Repository: upstash/degree-guru Branch: master Commit: f063f69e0071 Files: 38 Total size: 50.0 KB Directory structure: gitextract_e5_f89hi/ ├── .eslintrc.json ├── .gitignore ├── .prettierignore ├── .prettierrc ├── README.md ├── degreegurucrawler/ │ ├── .gitignore │ ├── Dockerfile │ ├── degreegurucrawler/ │ │ ├── __init__.py │ │ ├── items.py │ │ ├── middlewares.py │ │ ├── pipelines.py │ │ ├── settings.py │ │ ├── spiders/ │ │ │ ├── __init__.py │ │ │ └── configurable.py │ │ └── utils/ │ │ ├── config.py │ │ ├── crawler.yaml │ │ └── upstash_vector_store.py │ ├── docker-compose.yml │ ├── requirements.txt │ └── scrapy.cfg ├── next.config.js ├── package.json ├── postcss.config.js ├── src/ │ ├── app/ │ │ ├── api/ │ │ │ └── guru/ │ │ │ └── route.tsx │ │ ├── globals.css │ │ ├── layout.tsx │ │ ├── page.tsx │ │ └── vectorstore/ │ │ ├── UpstashVectorStore.d.ts │ │ └── UpstashVectorStore.js │ ├── components/ │ │ ├── form.tsx │ │ ├── message-loading.tsx │ │ ├── message.tsx │ │ ├── powered-by.tsx │ │ └── upstash-logo.tsx │ └── utils/ │ ├── const.ts │ └── cx.ts ├── tailwind.config.ts └── tsconfig.json ================================================ FILE CONTENTS ================================================ ================================================ FILE: .eslintrc.json ================================================ { "extends": "next/core-web-vitals" } ================================================ FILE: .gitignore ================================================ # See https://help.github.com/articles/ignoring-files/ for more about ignoring files. # dependencies /node_modules /.pnp .pnp.js .yarn/install-state.gz # testing /coverage # next.js /.next/ /out/ # production /build # misc .DS_Store *.pem .idea # debug npm-debug.log* yarn-debug.log* yarn-error.log* # local env files .env*.local # vercel .vercel # typescript *.tsbuildinfo next-env.d.ts ================================================ FILE: .prettierignore ================================================ # Ignore artifacts: degreegurucrawler figs ================================================ FILE: .prettierrc ================================================ { "arrowParens": "always", "bracketSameLine": false, "bracketSpacing": true, "semi": true, "singleQuote": false, "jsxSingleQuote": false, "quoteProps": "as-needed", "trailingComma": "all", "singleAttributePerLine": false, "htmlWhitespaceSensitivity": "css", "vueIndentScriptAndStyle": false, "proseWrap": "preserve", "insertPragma": false, "printWidth": 80, "requirePragma": false, "tabWidth": 2, "useTabs": false, "embeddedLanguageFormatting": "auto", "jsxBracketSameLine": false, "fluid": false, "importOrderSeparation": true, "importOrderSortSpecifiers": true, "importOrderBuiltinModulesToTop": true, "importOrderParserPlugins": ["typescript", "jsx"] } ================================================ FILE: README.md ================================================ # DegreeGuru ## Build a RAG Chatbot using Vercel AI SDK, Langchain, Upstash Vector and OpenAI [![Deploy with Vercel](https://vercel.com/button)](https://vercel.com/new/clone?repository-url=https%3A%2F%2Fgithub.com%2Fupstash%2Fdegreeguru&env=UPSTASH_REDIS_REST_URL,UPSTASH_REDIS_REST_TOKEN,UPSTASH_VECTOR_REST_URL,UPSTASH_VECTOR_REST_TOKEN,OPENAI_API_KEY&demo-title=DegreeGuru%20Demo&demo-description=A%20Demo%20Showcasing%20the%20DegreeGuru%20App&demo-url=https%3A%2F%2Fdegreeguru.vercel.app%2F&demo-image=https%3A%2F%2Fupstash.com%2Ficons%2Ffavicon-32x32.png) ![overview](figs/overview.gif) > [!NOTE] > **This project is a Community Project.** > > The project is maintained and supported by the community. Upstash may contribute but does not officially support or assume responsibility for it. **DegreeGuru** is a project designed to teach you making your own AI RAG chatbot on any custom data. Some of our favorite features: - 🕷️ Built-in crawler that scrapes the website you point it to, automatically making this data available for the AI - ⚡ Fast answers using Upstash Vector and real-time data streaming - 🛡️ Includes rate limiting to prevent API abuse This chatbot is trained on data from Stanford University as an example, but is totally domain agnostic. We've created this project so you can turn it into a chatbot with your very own data by simply modifying the `crawler.yaml` file. ## Overview 1. [Stack](#stack) 2. [Quickstart](#quickstart) 1. [Crawler](#crawler) 2. [ChatBot](#chatbot) 3. [Conclusion](#conclusion) 4. [Shortcomings](#shortcomings) ## Stack - Crawler: [scrapy](https://scrapy.org/) - Chatbot App: [Next.js](https://nextjs.org/) - Vector DB: [Upstash](https://upstash.com/) - LLM Orchestration: [Langchain.js](https://js.langchain.com) - Generative Model: [OpenAI](https://openai.com/), [gpt-3.5-turbo-1106](https://platform.openai.com/docs/models) - Embedding Model: [OpenAI](https://openai.com/), [text-embedding-ada-002](https://platform.openai.com/docs/guides/embeddings) - Text Streaming: [Vercel AI](https://vercel.com/ai) - Rate Limiting: [Upstash](https://upstash.com/) ## Quickstart For local development, we recommend forking this project and cloning the forked repository to your local machine by running the following command: ``` git clone git@github.com:[YOUR_GITHUB_ACCOUNT]/DegreeGuru.git ``` This project contains two primary components: the crawler and the chatbot. First, we'll take a look at how the crawler extracts information from any website you point it to. This data is automatically stored in an Upstash Vector database. If you already have a vector database available, the crawling stage can be skipped. ### Step 1: Crawler ![crawler-diagram](figs/how-this-project-works.png) The crawler is developed using Python, by [initializing a Scrapy project](https://docs.scrapy.org/en/latest/intro/tutorial.html#creating-a-project) and implementing a [custom spider](https://github.com/upstash/degreeguru/blob/master/degreegurucrawler/degreegurucrawler/spiders/configurable.py). The spider is equipped with [the `parse_page` function](https://github.com/upstash/degreeguru/blob/master/degreegurucrawler/degreegurucrawler/spiders/configurable.py#L42), invoked each time the spider visits a webpage. This callback function splits the text on the webpage into chunks, generates vector embeddings for each chunk, and upserts those vectors into your Upstash Vector Database. Each vector stored in our database includes the original text and website URL as metadata.
