quickstart hackathon

aboutphilippe · aboutphilippe · commit fdd07ada8dee · 2024-11-08T12:45:37.000-08:00
diff --git a/examples/llama_quickstart/.env.example b/examples/llama_quickstart/.env.example
@@ -0,0 +1 @@
+TOGETHER_API_KEY=<your-together-api-key>
diff --git a/examples/llama_quickstart/.gitignore b/examples/llama_quickstart/.gitignore
@@ -0,0 +1,3 @@
+.DS_Store
+.env
+poetry.lock
diff --git a/examples/llama_quickstart/README.md b/examples/llama_quickstart/README.md
@@ -0,0 +1,79 @@
+# Quiuckstart Llama Hackathon
+
+Restack AI - Streamlit + FastApi + TogetherAI with LlamaIndex Example
+
+The AI workflow will search hacker news based on a query, crawl each project website, make summaries for the user
+
+## Prerequisites
+
+- Python 3.9 or higher
+- Poetry (for dependency management)
+- Docker (for running the Restack services)
+- Active [Together AI](https://together.ai) account with API key
+
+## Usage
+
+1. Run Restack local engine with Docker:
+
+   ```bash
+   docker run -d --pull always --name studio -p 5233:5233 -p 6233:6233 -p 7233:7233 ghcr.io/restackio/engine:main
+   ```
+
+2. Open the Web UI to see the workflows:
+
+   ```bash
+   http://localhost:5233
+   ```
+
+3. Clone this repository:
+
+   ```bash
+   git clone https://github.com/restackio/examples-python
+   cd examples-python/examples/llama_quickstart
+   ```
+
+4. Install dependencies using Poetry:
+
+   ```bash
+   poetry install
+   ```
+
+5. Set up your environment variables:
+
+   Copy `.env.example` to `.env` and add your Together AI API key:
+
+   ```bash
+   cp .env.example .env
+   # Edit .env and add your TOGETHER_API_KEY
+   ```
+
+6. Run the services:
+
+   ```bash
+   poetry run services
+   ```
+
+   This will start the Restack service with the defined workflows and functions.
+
+7. In a new terminal, run fastapi app:
+
+   ```bash
+   poetry run app
+   ```
+
+8. In a new terminal, run the streamlit frontend
+
+   ```bash
+   poetry run streamlit run frontend.py
+   ```
+
+9. You can test api endpoint without the streamlit UI with:
+
+   ```bash
+   curl -X POST \
+     http://localhost:8000/api/schedule \
+     -H "Content-Type: application/json" \
+     -d '{"query": "AI", "count": 5}'
+   ```
+
+   This will schedule the workflow and return the result.
diff --git a/examples/llama_quickstart/frontend.py b/examples/llama_quickstart/frontend.py
@@ -0,0 +1,52 @@
+import streamlit as st
+import requests
+
+# Set page title and header
+st.title("LLama Hackathon Quickstart")
+st.text("FastAPI, Restack, Together AI, LLamaIndex")
+
+# Create text area for user input with session state
+if "user_input" not in st.session_state:
+    st.session_state.user_input = ""
+
+query = st.text_input("Query HN", key="query", value="ai")
+count = st.number_input("Number of results", key="count", value=5)
+
+# Initialize response history in session state
+if "response_history" not in st.session_state:
+    st.session_state.response_history = []
+
+# Create button to send request
+if st.button("Search HN"):
+    if query:
+        try:
+            with st.spinner('Searching...'):
+                # Make POST request to FastAPI backend
+                response = requests.post(
+                    "http://localhost:8000/api/schedule",
+                    json={"query": query, "count": count}
+                )
+                
+                if response.status_code == 200:
+                    st.success("Response received!")
+                    # Add the new response to history with the original prompt
+                    st.session_state.response_history.append({
+                        "query": query,
+                        "count": count,
+                        "response": response.json()["result"]
+                    })
+                else:
+                    st.error(f"Error: {response.status_code}")
+                    
+        except requests.exceptions.ConnectionError:
+            st.error("Failed to connect to the server. Make sure the FastAPI server is running.")
+    else:
+        st.warning("Please enter a prompt before submitting.")
