b[1;32mc:\\Users\\Dr-Lo\\miniconda3\\envs\\ML11\\Lib\\site-packages\\langchain_core\\runnables\\base.py:1625\u001b[0m, in \u001b[0;36mRunnable._call_with_config\u001b[1;34m(self, func, input, config, run_type, **kwargs)\u001b[0m\n\u001b[0;32m   1621\u001b[0m     context \u001b[38;5;241m=\u001b[39m copy_context()\n\u001b[0;32m   1622\u001b[0m     context\u001b[38;5;241m.\u001b[39mrun(var_child_runnable_config\u001b[38;5;241m.\u001b[39mset, child_config)\n\u001b[0;32m   1623\u001b[0m     output \u001b[38;5;241m=\u001b[39m cast(\n\u001b[0;32m   1624\u001b[0m         Output,\n\u001b[1;32m-> 1625\u001b[0m         \u001b[43mcontext\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mrun\u001b[49m\u001b[43m(\u001b[49m\n\u001b[0;32m   1626\u001b[0m \u001b[43m            \u001b[49m\u001b[43mcall_func_with_variable_args\u001b[49m\u001b[43m,\u001b[49m\u001b[43m  \u001b[49m\u001b[38;5;66;43;03m# type: ignore[arg-type]\u001b[39;49;00m\n\u001b[0;32m   1627\u001b[0m \u001b[43m            \u001b[49m\u001b[43mfunc\u001b[49m\u001b[43m,\u001b[49m\u001b[43m  \u001b[49m\u001b[38;5;66;43;03m# type: ignore[arg-type]\u001b[39;49;00m\n\u001b[0;32m   1628\u001b[0m \u001b[43m            \u001b[49m\u001b[38;5;28;43minput\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m  \u001b[49m\u001b[38;5;66;43;03m# type: ignore[arg-type]\u001b[39;49;00m\n\u001b[0;32m   1629\u001b[0m \u001b[43m            \u001b[49m\u001b[43mconfig\u001b[49m\u001b[43m,\u001b[49m\n\u001b[0;32m   1630\u001b[0m \u001b[43m            \u001b[49m\u001b[43mrun_manager\u001b[49m\u001b[43m,\u001b[49m\n\u001b[0;32m   1631\u001b[0m \u001b[43m            \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m,\u001b[49m\n\u001b[0;32m   1632\u001b[0m \u001b[43m        \u001b[49m\u001b[43m)\u001b[49m,\n\u001b[0;32m   1633\u001b[0m     )\n\u001b[0;32m   1634\u001b[0m \u001b[38;5;28;01mexcept\u001b[39;00m \u001b[38;5;167;01mBaseException\u001b[39;00m \u001b[38;5;28;01mas\u001b[39;00m e:\n\u001b[0;32m   1635\u001b[0m     run_manager\u001b[38;5;241m.\u001b[39mon_chain_error(e)\n",
-      "File \u001b[1;32mc:\\Users\\Dr-Lo\\miniconda3\\envs\\ML11\\Lib\\site-packages\\langchain_core\\runnables\\config.py:347\u001b[0m, in \u001b[0;36mcall_func_with_variable_args\u001b[1;34m(func, input, config, run_manager, **kwargs)\u001b[0m\n\u001b[0;32m    345\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m run_manager \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m \u001b[38;5;129;01mand\u001b[39;00m accepts_run_manager(func):\n\u001b[0;32m    346\u001b[0m     kwargs[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mrun_manager\u001b[39m\u001b[38;5;124m\"\u001b[39m] \u001b[38;5;241m=\u001b[39m run_manager\n\u001b[1;32m--> 347\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mfunc\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;28;43minput\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n",
-      "File \u001b[1;32mc:\\Users\\Dr-Lo\\miniconda3\\envs\\ML11\\Lib\\site-packages\\langserve\\client.py:343\u001b[0m, in \u001b[0;36mRemoteRunnable._invoke\u001b[1;34m(self, input, run_manager, config, **kwargs)\u001b[0m\n\u001b[0;32m    334\u001b[0m \u001b[38;5;250m\u001b[39m\u001b[38;5;124;03m\"\"\"Invoke the runnable with the given input and config.\"\"\"\u001b[39;00m\n\u001b[0;32m    335\u001b[0m response \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39msync_client\u001b[38;5;241m.\u001b[39mpost(\n\u001b[0;32m    336\u001b[0m     \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m/invoke\u001b[39m\u001b[38;5;124m\"\u001b[39m,\n\u001b[0;32m    337\u001b[0m     json\u001b[38;5;241m=\u001b[39m{\n\u001b[1;32m   (...)\u001b[0m\n\u001b[0;32m    341\u001b[0m     },\n\u001b[0;32m    342\u001b[0m )\n\u001b[1;32m--> 343\u001b[0m output, callback_events \u001b[38;5;241m=\u001b[39m \u001b[43m_decode_response\u001b[49m\u001b[43m(\u001b[49m\n\u001b[0;32m    344\u001b[0m \u001b[43m    \u001b[49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_lc_serializer\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mresponse\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mis_batch\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;28;43;01mFalse\u001b[39;49;00m\n\u001b[0;32m    345\u001b[0m \u001b[43m\u001b[49m\u001b[43m)\u001b[49m\n\u001b[0;32m    347\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_use_server_callback_events \u001b[38;5;129;01mand\u001b[39;00m callback_events:\n\u001b[0;32m    348\u001b[0m     handle_callbacks(run_manager, callback_events)\n",