To run the crawler, follow these steps: > [!TIP] > If you have docker installed, you can skip the "Configure Environment Variables" and "Install Required Python Libraries" sections. Instead you can simply update the environment variables in [docker-compose.yml](https://github.com/upstash/DegreeGuru/blob/master/degreegurucrawler/docker-compose.yml) and run `docker-compose up`. This will create a container running our crawler. Don't forget to configure the crawler as explained in the following sections!
Configure Environment Variables Before we can run our crawler, we need to configure environment variables. They let us securely store sensitive information, such as the API keys we need to communicate with OpenAI or Upstash Vector. If you don't already have an Upstash Vector Database, create one [here](https://console.upstash.com/vector) and set 1536 as the vector dimensions. We set 1536 here because that is the amount needed by the embedding model we will use. ![vector-db-create](figs/vector-db-create.png) The following environment variables should be set: ``` # Upstash Vector credentials retrieved here: https://console.upstash.com/vector UPSTASH_VECTOR_REST_URL=**** UPSTASH_VECTOR_REST_TOKEN=**** # OpenAI key retrieved here: https://platform.openai.com/api-keys OPENAI_API_KEY=**** ```
Install Required Python Libraries To install the libraries, we suggest setting up a virtual Python environment. Before starting the installation, navigate to the `degreegurucrawler` directory. To setup a virtual environment, first install `virtualenv` package: ```bash pip install virtualenv ``` Then, create a new virtual environment and activate it: ```bash # create environment python3 -m venv venv # activate environment source venv/bin/activate ``` Finally, use [the `requirements.txt`](https://github.com/upstash/degreeguru/blob/master/degreegurucrawler/requirements.txt) to install the required libraries: ```bash pip install -r requirements.txt ```

After setting these environment variables, we are almost ready to run the crawler. The subsequent step involves configuring the crawler itself, primarily accomplished through the `crawler.yaml` file located in the `degreegurucrawler/utils` directory. Additionally, it is imperative to address a crucial setting within the `settings.py` file.
Configuring the crawler in `crawler.yaml` The crawler.yaml has two main sections: `crawler` and `index`: ```yaml crawler: start_urls: - https://www.some.domain.com link_extractor: allow: '.*some\.domain.*' deny: - "#" - '\?' - about index: openAI_embedding_model: text-embedding-ada-002 text_splitter: chunk_size: 1000 chunk_overlap: 100 ``` In the `crawler` section, there are two subsections: - `start_urls`: the entrypoints our crawler will start searching from - `link_extractor`: a dictionary passed as arguments to [`scrapy.linkextractors.LinkExtractor`](https://docs.scrapy.org/en/latest/topics/link-extractors.html). Some important parameters are: - `allow`: Only extracts links matching the given regex(s) - `allow_domains`: Only extract links matching the given domain(s) - `deny`: Deny links matching the given regex(s) In the `index` section, there are two subsections: - `openAI_embedding_model`: The embedding model to use - `test_splitter`: a dictionary passed as arguments to [`langchain.text_splitter.RecursiveCharacterTextSplitter`](https://api.python.langchain.com/en/latest/text_splitter/langchain.text_splitter.RecursiveCharacterTextSplitter.html)
Configuring crawl depth via `settings.py` `settings.py` file has an important setting called `DEPTH_LIMIT` which determines how many consecutive links our spider can crawl. A high value lets our crawler visit the deepest corners of a website, taking longer to finish with possibly diminishing returns. A low value could end the crawl before extracting relevant information. If pages are skipped due to the `DEPTH_LIMIT`, Scrapy logs those skipped URLs for us. Because this usually causes a lot of logs, we've disabled this option in our project. If you'd like to keep it enabled, remove [the `"scrapy.spidermiddlewares.depth"` from the `disable_loggers` in `degreegurucrawler/spider/configurable.py` file](https://github.com/upstash/degreeguru/blob/master/degreegurucrawler/degreegurucrawler/spiders/configurable.py#L22).