+
+# Display response history
+if st.session_state.response_history:
+    st.subheader("Response History")
+    for i, item in enumerate(st.session_state.response_history, 1):
+        st.markdown(f"**Query {i}:** {item['query']}")
+        st.markdown(f"**Response {i}:** {item['response']}")
+        st.markdown("---")
diff --git a/examples/llama_quickstart/pyproject.toml b/examples/llama_quickstart/pyproject.toml
@@ -0,0 +1,29 @@
+[tool.poetry]
+name = "streamlit_fastapi_togetherai_llama"
+version = "0.0.1"
+description = "A simple example showcasing streamlit, fastapi, togetherai and llamaindex"
+authors = [
+    "Restack Team <service@restack.io>",
+]
+readme = "README.md"
+packages = [{include = "src"}]
+
+[tool.poetry.dependencies]
+python = "^3.12"
+pydantic = "^2.9.2"
+restack-ai = "^0.0.25"
+fastapi = "^0.115.4"
+llama-index = "^0.11.22"
+llama-index-llms-together = "^0.2.0"
+uvicorn = "^0.32.0"
+python-dotenv = "0.19"
+streamlit = "^1.40.0"
+requests = "^2.32.3"
+
+[build-system]
+requires = ["poetry-core"]
+build-backend = "poetry.core.masonry.api"
+
+[tool.poetry.scripts]
+services = "src.services:run_services"
+app = "src.app:run_app"
diff --git a/examples/llama_quickstart/src/__init__.py b/examples/llama_quickstart/src/__init__.py
diff --git a/examples/llama_quickstart/src/app.py b/examples/llama_quickstart/src/app.py
@@ -0,0 +1,59 @@
+from fastapi import FastAPI, HTTPException
+from fastapi.middleware.cors import CORSMiddleware
+from pydantic import BaseModel
+import time
+from restack_ai import Restack
+import uvicorn
+
+# Define request model
+class QueryRequest(BaseModel):
+    query: str
+    count: int
+
+app = FastAPI()
+
+# Add CORS middleware
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],  # Adjust this in production
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+@app.get("/")
+async def home():
+    return "Welcome to the TogetherAI LlamaIndex FastAPI App!"
+
+@app.post("/api/schedule")
+async def schedule_workflow(request: QueryRequest):
+    try:
+        client = Restack()
+        workflow_id = f"{int(time.time() * 1000)}-llm_complete_workflow"
+        
+        runId = await client.schedule_workflow(
+            workflow_name="hn_workflow",
+            workflow_id=workflow_id,
+            input={"query": request.query, "count": request.count}
+        )
+        print("Scheduled workflow", runId)
+        
+        result = await client.get_workflow_result(
+            workflow_id=workflow_id,
+            run_id=runId
+        )
+        
+        return {
+            "result": result,
+            "workflow_id": workflow_id,
+            "run_id": runId
+        }
+    except Exception as e:
+        raise HTTPException(status_code=400, detail=str(e))
+
+# Remove Flask-specific run code since FastAPI uses uvicorn
+def run_app():
+    uvicorn.run("src.app:app", host="0.0.0.0", port=8000, reload=True)
+
+if __name__ == '__main__':
+    run_app()
diff --git a/examples/llama_quickstart/src/client.py b/examples/llama_quickstart/src/client.py
@@ -0,0 +1,3 @@
+from restack_ai import Restack
+
+client = Restack()
diff --git a/examples/llama_quickstart/src/functions/__init__.py b/examples/llama_quickstart/src/functions/__init__.py
diff --git a/examples/llama_quickstart/src/functions/crawl/__init__.py b/examples/llama_quickstart/src/functions/crawl/__init__.py
diff --git a/examples/llama_quickstart/src/functions/crawl/website.py b/examples/llama_quickstart/src/functions/crawl/website.py
@@ -0,0 +1,25 @@
+from restack_ai.function import function, log
+import requests
+from bs4 import BeautifulSoup
+
+@function.defn(name="crawl_website")
+async def crawl_website(url):
+    try:
+        # Send a GET request to the URL
+        response = requests.get(url)
+        response.raise_for_status()  # Raise an error for bad responses
+
+        # Parse the content with BeautifulSoup
+        soup = BeautifulSoup(response.content, 'html.parser')
+
+        # Extract the text content from the page
+        content = soup.get_text()
+
+        log.info("crawl_website", extra={"content": content})
+
+        return content
+
+    except requests.exceptions.RequestException as e:
+        # Handle any exceptions that occur during the request
+        print(f"An error occurred: {e}")
+        return None
diff --git a/examples/llama_quickstart/src/functions/hn/__init__.py b/examples/llama_quickstart/src/functions/hn/__init__.py
diff --git a/examples/llama_quickstart/src/functions/hn/schema.py b/examples/llama_quickstart/src/functions/hn/schema.py
@@ -0,0 +1,5 @@
+from pydantic import BaseModel, Field
+
+class HnSearchInput(BaseModel):
+    query: str = Field(default=None, description="The query for search")
+    count: int = Field(default=5, description="The number of results to return")
diff --git a/examples/llama_quickstart/src/functions/hn/search.py b/examples/llama_quickstart/src/functions/hn/search.py
@@ -0,0 +1,18 @@
+import requests
+from restack_ai.function import function, log
+from src.functions.hn.schema import HnSearchInput
+
+@function.defn(name="hn_search")
+async def hn_search(input: HnSearchInput):
+    try:
+        # Fetch the latest stories IDs
+        response = requests.get(
+            f"https://hn.algolia.com/api/v1/search_by_date?tags=show_hn&query={input.query}&hitsPerPage={input.count}&numericFilters=points>2"
+        )
+        data = response.json()
+
+        log.info("hnSearch", extra={"data": data})
+        return data
+    except Exception as error:
+        log.error("Error fetching Hacker News posts:", exc_info=True)
+        return []
diff --git a/examples/llama_quickstart/src/functions/llm/__init__.py b/examples/llama_quickstart/src/functions/llm/__init__.py
diff --git a/examples/llama_quickstart/src/functions/llm/chat.py b/examples/llama_quickstart/src/functions/llm/chat.py
@@ -0,0 +1,38 @@
+from llama_index.llms.together import TogetherLLM
+from restack_ai.function import function, log, FunctionFailure
+from llama_index.core.llms import ChatMessage, MessageRole
+import os
+from pydantic import BaseModel
+from dotenv import load_dotenv
+
+load_dotenv()
+
+class FunctionInputParams(BaseModel):
+    system_prompt: str
+    user_prompt: str
+
+@function.defn(name="llm_chat")
+async def llm_chat(input: FunctionInputParams):
+    try:
+        api_key = os.getenv("TOGETHER_API_KEY")
+        if not api_key:
+            log.error("TOGETHER_API_KEY environment variable is not set.")