-      "File \u001b[1;32mc:\\Users\\Dr-Lo\\miniconda3\\envs\\ML11\\Lib\\site-packages\\langserve\\client.py:230\u001b[0m, in \u001b[0;36m_decode_response\u001b[1;34m(serializer, response, is_batch)\u001b[0m\n\u001b[0;32m    223\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21m_decode_response\u001b[39m(\n\u001b[0;32m    224\u001b[0m     serializer: Serializer,\n\u001b[0;32m    225\u001b[0m     response: httpx\u001b[38;5;241m.\u001b[39mResponse,\n\u001b[0;32m    226\u001b[0m     \u001b[38;5;241m*\u001b[39m,\n\u001b[0;32m    227\u001b[0m     is_batch: \u001b[38;5;28mbool\u001b[39m \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;01mFalse\u001b[39;00m,\n\u001b[0;32m    228\u001b[0m ) \u001b[38;5;241m-\u001b[39m\u001b[38;5;241m>\u001b[39m Tuple[Any, Union[List[CallbackEventDict], List[List[CallbackEventDict]]]]:\n\u001b[0;32m    229\u001b[0m \u001b[38;5;250m    \u001b[39m\u001b[38;5;124;03m\"\"\"Decode the response.\"\"\"\u001b[39;00m\n\u001b[1;32m--> 230\u001b[0m     \u001b[43m_raise_for_status\u001b[49m\u001b[43m(\u001b[49m\u001b[43mresponse\u001b[49m\u001b[43m)\u001b[49m\n\u001b[0;32m    231\u001b[0m     obj \u001b[38;5;241m=\u001b[39m response\u001b[38;5;241m.\u001b[39mjson()\n\u001b[0;32m    232\u001b[0m     \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28misinstance\u001b[39m(obj, \u001b[38;5;28mdict\u001b[39m):\n",
-      "File \u001b[1;32mc:\\Users\\Dr-Lo\\miniconda3\\envs\\ML11\\Lib\\site-packages\\langserve\\client.py:165\u001b[0m, in \u001b[0;36m_raise_for_status\u001b[1;34m(response)\u001b[0m\n\u001b[0;32m    162\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m e\u001b[38;5;241m.\u001b[39mresponse\u001b[38;5;241m.\u001b[39mtext:\n\u001b[0;32m    163\u001b[0m     message \u001b[38;5;241m+\u001b[39m\u001b[38;5;241m=\u001b[39m \u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m for \u001b[39m\u001b[38;5;132;01m{\u001b[39;00me\u001b[38;5;241m.\u001b[39mresponse\u001b[38;5;241m.\u001b[39mtext\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m\"\u001b[39m\n\u001b[1;32m--> 165\u001b[0m \u001b[38;5;28;01mraise\u001b[39;00m httpx\u001b[38;5;241m.\u001b[39mHTTPStatusError(\n\u001b[0;32m    166\u001b[0m     message\u001b[38;5;241m=\u001b[39mmessage,\n\u001b[0;32m    167\u001b[0m     request\u001b[38;5;241m=\u001b[39m_sanitize_request(e\u001b[38;5;241m.\u001b[39mrequest),\n\u001b[0;32m    168\u001b[0m     response\u001b[38;5;241m=\u001b[39me\u001b[38;5;241m.\u001b[39mresponse,\n\u001b[0;32m    169\u001b[0m )\n",
-      "\u001b[1;31mHTTPStatusError\u001b[0m: Server error '500 Internal Server Error' for url 'http://localhost:8000/invoke'\nFor more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/500 for Internal Server Error"
-     ]
-    }
-   ],
-   "source": [
-    "remote_runnable.invoke(dict(input = 'Do you know the tragedy of Darth Plagueis the Wise?', chat_history = []))\n",
-    "# remote_runnable.invoke(dict(input = 'What power source did the Death Star use?', chat_history = []))"
-   ]
-  }
- ],
- "metadata": {
-  "kernelspec": {
-   "display_name": "ML11",
-   "language": "python",
-   "name": "python3"
-  },
-  "language_info": {
-   "codemirror_mode": {
-    "name": "ipython",
-    "version": 3
-   },
-   "file_extension": ".py",
-   "mimetype": "text/x-python",
-   "name": "python",
-   "nbconvert_exporter": "python",
-   "pygments_lexer": "ipython3",
-   "version": "3.11.7"
-  }
- },
- "nbformat": 4,
- "nbformat_minor": 2
-}