That's it! 🎉 We've configured our crawler and are ready to run it using the following command: ``` scrapy crawl configurable --logfile degreegurucrawl.log ``` Note that running this might take time. You can monitor the progress by looking at the log file `degreegurucrawl.log` or the metrics of your Upstash Vector Database dashboard as shown below. ![vector-db](figs/vector-db.png) > [!TIP] > If you want to do a dry run (without creating embeddings or a vector database), simply comment out [the line where we pass the `callback` parameter to the `Rule` object in `ConfigurableSpider`](https://github.com/upstash/degreeguru/blob/master/degreegurucrawler/degreegurucrawler/spiders/configurable.py#L38) ### Step 2: Chatbot In this section, we'll explore how to chat with the data we've just crawled and stored in our vector database. Here's an overview of what this will look like architecturally: ![chatbot-diagram](figs/infrastructure.png) Before we can run the chatbot locally, we need to set the environment variables as shown in the [`.env.local.example`](https://github.com/upstash/degreeguru/blob/master/.env.local.example) file. Rename this file and remove the `.example` ending, leaving us with `.env.local`. Your `.env.local` file should look like this: ``` # Redis tokens retrieved here: https://console.upstash.com/ UPSTASH_REDIS_REST_URL= UPSTASH_REDIS_REST_TOKEN= # Vector database tokens retrieved here: https://console.upstash.com/vector UPSTASH_VECTOR_REST_URL= UPSTASH_VECTOR_REST_TOKEN= # OpenAI key retrieved here: https://platform.openai.com/api-keys OPENAI_API_KEY= ``` The first four variables are provided by Upstash, you can visit the commented links for the place to retrieve these tokens. You can find the vector database tokens here: ![vector-db-read-only](figs/vector-db-read-only.png) The `UPSTASH_REDIS_REST_URL` and `UPSTASH_REDIS_REST_TOKEN` are needed for rate-limiting based on IP address. In order to get these secrets, go to Upstash dashboard and create a Redis database. ![redis-create](figs/redis-create.png) Finally, set the `OPENAI_API_KEY` environment variable you can get [here](https://platform.openai.com/api-keys) which allows us to vectorize user queries and generate responses. That's the setup done! 🎉 We've configured our crawler, set up all neccessary environment variables are after running `npm install` to install all local packages needed to run the app, we can start our chatbot using the command: ```bash npm run dev ``` Visit `http://localhost:3000` to see your chatbot live in action! ### Step 3: Optional tweaking You can use this chatbot in two different modes: - Streaming Mode: model responses are streamed to the web application in real-time as the model generates them. Interaction with the app is more fluid. - Non-Streaming Mode: Model responses are shown to the user once entirely generated. In this mode, DegreeGuru can explicitly provide the URLs of the web pages it uses as context.
Changing streaming mode To turn streaming on/off, navigate to `src/app/route/guru` and open the `route.tsx` file. Setting [`returnIntermediateSteps`](https://github.com/upstash/degreeguru/blob/master/src/app/api/guru/route.tsx#L64) to `true` disables streaming, setting it to `false` enables streaming.
To customize the chatbot further, you can update the [AGENT_SYSTEM_TEMPLATE in your route.tsx file](https://github.com/upstash/DegreeGuru/blob/master/src/app/api/guru/route.tsx#L101) to better match your specific use case.
## Conclusion Congratulations on setting up your own AI chatbot! We hope you learned a lot by following along and seeing how the different parts of this app, namely the crawler, vector database, and LLM, play together. A major focus in developing this project was on its user-friendly design and adaptable settings to make this project perfect for your use case. ## Limitations The above implementation works great for a variety of use cases. There are a few limitations I'd like to mention: - Because the Upstash LangChain integration is a work-in-progress, the [`UpstashVectorStore`](https://github.com/upstash/degreeguru/blob/master/src/app/vectorstore/UpstashVectorStore.js) used with LangChain currently only implements the `similaritySearchVectorWithScore` method needed for our agent. Once we're done developing our native LangChain integration, we'll update this project accordingly. - When the non-streaming mode is enabled, the message history can cause an error after the user enters another query. - Our sources are available as URLs in the Upstash Vector Database, but we cannot show the sources explicitly when streaming. Instead, we provide the links to the chatbot as context and expect the bot to include the links in the response. ================================================ FILE: degreegurucrawler/.gitignore ================================================ *.log __pycache__ degreegurudb ================================================ FILE: degreegurucrawler/Dockerfile ================================================ FROM python:3.8-slim # copy directory into dockerfile COPY . crawler WORKDIR crawler # install requirements RUN pip install -r requirements.txt CMD ["scrapy", "crawl", "configurable"] ================================================ FILE: degreegurucrawler/degreegurucrawler/__init__.py ================================================ ================================================ FILE: degreegurucrawler/degreegurucrawler/items.py ================================================ # Define here the models for your scraped items # # See documentation in: # https://docs.scrapy.org/en/latest/topics/items.html import scrapy class DegreegurucrawlerItem(scrapy.Item): # define the fields for your item here like: # name = scrapy.Field() pass ================================================ FILE: degreegurucrawler/degreegurucrawler/middlewares.py ================================================ # Define here