+            raise ValueError("TOGETHER_API_KEY environment variable is required.")
+    
+        llm = TogetherLLM(
+            model="meta-llama/Llama-3.2-11B-Vision-Instruct-Turbo", api_key=api_key
+        )
+        messages = [
+            ChatMessage(
+                role=MessageRole.SYSTEM, content=input.system_prompt
+            ),
+            ChatMessage(
+                role=MessageRole.USER, content=input.user_prompt
+            ),
+        ]
+        resp = llm.chat(messages)
+        return resp.message.content
+    except Exception as e:
+        log.error(f"Error interacting with llm: {e}")
+        raise FunctionFailure(f"Error interacting with llm: {e}", non_retryable=True)
+  
diff --git a/examples/llama_quickstart/src/services.py b/examples/llama_quickstart/src/services.py
@@ -0,0 +1,29 @@
+import asyncio
+from src.client import client
+from src.functions.llm.chat import llm_chat
+from src.functions.hn.search import hn_search
+from src.workflows.workflow import hn_workflow
+from src.functions.crawl.website import crawl_website
+from restack_ai.restack import ServiceOptions
+
+async def main():
+    await asyncio.gather(
+        client.start_service(
+            workflows=[hn_workflow],
+            functions=[hn_search, crawl_website]
+        ),
+        client.start_service(
+            functions=[llm_chat],
+            task_queue="llm_chat",
+            options=ServiceOptions(
+                rate_limit=1,
+                max_concurrent_function_runs=1
+            )
+        )
+    )
+
+def run_services():
+    asyncio.run(main())
+
+if __name__ == "__main__":
+    run_services()
diff --git a/examples/llama_quickstart/src/workflows/__init__.py b/examples/llama_quickstart/src/workflows/__init__.py
diff --git a/examples/llama_quickstart/src/workflows/workflow.py b/examples/llama_quickstart/src/workflows/workflow.py
@@ -0,0 +1,38 @@
+from datetime import timedelta
+from restack_ai.workflow import workflow, import_functions, log
+
+with import_functions():
+    from src.functions.hn.search import hn_search
+    from src.functions.hn.schema import HnSearchInput
+    from src.functions.crawl.website import crawl_website
+    from src.functions.llm.chat import llm_chat, FunctionInputParams
+    
+
+@workflow.defn(name="hn_workflow")
+class hn_workflow:
+    @workflow.run
+    async def run(self, input: dict):
+
+        query = input["query"]
+        count = input["count"]
+        hn_results = await workflow.step(hn_search, HnSearchInput(query=query, count=count), start_to_close_timeout=timedelta(seconds=10))
+        urls = [hit['url'] for hit in hn_results['hits'] if 'url' in hit]
+
+        crawled_contents = []
+        for url in urls:  # Use the extracted URLs
+            log.info("hn_result", extra={"url": url})
+            if url:
+                content = await workflow.step(crawl_website, url, start_to_close_timeout=timedelta(seconds=30))
+                crawled_contents.append(content)
+        
+        summaries = []
+        for content in crawled_contents:
+            system_prompt = f"Provide a summary of the website for project found on Hacker news"
+            user_prompt = f"Summarize the following content: {content}"
+            summary = await workflow.step(llm_chat, FunctionInputParams(system_prompt=system_prompt, user_prompt=user_prompt), task_queue="llm_chat",start_to_close_timeout=timedelta(seconds=120))
+            summaries.append(summary)
+
+        system_prompt = f"You are a personal assistant. Provide a summary of the latest hacker news and the summaries of the websites. Structure your response with the title of the project, then a short description and a list of actionable bullet points."
+        user_prompt = f"Here is the latest hacker news data: {str(hn_results)} and summaries of the websites: {str(summaries)}"
+
+        return await workflow.step(llm_chat, FunctionInputParams(system_prompt=system_prompt,user_prompt=user_prompt), task_queue="llm_chat", start_to_close_timeout=timedelta(seconds=120))

Original file line number	Diff line number	Diff line change
`@@ -0,0 +1 @@`
	`1`	`+TOGETHER_API_KEY=<your-together-api-key>`
Original file line number	Diff line number	Diff line change
`@@ -0,0 +1,3 @@`
	`1`	`+from restack_ai import Restack`
	`2`	`+`
	`3`	`+client = Restack()`