explore.ipynb

@@ -61,6 +61,7 @@
     "\n",
     "import os\n",
     "import shutil\n",
+    "from pathlib import Path\n",
     "import re\n",
     "\n",
     "import dotenv\n",
@@ -91,7 +92,8 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "The script database contains 1251 chunks, with mean length of 895 characters\n"
+      "Deleting the previous database and creating a new one (because otherwise content is duplicated in the db every time this block is run)\n",
+      "The script database contains 1260 chunks, with mean length of 892 characters\n"
      ]
     }
    ],
@@ -106,7 +108,7 @@
     "\n",
     "REGENERATE_SCRIPT_DATABASE = False\n",
     "\n",
-    "if (db_exists := os.path.exists(db_dir := r'scripts\\db')):\n",
+    "if (db_exists := os.path.exists(db_dir := str(Path('scripts') / 'db')):\n",
     "    if REGENERATE_SCRIPT_DATABASE:\n",
     "        print('Deleting the previous database and creating a new one (because otherwise content is duplicated in the db every time this block is run)')\n",
     "        shutil.rmtree(db_dir)\n",
@@ -114,7 +116,7 @@
     "\n",
     "if not db_exists or (db_exists and REGENERATE_SCRIPT_DATABASE): # Unfortunate disjoining of the two conditional blocks\n",
     "    scripts = DirectoryLoader('scripts', glob = '*.txt', loader_cls = TextLoader).load()\n",
-    "    for s in scripts: s.page_content = re.sub(r'\\t+|[ ]{2,}', '', s.page_content) # Spacing to centre text noise\n",
+    "    for s in scripts: s.page_content = re.sub(r'^[\\t ]+', '', s.page_content, flags = re.MULTILINE) # Spacing to centre text noise\n",
     "\n",
     "    script_chunks = RecursiveCharacterTextSplitter(chunk_size = 1000, chunk_overlap = 200, separators = ['\\n\\n\\n', '\\n\\n', '\\n']).split_documents(scripts)\n",
     "        # Why not some overlap for extra context just in case?\n",
@@ -588,7 +590,7 @@
    "source": [
     "REGENERATE_WOOKIEEPEDIA_DATABASE = False\n",
     "\n",
-    "if (db_exists := os.path.exists(db_dir := r'wookieepedia\\db')):\n",
+    "if (db_exists := os.path.exists(db_dir := str(Path('wookieepedia') / 'db'))):\n",
     "    if REGENERATE_WOOKIEEPEDIA_DATABASE:\n",
     "        print('Deleting the previous database and creating a new one (because otherwise content is duplicated in the db every time this block is run)')\n",
     "        shutil.rmtree(db_dir)\n",
@@ -732,9 +734,164 @@
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 7,
    "metadata": {},
-   "outputs": [],
+   "outputs": [
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "da3292f4ae9e4d6c912f226b91a4412e",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "modules.json:   0%|          | 0.00/349 [00:00<?, ?B/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "2f3dcddcf3ce4fbbac71999124541281",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "config_sentence_transformers.json:   0%|          | 0.00/116 [00:00<?, ?B/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "70c10212924a487d878d8f87c43bdd90",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "README.md:   0%|          | 0.00/10.7k [00:00<?, ?B/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "8d04da9721ec4e86b01068d16a8588fa",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "sentence_bert_config.json:   0%|          | 0.00/53.0 [00:00<?, ?B/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "8dfd7c22b7a44ea9a16fbd6573635186",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "config.json:   0%|          | 0.00/612 [00:00<?, ?B/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "f5fabe250679451da00677c57d08abf1",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "model.safetensors:   0%|          | 0.00/90.9M [00:00<?, ?B/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "0e978ece6930478b9e7106020bfa1ebf",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "tokenizer_config.json:   0%|          | 0.00/350 [00:00<?, ?B/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "db27fca31ec844e888ed7f6f6aecbe2b",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "vocab.txt:   0%|          | 0.00/232k [00:00<?, ?B/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "d1570b2dd7d7413ba72777799fa7e2a5",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "tokenizer.json:   0%|          | 0.00/466k [00:00<?, ?B/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "2ec46385e4f44907991e3f498bc4b77d",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "special_tokens_map.json:   0%|          | 0.00/112 [00:00<?, ?B/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "0f1dddc6b2ee4ebf8eb37e356b9c3e41",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "1_Pooling/config.json:   0%|          | 0.00/190 [00:00<?, ?B/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
    "source": [
     "# from sentence_transformers import SentenceTransformer\n",
     "\n",
@@ -1196,7 +1353,7 @@
    "source": [
     "compound_chain = create_retrieval_chain(compound_retriever, document_chain)\n",
     "\n",
-    "compound_chain.invoke(dict(input = 'Do you know the tragedy of Darth Plagueis the Wise?'))"
+    "# compound_chain.invoke(dict(input = 'Do you know the tragedy of Darth Plagueis the Wise?'))"
    ]
   }
  ],

