S2 demo

cookbook/integrations/singlestore/auto_rag/README.md

@@ -0,0 +1,76 @@
+# Autonomous RAG with SingleStore and Phidata
+
+This cookbook shows how to build an Autonomous RAG application with SingleStore and Phidata.
+
+We'll use the following LLMs:
+- GPT-4 by OpenAI (needs an API key)
+- Hermes2 running locally using Ollama (no API key needed)
+- Claude3 by Anthropic (needs an API key)
+
+> Note: Fork and clone this repository if needed
+
+### 1. Create a virtual environment
+
+```shell
+python3 -m venv ~/.venvs/aienv
+source ~/.venvs/aienv/bin/activate
+```
+
+### 2. Install libraries
+
+```shell
+pip install -r cookbook/integrations/singlestore/auto_rag/requirements.txt
+```
+
+### 3. To use a local model, install and run Ollama
+
+- [Install](https://github.com/ollama/ollama?tab=readme-ov-file#macos) ollama
+
+- Run you embedding model
+
+```shell
+ollama run nomic-embed-text
+```
+
+- Run your chat model
+
+```shell
+ollama run adrienbrault/nous-hermes2pro:Q8_0 'Hey!'
+```
+
+### 4. Provide credentials
+
+- For SingleStore
+
+> Note: Please make sure you provide a certificate file for SSL connection [Read more](https://docs.singlestore.com/cloud/connect-to-your-workspace/connect-with-mysql/connect-with-mysql-client/connect-to-singlestore-helios-using-tls-ssl/)
+
+```shell
+export SINGLESTORE_HOST="host"
+export SINGLESTORE_PORT="3333"
+export SINGLESTORE_USERNAME="user"
+export SINGLESTORE_PASSWORD="password"
+export SINGLESTORE_DATABASE="db"
+export SINGLESTORE_SSL_CA=".certs/singlestore_bundle.pem"
+```
+
+- To use OpenAI GPT-4, provide your OPENAI_API_KEY
+
+```shell
+export OPENAI_API_KEY="sk-..."
+```
+
+- To use Anthropic Claude3, provide your ANTHROPIC_API_KEY
+
+```shell
+export ANTHROPIC_API_KEY="sk-..."
+```
+
+### 6. Run Streamlit application
+
+```shell
+streamlit run cookbook/integrations/singlestore/auto_rag/app.py
+```
+
+### 7. Message us on [discord](https://discord.gg/4MtYHHrgA8) if you have any questions
+
+### 8. Star ⭐️ the project if you like it.