the models for your spider middleware # # See documentation in: # https://docs.scrapy.org/en/latest/topics/spider-middleware.html from scrapy import signals # useful for handling different item types with a single interface from itemadapter import is_item, ItemAdapter class DegreegurucrawlerSpiderMiddleware: # Not all methods need to be defined. If a method is not defined, # scrapy acts as if the spider middleware does not modify the # passed objects. @classmethod def from_crawler(cls, crawler): # This method is used by Scrapy to create your spiders. s = cls() crawler.signals.connect(s.spider_opened, signal=signals.spider_opened) return s def process_spider_input(self, response, spider): # Called for each response that goes through the spider # middleware and into the spider. # Should return None or raise an exception. return None def process_spider_output(self, response, result, spider): # Called with the results returned from the Spider, after # it has processed the response. # Must return an iterable of Request, or item objects. for i in result: yield i def process_spider_exception(self, response, exception, spider): # Called when a spider or process_spider_input() method # (from other spider middleware) raises an exception. # Should return either None or an iterable of Request or item objects. pass def process_start_requests(self, start_requests, spider): # Called with the start requests of the spider, and works # similarly to the process_spider_output() method, except # that it doesn’t have a response associated. # Must return only requests (not items). for r in start_requests: yield r def spider_opened(self, spider): spider.logger.info("Spider opened: %s" % spider.name) class DegreegurucrawlerDownloaderMiddleware: # Not all methods need to be defined. If a method is not defined, # scrapy acts as if the downloader middleware does not modify the # passed objects. @classmethod def from_crawler(cls, crawler): # This method is used by Scrapy to create your spiders. s = cls() crawler.signals.connect(s.spider_opened, signal=signals.spider_opened) return s def process_request(self, request, spider): # Called for each request that goes through the downloader # middleware. # Must either: # - return None: continue processing this request # - or return a Response object # - or return a Request object # - or raise IgnoreRequest: process_exception() methods of # installed downloader middleware will be called return None def process_response(self, request, response, spider): # Called with the response returned from the downloader. # Must either; # - return a Response object # - return a Request object # - or raise IgnoreRequest return response def process_exception(self, request, exception, spider): # Called when a download handler or a process_request() # (from other downloader middleware) raises an exception. # Must either: # - return None: continue processing this exception # - return a Response object: stops process_exception() chain # - return a Request object: stops process_exception() chain pass def spider_opened(self, spider): spider.logger.info("Spider opened: %s" % spider.name) ================================================ FILE: degreegurucrawler/degreegurucrawler/pipelines.py ================================================ # Define your item pipelines here # # Don't forget to add your pipeline to the ITEM_PIPELINES setting # See: https://docs.scrapy.org/en/latest/topics/item-pipeline.html # useful for handling different item types with a single interface from itemadapter import ItemAdapter class DegreegurucrawlerPipeline: def process_item(self, item, spider): return item ================================================ FILE: degreegurucrawler/degreegurucrawler/settings.py ================================================ # Scrapy settings for degreegurucrawler project # # For simplicity, this file contains only settings considered important or # commonly used. You can find more settings consulting the documentation: # # https://docs.scrapy.org/en/latest/topics/settings.html # https://docs.scrapy.org/en/latest/topics/downloader-middleware.html # https://docs.scrapy.org/en/latest/topics/spider-middleware.html BOT_NAME = "degreegurucrawler" SPIDER_MODULES = ["degreegurucrawler.spiders"] NEWSPIDER_MODULE = "degreegurucrawler.spiders" DEPTH_LIMIT = 3 # Crawl responsibly by identifying yourself (and your website) on the user-agent #USER_AGENT = "degreegurucrawler (+http://www.yourdomain.com)" # Obey robots.txt rules ROBOTSTXT_OBEY = True # Configure maximum concurrent requests performed by Scrapy (default: 16) #CONCURRENT_REQUESTS = 32 # Configure a delay for requests for the same website (default: 0) # See https://docs.scrapy.org/en/latest/topics/settings.html#download-delay # See also autothrottle settings and docs #DOWNLOAD_DELAY = 3 # The download delay setting will honor only one of: #CONCURRENT_REQUESTS_PER_DOMAIN = 16 #CONCURRENT_REQUESTS_PER_IP = 16 # Disable cookies (enabled by default) #COOKIES_ENABLED = False # Disable Telnet Console (enabled by default) #TELNETCONSOLE_ENABLED = False # Override the default request headers: #DEFAULT_REQUEST_HEADERS = { # "Accept": "text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8", # "Accept-Language": "en", #} # Enable or disable spider middlewares # See https://docs.scrapy.org/en/latest/topics/spider-middleware.html #SPIDER_MIDDLEWARES = { # "degreegurucrawler.middlewares.DegreegurucrawlerSpiderMiddleware": 543, #} # Enable or disable downloader middlewares # See https://docs.scrapy.org/en/latest/topics/downloader-middleware.html #DOWNLOADER_MIDDLEWARES = { # "degreegurucrawler.middlewares.DegreegurucrawlerDownloaderMiddleware": 543, #} # Enable or disable extensions # See https://docs.scrapy.org/en/latest/topics/extensions.html #EXTENSIONS = { # "scrapy.extensions.telnet.TelnetConsole": None, #} # Configure item pipelines # See https://docs.scrapy.org/en/latest/topics/item-pipeline.html #ITEM_PIPELINES = { # "degreegurucrawler.pipelines.DegreegurucrawlerPipeline": 300, #} # Enable and configure the AutoThrottle extension (disabled by default) # See https://docs.scrapy.org/en/latest/topics/autothrottle.html #AUTOTHROTTLE_ENABLED = True # The initial download delay #AUTOTHROTTLE_START_DELAY = 5 # The maximum download delay to be set in case of high latencies #AUTOTHROTTLE_MAX_DELAY = 60 # The average number of requests Scrapy should be sending in parallel to # each remote server #AUTOTHROTTLE_TARGET_CONCURRENCY = 1.0 # Enable showing throttling stats for every response received: #AUTOTHROTTLE_DEBUG = False # Enable and configure HTTP caching (disabled by default) # See https://docs.scrapy.org/en/latest/topics/downloader-middleware.html#httpcache-middleware-settings #HTTPCACHE_ENABLED = True #HTTPCACHE_EXPIRATION_SECS = 0 #HTTPCACHE_DIR = "httpcache" #HTTPCACHE_IGNORE_HTTP_CODES = [] #HTTPCACHE_STORAGE = "scrapy.extensions.httpcache.FilesystemCacheStorage" # Set settings whose default value is deprecated to a future-proof value REQUEST_FINGERPRINTER_IMPLEMENTATION = "2.7" TWISTED_REACTOR = "twisted.internet.asyncioreactor.AsyncioSelectorReactor" FEED_EXPORT_ENCODING = "utf-8" ================================================ FILE: degreegurucrawler/degreegurucrawler/spiders/__init__.py ================================================ # This package will contain the spiders of your Scrapy project # # Please refer to the documentation for information on how to create and manage # your spiders. ================================================ FILE: degreegurucrawler/degreegurucrawler/spiders/configurable.py ================================================ import os import uuid import logging from ..utils.upstash_vector_store import UpstashVectorStore from ..utils.config import text_splitter_config, crawler_config from scrapy.spiders import CrawlSpider, Rule from scrapy.linkextractors import LinkExtractor from langchain.text_splitter import RecursiveCharacterTextSplitter class ConfigurableSpider(CrawlSpider): name = "configurable" start_urls = crawler_config["start_urls"] rules = ( Rule( LinkExtractor( **crawler_config["link_extractor"] ), callback="parse_page", follow=True # to enable following links on each page when callback is provided ), ) def __init__(self, *a, **kw): super().__init__(*a, **kw) self.vectorstore = UpstashVectorStore( url=os.environ.get("UPSTASH_VECTOR_REST_URL"), token=os.environ.get("UPSTASH_VECTOR_REST_TOKEN") ) print( f"Creating a vector index at {os.environ.get('UPSTASH_VECTOR_REST_URL')}.\n" f" Vector store info before crawl: {self.vectorstore.index.info()}" ) self.text_splitter = RecursiveCharacterTextSplitter( **text_splitter_config ) self._disable_loggers() def _disable_loggers(self): """ disables some of the loggers to keep the log clean """ disable_loggers = [ "scrapy.spidermiddlewares.depth", "protego", "httpcore.http11", "httpx", "openai._base_client", "urllib3.connectionpool" ] for logger in disable_loggers: logging.getLogger(logger).setLevel(logging.WARNING) def parse_page(self, response): """ Creates chunks out of the crawled webpage and adds them to the vector store. """ # extract text content text_content = response.xpath('//p').getall() text_content = '\n'.join(text_content) # split documents documents = self.text_splitter.split_text(text_content) if len(documents) == 0: return # get source url link = response.url # add documents to vector store self.vectorstore.add( ids=[str(uuid.uuid4())[:8] for doc in documents], documents=documents, link=link ) ================================================ FILE: degreegurucrawler/degreegurucrawler/utils/config.py ================================================ import os import yaml config_path = "degreegurucrawler/utils/crawler.yaml" with open(config_path, 'r') as file: config = yaml.load(file, Loader=yaml.FullLoader) embedding_function_config = { "api_key": os.environ.get('OPENAI_API_KEY'), "model_name": config["index"]["openAI_embedding_model"] } crawler_config = config["crawler"] text_splitter_config = config["index"]["text_splitter"] ================================================ FILE: degreegurucrawler/degreegurucrawler/utils/crawler.yaml ================================================ crawler: start_urls: - https://www.some.domain.com link_extractor: allow: '.*some\.domain.*' deny: - "#" - '\?' - course - search - subjects - degree-charts - archive - news - alumni - announcement - people - topics - membership - section - about - letter - member - committee - book - year - project - user - page - event - resource - login index: openAI_embedding_model: text-embedding-ada-002 text_splitter: chunk_size: 1000 chunk_overlap: 100 ================================================ FILE: degreegurucrawler/degreegurucrawler/utils/upstash_vector_store.py ================================================ from typing import List from openai import OpenAI from upstash_vector import Index class UpstashVectorStore: def __init__( self, url: str, token: str ): self.client = OpenAI() self.index = Index(url=url, token=token) def get_embeddings( self, documents: List[str], model: str = "text-embedding-ada-002" ) -> List[List[float]]: """ Given a list of documents, generates and returns a list of embeddings """ documents = [document.replace("\n", " ") for document in documents] embeddings = self.client.embeddings.create( input = documents, model=model ) return [data.embedding for data in embeddings.data] def add( self, ids: List[str], documents: List[str], link: str ) -> None: """ Adds a list of documents to the Upstash Vector Store """ embeddings = self.get_embeddings(documents) self.index.upsert( vectors=[ ( id, embedding, { "text": document, "url": link } ) for id, embedding, document in zip(ids, embeddings, documents) ] ) ================================================ FILE: degreegurucrawler/docker-compose.yml ================================================ version: '3' services: my_service: image: degreegurucrawler build: context: . dockerfile: Dockerfile environment: - UPSTASH_VECTOR_REST_URL=**** - UPSTASH_VECTOR_REST_TOKEN=**** - OPENAI_API_KEY=**** ================================================ FILE: degreegurucrawler/requirements.txt ================================================ upstash_vector scrapy==2.11.0 langchain==0.1.0 openai==1.7.2 ================================================ FILE: degreegurucrawler/scrapy.cfg ================================================ # Automatically created by: scrapy startproject # # For more information about the [deploy] section see: # https://scrapyd.readthedocs.io/en/latest/deploy.html [settings] default = degreegurucrawler.settings [deploy] #url = http://localhost:6800/ project = degreegurucrawler ================================================ FILE: next.config.js ================================================ /** @type {import('next').NextConfig} */ const nextConfig = {}; module.exports = nextConfig; ================================================ FILE: package.json ================================================ { "name": "degreeguru", "version": "0.1.0", "private": true, "scripts": { "dev": "next dev", "build": "next build", "start": "next start", "lint": "next lint" }, "dependencies": { "@tabler/icons-react": "^2.47.0", "@upstash/ratelimit": "^2.0.3", "@upstash/redis": "^1.34.0", "@upstash/vector": "^0.1.0-alpha-13", "ai": "^2.2.31", "langchain": "^0.1.5", "markdown-to-jsx": "^7.4.0", "next": "14.2.35", "react": "^18", "react-dom": "^18" }, "devDependencies": { "@types/node": "^20", "@types/react": "^18", "@types/react-dom": "^18", "autoprefixer": "^10.0.1", "clsx": "^2.1.0", "eslint": "^8", "eslint-config-next": "14.0.4", "postcss": "^8", "prettier": "^3.2.5", "tailwind-merge": "^2.2.1", "tailwindcss": "^3.3.0", "typescript": "^5" } } ================================================ FILE: postcss.config.js ================================================ module.exports = { plugins: { tailwindcss: {}, autoprefixer: {}, }, }; ================================================ FILE: src/app/api/guru/route.tsx ================================================ import { NextRequest, NextResponse } from "next/server"; import { Ratelimit } from "@upstash/ratelimit"; import { Redis } from "@upstash/redis"; import { Message as VercelChatMessage, StreamingTextResponse } from "ai"; import { AIMessage, ChatMessage, HumanMessage } from "@langchain/core/messages"; import { ChatOpenAI, OpenAIEmbeddings } from "@langchain/openai"; import { createRetrieverTool } from "langchain/tools/retriever"; import { AgentExecutor, createOpenAIFunctionsAgent } from "langchain/agents"; import { ChatPromptTemplate, MessagesPlaceholder, } from "@langchain/core/prompts"; import { UpstashVectorStore } from "@/app/vectorstore/UpstashVectorStore"; export const runtime = "edge"; const redis = Redis.fromEnv(); const ratelimit = new Ratelimit({ redis: redis, limiter: Ratelimit.slidingWindow(1, "10 s"), }); const convertVercelMessageToLangChainMessage = (message: VercelChatMessage) => { if (message.role === "user") { return new HumanMessage(message.content); } else if (message.role === "assistant") { return new AIMessage(message.content); } else { return new ChatMessage(message.content, message.role); } }; export async function POST(req: NextRequest) { try { const ip = req.ip ?? "127.0.0.1"; const { success } = await ratelimit.limit(ip); if (!success) { const textEncoder = new TextEncoder(); const customString = "Oops! It seems you've reached the rate limit. Please try again later."; const transformStream = new ReadableStream({ async start(controller) { controller.enqueue(textEncoder.encode(customString)); controller.close(); }, }); return new StreamingTextResponse(transformStream); } const body = await req.json(); /** * We represent intermediate steps as system messages for display purposes, * but don't want them in the chat history. */ const messages = (body.messages ?? []).filter( (message: VercelChatMessage) => message.role === "user" || message.role === "assistant", ); const returnIntermediateSteps = false; const previousMessages = messages .slice(0, -1) .map(convertVercelMessageToLangChainMessage); const currentMessageContent = messages[messages.length - 1].content; const chatModel = new ChatOpenAI({ modelName: "gpt-3.5-turbo-1106", temperature: 0.2, // IMPORTANT: Must "streaming: true" on OpenAI to enable final output streaming below. streaming: true, }, { apiKey: process.env.OPENAI_API_KEY, organization: process.env.OPENAI_ORGANIZATION }); /** * Create vector store and retriever */ const vectorstore = await new UpstashVectorStore(new OpenAIEmbeddings()); const retriever = vectorstore.asRetriever( { k: 6, searchType: "mmr", searchKwargs: { fetchK: 20, lambda: 0.5 }, verbose: false }, ); /** * Wrap the retriever in a tool to present it to the agent in a * usable form. */ const tool = createRetrieverTool(retriever, { name: "search_latest_knowledge", description: "Searches and returns up-to-date general information.", }); /** * Based on https://smith.langchain.com/hub/hwchase17/openai-functions-agent * * This default prompt for the OpenAI functions agent has a placeholder * where chat messages get inserted as "chat_history". * * You can customize this prompt yourself! */ const AGENT_SYSTEM_TEMPLATE = ` You are an artificial intelligence university bot named DegreeGuru, programmed to respond to inquiries about Stanford in a highly systematic and data-driven manner. Begin your answers with a formal greeting and sign off with a closing statement about promoting knowledge. Your responses should be precise and factual, with an emphasis on using the context provided and providing links from the context whenever posible. If some link does not look like it belongs to stanford, don't use the link and the information in your