ingredients.py

@@ -29,6 +29,7 @@ from bs4 import BeautifulSoup
 
 import os
 import shutil
+from pathlib import Path
 import re
 
 import dotenv
@@ -39,46 +40,22 @@ dotenv.load_dotenv()
 
 ## Vector stores
 
-script_db = Chroma(embedding_function = SentenceTransformerEmbeddings(model_name = 'all-MiniLM-L6-v2'), persist_directory = r'scripts\db')
-woo_db = Chroma(embedding_function = SentenceTransformerEmbeddings(model_name = 'all-MiniLM-L6-v2'), persist_directory = r'wookieepedia\db')
+# Non-persistent; build from documents
 
+# scripts = DirectoryLoader('scripts', glob = '*.txt', loader_cls = TextLoader).load()
+# for s in scripts: s.page_content = re.sub(r'^[\t ]+', '', s.page_content, flags = re.MULTILINE)  # Spacing to centre text noise
+# script_chunks = RecursiveCharacterTextSplitter(chunk_size = 1000, chunk_overlap = 200, separators = ['\n\n\n', '\n\n', '\n']).split_documents(scripts)
+# script_db = Chroma.from_documents(script_chunks, SentenceTransformerEmbeddings(model_name = 'all-MiniLM-L6-v2'))
 
+# pages = DirectoryLoader('wookieepedia', glob = '*.txt', loader_cls = TextLoader).load()
+# page_chunks = RecursiveCharacterTextSplitter(chunk_size = 1000, chunk_overlap = 200, separators = ['\n\n\n', '\n\n', '\n']).split_documents(pages)
+# woo_db = Chroma.from_documents(page_chunks, SentenceTransformerEmbeddings(model_name = 'all-MiniLM-L6-v2'))
 