cookbook/integrations/singlestore/auto_rag/app.py

@@ -0,0 +1,155 @@
+from typing import List
+
+import streamlit as st
+from phi.assistant import Assistant
+from phi.document import Document
+from phi.document.reader.website import WebsiteReader
+from phi.tools.streamlit.components import (
+    check_password,
+    reload_button_sidebar,
+    get_username_sidebar,
+)
+
+from assistant import get_local_rag_assistant  # Adjust the import statement as needed
+from logging import getLogger
+
+logger = getLogger(__name__)
+
+st.set_page_config(
+    page_title="Local RAG with Web Scraping \n Using SingleStore as a backend databse",
+    page_icon=":orange_heart:",
+)
+st.title("Local RAG with Web Scraping")
+st.markdown("##### :orange_heart: Built using [phidata](https://github.com/phidatahq/phidata)")
+def restart_assistant():
+    st.session_state["web_assistant"] = None
+    st.session_state["web_assistant_run_id"] = None
+    st.session_state["url_scrape_key"] += 1
+    st.rerun()
+
+
+def main() -> None:
+    # Get username
+    username = get_username_sidebar()
+    if username:
+        st.sidebar.info(f":technologist: User: {username}")
+    else:
+        st.write(":technologist: Please enter a username")
+        return
+
+    # Get model
+    local_rag_model = st.sidebar.selectbox("Select Model", options=["openhermes", "codellama"])
+    # Set assistant_type in session state
+    if "local_rag_model" not in st.session_state:
+        st.session_state["local_rag_model"] = local_rag_model
+    # Restart the assistant if assistant_type has changed
+    elif st.session_state["local_rag_model"] != local_rag_model:
+        st.session_state["local_rag_model"] = local_rag_model
+        restart_assistant()
+
+    # Get the assistant
+    web_assistant: Assistant
+    if "web_assistant" not in st.session_state or st.session_state["web_assistant"] is None:
+        logger.info("---*--- Creating Web Assistant ---*---")
+        web_assistant = get_local_rag_assistant(
+            model=local_rag_model,
+            user_id=username,
+            debug_mode=True,
+        )
+        st.session_state["web_assistant"] = web_assistant
+    else:
+        web_assistant = st.session_state["web_assistant"]
+
+    # Create assistant run (i.e. log to database) and save run_id in session state
+    try:
+        st.session_state["web_assistant_run_id"] = web_assistant.create_run()
+    except Exception:
+        st.warning("Could not create assistant, is the database running?")
+        return
+
+    # Load existing messages
+    assistant_chat_history = web_assistant.memory.get_chat_history()
+    if len(assistant_chat_history) > 0:
+        logger.debug("Loading chat history")
+        st.session_state["messages"] = assistant_chat_history
+    else:
+        logger.debug("No chat history found")
+        st.session_state["messages"] = [{"role": "web_assistant", "content": "Ask me anything..."}]
+
+    # Prompt for user input
+    if prompt := st.chat_input():
+        st.session_state["messages"].append({"role": "user", "content": prompt})
+
+    # Display existing chat messages
+    for message in st.session_state["messages"]:
+        if message["role"] == "system":
+            continue
+        with st.chat_message(message["role"]):
+            st.write(message["content"])
+
+    # If last message is from a user, generate a new response
+    last_message = st.session_state["messages"][-1]
+    if last_message.get("role") == "user":
+        question = last_message["content"]
+        with st.chat_message("web_assistant"):
+            response = ""
+            resp_container = st.empty()
+            for delta in web_assistant.run(question):
+                response += delta  # type: ignore
+                resp_container.markdown(response)
+
+            st.session_state["messages"].append({"role": "web_assistant", "content": response})
+
+    if st.sidebar.button("New Run"):
+        restart_assistant()
+
+    if st.sidebar.button("Auto Rename"):
+        web_assistant.auto_rename_run()
+
+    # Upload Web Content
+    if web_assistant.knowledge_base:
+        if "url_scrape_key" not in st.session_state:
+            st.session_state["url_scrape_key"] = 0
+
+        scraped_url = st.sidebar.text_input(
+            "Input URL",
+            type="default",
+            key=st.session_state["url_scrape_key"]
+        )
+        append_button = st.sidebar.button("Search URL")
+        if append_button:
+            if scraped_url is not None:
+                alert = st.sidebar.info("Processing URLs...", icon="🧠")
+                if f"{scraped_url}_scraped" not in st.session_state:
+                    scraper = WebsiteReader()
+                    web_documents: List[Document] = scraper.read(scraped_url)
+                    if web_documents:
+                        web_assistant.knowledge_base.load_documents(web_documents, upsert=True, skip_existing=True)
+                    else:
+                        st.sidebar.error("Could not read Website")
+                    st.session_state[f"{scraped_url}_uploaded"] = True
+                alert.empty()
+
+    if web_assistant.storage:
+        web_assistant_run_ids: List[str] = web_assistant.storage.get_all_run_ids(user_id=username)
+        new_web_assistant_run_id = st.sidebar.selectbox("Run ID", options=web_assistant_run_ids)
+        if st.session_state["web_assistant_run_id"] != new_web_assistant_run_id:
+            logger.info(f"---*--- Loading run: {new_web_assistant_run_id} ---*---")
+            st.session_state["web_assistant"] = get_local_rag_assistant(
+                model=local_rag_model,
+                user_id=username,
+                run_id=new_web_assistant_run_id,
+                debug_mode=True,
+            )
+            st.rerun()
+
+    web_assistant_run_name = web_assistant.run_name
+    if web_assistant_run_name:
+        st.sidebar.write(f":thread: {web_assistant_run_name}")
+
+    # Show reload button
+    reload_button_sidebar()
+
+
+if check_password():
+    main()