response. Don't repeat yourself in your responses even if some information is repeated in the context. Reply with apologies and tell the user that you don't know the answer only when you are faced with a question whose answer is not available in the context. `; const prompt = ChatPromptTemplate.fromMessages([ ["system", AGENT_SYSTEM_TEMPLATE], new MessagesPlaceholder("chat_history"), ["human", "{input}"], new MessagesPlaceholder("agent_scratchpad"), ]); const agent = await createOpenAIFunctionsAgent({ llm: chatModel, tools: [tool], prompt, }); const agentExecutor = new AgentExecutor({ agent, tools: [tool], // Set this if you want to receive all intermediate steps in the output of .invoke(). returnIntermediateSteps, }); if (!returnIntermediateSteps) { /** * Agent executors also allow you to stream back all generated tokens and steps * from their runs. * * This contains a lot of data, so we do some filtering of the generated log chunks * and only stream back the final response. * * This filtering is easiest with the OpenAI functions or tools agents, since final outputs * are log chunk values from the model that contain a string instead of a function call object. * * See: https://js.langchain.com/docs/modules/agents/how_to/streaming#streaming-tokens */ const logStream = await agentExecutor.streamLog({ input: currentMessageContent, chat_history: previousMessages, }); const textEncoder = new TextEncoder(); const transformStream = new ReadableStream({ async start(controller) { for await (const chunk of logStream) { if (chunk.ops?.length > 0 && chunk.ops[0].op === "add") { const addOp = chunk.ops[0]; if ( addOp.path.startsWith("/logs/ChatOpenAI") && typeof addOp.value === "string" && addOp.value.length ) { controller.enqueue(textEncoder.encode(addOp.value)); } } } controller.close(); }, }); return new StreamingTextResponse(transformStream); } else { /** * Intermediate steps are the default outputs with the executor's `.stream()` method. * We could also pick them out from `streamLog` chunks. * They are generated as JSON objects, so streaming them is a bit more complicated. */ const result = await agentExecutor.invoke({ input: currentMessageContent, chat_history: previousMessages, }); const urls = JSON.parse( `[${result.intermediateSteps[0]?.observation.replaceAll("}\n\n{", "}, {")}]`, ).map((source: { url: any }) => source.url); return NextResponse.json( { _no_streaming_response_: true, output: result.output, sources: urls, }, { status: 200 }, ); } } catch (e: any) { console.log(e.message); return NextResponse.json({ error: e.message }, { status: 500 }); } } ================================================ FILE: src/app/globals.css ================================================ @tailwind base; @tailwind components; @tailwind utilities; @layer base { a, button { @apply transition; } a { @apply text-emerald-700 underline decoration-emerald-700/60 decoration-2 hover:decoration-emerald-700 hover:bg-emerald-200 outline-0 focus:ring-2 focus:ring-offset-1 focus:ring-emerald-500; } ::selection { @apply bg-emerald-200 text-emerald-950; } button:focus, input:focus { @apply outline-0 ring-2 ring-offset-1 ring-emerald-500 caret-emerald-500; } label, strong, b { @apply font-semibold; } h1, h2, h3, h4 { @apply text-balance; } p { @apply text-pretty; } } ================================================ FILE: src/app/layout.tsx ================================================ import type { Metadata } from "next"; import { Inter } from "next/font/google"; import "./globals.css"; import cx from "@/utils/cx"; const inter = Inter({ subsets: ["latin"] }); export const metadata: Metadata = { title: "DegreeGuru", description: "DegreeGuru ChatBot", }; export default function RootLayout({ children, }: { children: React.ReactNode; }) { return ( {children} ); } ================================================ FILE: src/app/page.tsx ================================================ "use client"; import React, { useCallback, useEffect, useRef, useState } from "react"; import { Message as MessageProps, useChat } from "ai/react"; import Form from "@/components/form"; import Message from "@/components/message"; import cx from "@/utils/cx"; import PoweredBy from "@/components/powered-by"; import MessageLoading from "@/components/message-loading"; import { INITIAL_QUESTIONS } from "@/utils/const"; export default function Home() { const formRef = useRef(null); const messagesEndRef = useRef(null); const [streaming, setStreaming] = useState(false); const { messages, input, handleInputChange, handleSubmit, setInput } = useChat({ api: "/api/guru", initialMessages: [ { id: "0", role: "system", content: `**Welcome to DegreeGuru** Your ultimate companion in navigating the academic landscape of Stanford.`, }, ], onResponse: () => { setStreaming(false); }, }); const onClickQuestion = (value: string) => { setInput(value); setTimeout(() => { formRef.current?.dispatchEvent( new Event("submit", { cancelable: true, bubbles: true, }), ); }, 1); }; useEffect(() => { if (messagesEndRef.current) { messagesEndRef.current.scrollIntoView(); } }, [messages]); const onSubmit = useCallback( (e: React.FormEvent) => { e.preventDefault(); handleSubmit(e); setStreaming(true); }, [handleSubmit], ); return (
{messages.map((message: MessageProps) => { return ; })} {/* loading */} {streaming && } {/* initial question */} {messages.length === 1 && (
{INITIAL_QUESTIONS.map((message) => { return ( ); })}
)} {/* bottom ref */}