-## Wookieepedia functions
-
-def first_wookieepedia_result(query: str) -> str:
-    '''Get the url of the first result when searching Wookieepedia for a query
-    (best for simple names as queries, ideally generated by the llm for something like
-    "Produce a input consisting of the name of the most important element in the query so that its article can be looked up")
-    '''
-    search_results = requests.get(f'https://starwars.fandom.com/wiki/Special:Search?query={"+".join(query.split(" "))}')
-    soup = BeautifulSoup(search_results.content, 'html.parser')
-    first_res = soup.find('a', class_ = 'unified-search__result__link')
-    return first_res['href']
-
-
-def get_new_wookieepedia_chunks(query: str, previous_sources: set[str]) -> list[Document]:
-    '''Retrieve and return chunks of the content of the first result of query on Wookieepedia, then return the closest matches for.
-    '''
-    url = first_wookieepedia_result(query)
-
-    if url in previous_sources: return []
-    else:
-        doc = WebBaseLoader(url).load()[0] # Only one url passed in => only one Document out; no need to assert
-    
-        # There probably is a very long preamble before the real content, however, if more than one gap then ignore and proceed with full document
-        trimmed = parts[1] if len(parts := doc.page_content.split('\n\n\n\n\n\n\n\n\n\n\n\n\n\n \xa0 \xa0')) == 2 else doc.page_content
-        doc.page_content = re.sub(r'[\n\t]{2,}', '\n', trimmed) # And remove excessive spacing
-
-        return RecursiveCharacterTextSplitter(chunk_size = 800, chunk_overlap = 100).split_documents([doc])
 
-def get_wookieepedia_context(original_query: str, simple_query: str, wdb: Chroma) -> list[Document]:
-    try:
-        new_chunks = get_new_wookieepedia_chunks(simple_query, previous_sources = set(md.get('source') for md in wdb.get()['metadatas']))
-        if new_chunks: wdb.add_documents(new_chunks)
-    except: return []
+# # Load pre-built persistent ones
 
-    return wdb.similarity_search(original_query, k = 10)
+script_db = Chroma(embedding_function = SentenceTransformerEmbeddings(model_name = 'all-MiniLM-L6-v2'), persist_directory = str(Path('scripts') / 'db'))
+woo_db = Chroma(embedding_function = SentenceTransformerEmbeddings(model_name = 'all-MiniLM-L6-v2'), persist_directory = str(Path('wookieepedia') / 'db'))
 
 
 
@@ -105,7 +82,6 @@ document_prompt = ChatPromptTemplate.from_messages([
 ])
 document_chain = create_stuff_documents_chain(llm, document_prompt)
 
-
 script_retriever_prompt = ChatPromptTemplate.from_messages([
     MessagesPlaceholder(variable_name = 'chat_history'),
     ('user', '{input}'),
@@ -114,7 +90,6 @@ script_retriever_prompt = ChatPromptTemplate.from_messages([
 ])
 script_retriever_chain = create_history_aware_retriever(llm, script_db.as_retriever(), script_retriever_prompt) # Essentially just: prompt | llm | StrOutputParser() | retriever
 
-
 woo_retriever_prompt = ChatPromptTemplate.from_messages([
     MessagesPlaceholder(variable_name = 'chat_history'),
     ('user', '{input}'),
@@ -122,7 +97,6 @@ woo_retriever_prompt = ChatPromptTemplate.from_messages([
 ])
 woo_retriever_chain = create_history_aware_retriever(llm, woo_db.as_retriever(), woo_retriever_prompt) # Essentially just: prompt | llm | StrOutputParser() | retriever
 
-
 # full_chain = create_retrieval_chain(script_retriever_chain, document_chain)
 full_chain = create_retrieval_chain(woo_retriever_chain, document_chain)
 
@@ -146,6 +120,7 @@ script_tool = create_retriever_tool(
     '''Search the Star Wars film scripts. This tool should be the first choice for Star Wars related questions.
     Queries passed to this tool should be lists of keywords likely to be in dialogue or scene descriptions, and should not include film titles.'''
 )
+
 woo_tool = create_retriever_tool(
     woo_db.as_retriever(search_kwargs = dict(k = 4)),
     'search_wookieepedia',
@@ -166,6 +141,7 @@ agent_prompt = ChatPromptTemplate.from_messages([
     ('placeholder', '{agent_scratchpad}') # Required for chat history and the agent's intermediate processing values
 ])
 agent = create_tool_calling_agent(llm, tools, agent_prompt)
+
 agent_executor = AgentExecutor(agent = agent, tools = tools, verbose = True)
 
 
@@ -204,3 +180,49 @@ def compound_retriever(question):
 compound_chain = create_retrieval_chain(compound_retriever, document_chain)
 