cookbook/integrations/singlestore/auto_rag/assistant.py

@@ -0,0 +1,58 @@
+from typing import Optional
+
+from phi.assistant import Assistant
+from phi.knowledge import AssistantKnowledge
+from phi.llm.ollama import Ollama
+from phi.embedder.ollama import OllamaEmbedder
+from phi.storage.assistant.singlestore import S2AssistantStorage
+from phi.vectordb.singlestore import S2VectorDb
+
+from resources import config  # type: ignore
+
+# Setup SingleStore connection
+db_url = (
+    f"mysql+pymysql://{config['username']}:{config['password']}@{config['host']}:{config['port']}/{config['database']}?charset=utf8mb4"
+)
+
+local_assistant_storage = S2AssistantStorage(
+    table_name="local_rag_assistant",
+    schema=config["database"],
+    db_url=db_url,
+)
+
+local_assistant_knowledge = AssistantKnowledge(
+    vector_db=S2VectorDb(
+        collection="web_documents_singlestore",
+        schema=config["database"],
+        db_url=db_url,
+        # Assuming OllamaEmbedder or a compatible embedder is used for SingleStore
+        embedder=OllamaEmbedder(model="nomic-embed-text", dimensions=768),
+    ),
+    num_documents=5,
+)
+
+
+def get_local_rag_assistant(
+    model: str = "openhermes",
+    user_id: Optional[str] = None,
+    run_id: Optional[str] = None,
+    debug_mode: bool = False,
+) -> Assistant:
+    """Get a Local URL RAG Assistant with SingleStore backend."""
+    return Assistant(
+        name="local_rag_assistant",
+        run_id=run_id,
+        user_id=user_id,
+        llm=Ollama(model=model),
+        storage=local_assistant_storage,
+        knowledge_base=local_assistant_knowledge,
+        add_chat_history_to_messages=False,
+        add_references_to_prompt=True,
+        num_history_messages=4,
+        markdown=True,
+        debug_mode=debug_mode,
+        description="You are an AI called 'Phi' designed to help users answer questions from a knowledge base.",
+        assistant_data={"assistant_type": "rag"},
+        use_tools=True,
+        show_tool_calls=True,
+    )

cookbook/integrations/singlestore/auto_rag/cli.py

@@ -0,0 +1,39 @@
+import typer
+from phi.assistant import Assistant
+from phi.llm.ollama import Ollama
+from phi.vectordb.singlestore import S2VectorDb
+from phi.embedder.ollama import OllamaEmbedder
+from phi.knowledge.pdf import PDFUrlKnowledgeBase
+from phi.storage.assistant.singlestore import S2AssistantStorage
+from resources import config  # type: ignore
+
+db_url = (
+    f"mysql+pymysql://{config['username']}:{config['password']}@{config['host']}:{config['port']}/{config['database']}"
+)
+storage = S2AssistantStorage(table_name="pdf_assistant", schema=config["database"], db_url=db_url)
+knowledge_base = PDFUrlKnowledgeBase(
+    urls=["https://phi-public.s3.amazonaws.com/recipes/ThaiRecipes.pdf"],
+    vector_db=S2VectorDb(
+        collection="recipes",
+        schema=config["database"],
+        db_url=db_url,
+        embedder=OllamaEmbedder(model="nomic-embed-text", dimensions=768),
+    ),
+)
+# Comment after first run to avoid reloading the knowledge base
+# knowledge_base.load(recreate=False)
+
+
+def local_assistant(model: str = "openhermes", debug: bool = False):
+    print(f"============= Running: {model} =============")
+    Assistant(
+        llm=Ollama(model=model),
+        storage=storage,
+        knowledge_base=knowledge_base,
+        add_references_to_prompt=True,
+        debug_mode=debug,
+    ).cli_app(markdown=True)
+
+
+if __name__ == "__main__":
+    typer.run(local_assistant)

cookbook/integrations/singlestore/auto_rag/requirements.in

@@ -0,0 +1,8 @@
+openai
+streamlit
+pypdf
+pymysql
+sqlalchemy
+phidata
+ollama
+anthropic