); } ================================================ FILE: src/app/vectorstore/UpstashVectorStore.d.ts ================================================ import { Index } from "@upstash/vector"; import { Document } from "@langchain/core/documents"; import { MaxMarginalRelevanceSearchOptions, VectorStore, } from "@langchain/core/vectorstores"; type UpstashMetadata = Record; export class UpstashVectorStore extends VectorStore { declare FilterType: PineconeMetadata; constructor(embeddings: any); index: Index; similaritySearchVectorWithScore( query: any, k: any, filter: any, ): Promise; maxMarginalRelevanceSearch( query: string, options: MaxMarginalRelevanceSearchOptions ): Promise } ================================================ FILE: src/app/vectorstore/UpstashVectorStore.js ================================================ import { VectorStore } from "@langchain/core/vectorstores"; import { Document } from "@langchain/core/documents"; import { Index } from "@upstash/vector"; import { maximalMarginalRelevance } from "@langchain/core/utils/math"; export class UpstashVectorStore extends VectorStore { _vectorstoreType() { return "upstash"; } constructor(embeddings) { super(embeddings); this.index = new Index({ url: process.env.UPSTASH_VECTOR_URL, token: process.env.UPSTASH_VECTOR_TOKEN, }); } async similaritySearchVectorWithScore(query, k, filter) { const result = await this.index.query({ vector: query, topK: k, includeVectors: false, includeMetadata: true, }); const results = []; for (let i = 0; i < result.length; i++) { results.push([ new Document({ pageContent: JSON.stringify(result[i]?.metadata) || "", }), ]); } return results; } async maxMarginalRelevanceSearch(query, options) { const queryEmbedding = await this.embeddings.embedQuery(query); const result = await this.index.query({ vector: queryEmbedding, topK: options.fetchK ?? 20, includeVectors: true, includeMetadata: true, }); const embeddingList = result.map((r) => r.vector) const mmrIndexes = maximalMarginalRelevance( queryEmbedding, embeddingList, options.lambda, options.k ); const topMmrMatches = mmrIndexes.map((idx) => result[idx]); const results = []; for (let i = 0; i < topMmrMatches.length; i++) { results.push( new Document({ pageContent: JSON.stringify(topMmrMatches[i]?.metadata) || "", }), ); } return results; } } ================================================ FILE: src/components/form.tsx ================================================ import { ComponentProps, forwardRef } from "react"; import { IconArrowBack } from "@tabler/icons-react"; import cx from "@/utils/cx"; export interface Props extends ComponentProps<"form"> { inputProps: ComponentProps<"input">; buttonProps: ComponentProps<"button">; } const Form = ({ inputProps, buttonProps, onSubmit }: Props, ref: any) => { return ( {/**/} ); }; export default forwardRef(Form); ================================================ FILE: src/components/message-loading.tsx ================================================ import React from "react"; import cx from "@/utils/cx"; import { Avatar } from "@/components/message"; const MessageLoading: React.FC = () => { return (
{/* https://github.com/n3r4zzurr0/svg-spinners/blob/main/svg-smil/3-dots-bounce.svg?short_path=50864c0 */}
); }; export default MessageLoading; ================================================ FILE: src/components/message.tsx ================================================ import React from "react"; import Markdown from "markdown-to-jsx"; import cx from "@/utils/cx"; import { Message as MessageProps } from "ai/react"; import UpstashLogo from "@/components/upstash-logo"; import { IconUser } from "@tabler/icons-react"; const Message: React.FC = ({ content, role }) => { const isUser = role === "user"; return (
    {children}
, ul: ({ children }) =>
    {children}
, }, }} > {content}
); }; const Avatar: React.FC<{ isUser?: boolean; className?: string }> = ({ isUser = false, className, }) => { return (
{isUser ? : }
); }; export default Message; export { Avatar }; ================================================ FILE: src/components/powered-by.tsx ================================================ const PoweredBy = () => { return (

This project is a prototype for a RAG chatbot.
Built using{" "} LangChain ,{" "} Upstash Vector {" "} and{" "} Vercel AI SDK {" "} ・{" "} Source Code

); }; export default PoweredBy; ================================================ FILE: src/components/upstash-logo.tsx ================================================ import React, { HTMLProps } from "react"; export interface Props extends HTMLProps { size?: number; } export default function UpstashLogo({ height = 20, ...props }: Props) { return ( ); } ================================================ FILE: src/utils/const.ts ================================================ export const INITIAL_QUESTIONS = [ { content: "Are there resources for students interested in creative writing?", }, { content: "Are there courses on environmental sustainability?", }, { content: "Are there any workshops or seminars on entrepreneurship for students?", }, { content: "What kinds of courses will I take as a philosophy major?", }, ]; ================================================ FILE: src/utils/cx.ts ================================================ import { ClassValue, clsx } from "clsx"; import { twMerge } from "tailwind-merge"; export default function cx(...inputs: ClassValue[]) { return twMerge(clsx(inputs)); } ================================================ FILE: tailwind.config.ts ================================================ import type { Config } from "tailwindcss"; const config: Config = { content: [ "./src/pages/**/*.{js,ts,jsx,tsx,mdx}", "./src/components/**/*.{js,ts,jsx,tsx,mdx}", "./src/app/**/*.{js,ts,jsx,tsx,mdx}", ], plugins: [], }; export default config; ================================================ FILE: tsconfig.json ================================================ { "compilerOptions": { "target": "es5", "lib": ["dom", "dom.iterable", "esnext"], "allowJs": true, "skipLibCheck": true, "strict": true, "noEmit": true, "esModuleInterop": true, "module": "esnext", "moduleResolution": "bundler", "resolveJsonModule": true, "isolatedModules": true, "jsx": "preserve", "incremental": true, "plugins": [ { "name": "next" } ], "paths": { "@/*": ["./src/*"] } }, "include": ["next-env.d.ts", "**/*.ts", "**/*.tsx", ".next/types/**/*.ts"], "exclude": ["node_modules"] }