 
+
+## Wookieepedia functions
+
+def first_wookieepedia_result(query: str) -> str:
+    '''Get the url of the first result when searching Wookieepedia for a query
+    (best for simple names as queries, ideally generated by the llm for something like
+    "Produce a input consisting of the name of the most important element in the query so that its article can be looked up")
+    '''
+    search_results = requests.get(f'https://starwars.fandom.com/wiki/Special:Search?query={"+".join(query.split(" "))}')
+    soup = BeautifulSoup(search_results.content, 'html.parser')
+    first_res = soup.find('a', class_ = 'unified-search__result__link')
+    return first_res['href']
+
+
+def get_wookieepedia_page_content(query: str, previous_sources: set[str]) -> Document | None:
+    '''Return cleaned content from a Wookieepedia page provided it was not already sourced
+    '''
+    url = first_wookieepedia_result(query)
+
+    if url in previous_sources: return None
+    else:
+        response = requests.get(url)
+        soup = BeautifulSoup(response.content, 'html.parser')
+        doc = soup.find('div', id = 'content').get_text()
+
+        # Cleaning
+        doc = doc.split('\n\n\n\n\n\n\n\n\n\n\n\n\n\n')[-1] # The (multiple) preambles are separated by these many newlines; no harm done if not present
+        doc = re.sub('\[\d*\]', '', doc) # References (and section title's "[]" suffixes) are noise
+        doc = doc.split('\nAppearances\n')[0] # Keep only content before these sections
+        doc = doc.split('\nSources\n')[0] # Technically no need to check this if successfully cut on appearances, but no harm done
+        doc = re.sub('Contents\n\n(?:[\d\.]+ [^\n]+\n+)+', '', doc) # Remove table of contents
+
+        return Document(page_content = doc, metadata = dict(source = url))
+
+def get_wookieepedia_context(original_query: str, simple_query: str, wdb: Chroma) -> list[Document]:
+    try:
+        doc = get_wookieepedia_page_content(simple_query, previous_sources = set(md.get('source') for md in wdb.get()['metadatas']))
+        if doc is not None:
+            new_chunks = RecursiveCharacterTextSplitter(chunk_size = 1000, chunk_overlap = 200).split_documents([doc])
+            wdb.add_documents(new_chunks)
+            print(f"Added new chunks (for '{simple_query}' -> {doc['metadata']['source']}) to the Wookieepedia database.")
+    except: return []
+
+    return wdb.similarity_search(original_query, k = 10)
+
+

langserve_app.py

@@ -0,0 +1,71 @@
+# Chains
+from langchain_core.pydantic_v1 import BaseModel, Field
+
+# To serve the app
+from fastapi import FastAPI
+from langchain_core.messages import BaseMessage
+from langserve import add_routes, CustomUserType
+
+import dotenv
+dotenv.load_dotenv()
+
+from ingredients import script_db, woo_db, full_chain, compound_chain, agent_executor
+
+
+
+## Type specifications (with unusual class-scope fields)
+
+class StrInput(BaseModel):
+    input: str
+
+class Input(BaseModel):
+    input: str
+    chat_history: list[BaseMessage] = Field(
+        ...,
+        extra = dict(widget = dict(type = 'chat', input = 'location')),
+    )
+
+class Output(BaseModel):
+    output: str
+
+
+
+## App definition
+# NOTE: The chat playground type has a web page issue (flashes and becomes white, hence non-interactable; this was supposedly solved in an issue late last year)
+
+app = FastAPI(
+    title = 'Star Wars Expert',
+    version = '1.0',
+    description = 'A Star Wars expert chatbot',
+)
+
+
+# Basic retriever versions
+
+# add_routes(app, script_db.as_retriever())
+# add_routes(app, woo_db.as_retriever())
+
+
+# History-aware retriever version
+# add_routes(app, full_chain.with_types(input_type = StrInput, output_type = Output), playground_type = 'default')
+
+
+# Agent version
+
+# add_routes(app, agent_executor, playground_type = 'chat')
+# add_routes(app, agent_executor.with_types(input_type = StrInput, output_type = Output))
+
+
+# Non-agent chain-logic version
+
+add_routes(app, compound_chain.with_types(input_type = StrInput))
+# add_routes(app, compound_chain.with_types(input_type = Input), playground_type = 'chat')
+
+
+
+if __name__ == '__main__':
+    import uvicorn
+
+    uvicorn.run(app, host = 'localhost', port = 8000)
+
+

requirements.txt

@@ -1,12 +1,14 @@
 beautifulsoup4>=4.12.3
-fastapi>=0.110.2
 langchain>=0.1.16
 langchain_community>=0.0.34
 langchain_core>=0.1.45
 langchain_openai>=0.1.3
-langserve>=0.1.0
 sentence-transformers>=2.7.0
 python-dotenv>=1.0.1
 Requests>=2.31.0
-uvicorn>=0.29.0
 chromadb>=0.4.24
+gradio>=4.27.0
+
+# fastapi>=0.110.2
+# langserve>=0.1.0
+# uvicorn>=0.29.0

scripts/db/{ee161592-5f4b-453f-af5c-10f9ed7a57a0 → 9a364cee-0b3d-4004-86d6-7bac58ecc92c}/data_level0.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:698c3f0c09428ffae03ea896c578be3cd69b1b3019d69354e05383663eaf26c3
+oid sha256:b4907f424ad43eafe0706f355d0345cd82ee7798a135f9995233ca7e42bc937d
 size 1676000

scripts/db/{ee161592-5f4b-453f-af5c-10f9ed7a57a0 → 9a364cee-0b3d-4004-86d6-7bac58ecc92c}/header.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:29f241a1a6775b5338a81594c5c0ce61c670b325cacc6d674250c7ae3c6b9682
+oid sha256:bc26d853c2df1dbd2161639e83f53b98808b632c106c529341d0b31f3a0b3e2a
 size 100

scripts/db/{ee161592-5f4b-453f-af5c-10f9ed7a57a0 → 9a364cee-0b3d-4004-86d6-7bac58ecc92c}/index_metadata.pickle

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5e7e2b8a066a226a9961812ad1ef171838def7f221c4f72639b0c1e9ff724e46
+oid sha256:7824365e98ba10f9c2112739ddbb7d87839f7c27a662e21ee2fe819677ce0e0e
 size 55974

scripts/db/{ee161592-5f4b-453f-af5c-10f9ed7a57a0 → 9a364cee-0b3d-4004-86d6-7bac58ecc92c}/length.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9e6471f6ab66dac4d5f196ef41fe5461b42784fefbaa4b5c456768ffabbbd3bc
+oid sha256:ac2283f2e028a485f5564c59ed1722b4e7c01c7dad86578d7572e39b4af15c74
 size 4000

scripts/db/{ee161592-5f4b-453f-af5c-10f9ed7a57a0 → 9a364cee-0b3d-4004-86d6-7bac58ecc92c}/link_lists.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cd75fcf8d5b138005794e3d58997ed37fc05c605f60f77b61399a03a2659333f
+oid sha256:f3ace8c9f24593fdd1d5ea5618e39cac109d9663f598c4e0e5ab668534f3b62e
 size 8624

scripts/db/chroma.sqlite3

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e0cc8b50d0ea411aef4effba6a6cdfad5257875b12e939e334531b534abc5aab
-size 12161024
+oid sha256:08c0800eb8542363f5030d5fa2c1807251c5827bd9a8a1888f923d85cc564951
+size 12173312

testing.ipynb

@@ -0,0 +1,132 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Testing of different versions"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Local environment"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from ingredients import script_db, woo_db, full_chain, compound_chain, agent_executor"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "compound_chain.invoke(dict(input = 'What power source did the Death Star use?', chat_history = []))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Gradio app"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Loaded as API: http://127.0.0.1:7860/ ✔\n"
+     ]
+    }
+   ],
+   "source": [
+    "from gradio_client import Client\n",
+    "\n",
+    "# client = Client('T-Flet/Star-Wars-Expert')\n",
+    "client = Client('http://127.0.0.1:7860')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "'Hello! How can I help you with Star Wars today?'"
+      ]
+     },
+     "execution_count": 4,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "client.predict(message = 'Do you know the tragedy of Darth Plagueis the Wise?', api_name = '/chat')"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## LangServe app"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from langserve import RemoteRunnable\n",
+    "\n",
+    "remote_runnable = RemoteRunnable('http://localhost:8000/')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# remote_runnable.invoke(dict(input = 'Do you know the tragedy of Darth Plagueis the Wise?', chat_history = []))\n",
+    "remote_runnable.invoke(dict(input = 'What power source did the Death Star use?', chat_history = []))"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "ML11",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.11.7"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}