<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Scott Regan on Medium]]></title>
        <description><![CDATA[Stories by Scott Regan on Medium]]></description>
        <link>https://medium.com/@scotttregan?source=rss-df3fd9625068------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*7oATcDFiqB9lcAP6xhqc1w.png</url>
            <title>Stories by Scott Regan on Medium</title>
            <link>https://medium.com/@scotttregan?source=rss-df3fd9625068------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 22:00:34 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@scotttregan/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Build an LLM app using Skeleton-of-Thought]]></title>
            <link>https://medium.com/logspace/build-an-llm-app-using-skeleton-of-thought-6aa7e2e54cd1?source=rss-df3fd9625068------2</link>
            <guid isPermaLink="false">https://medium.com/p/6aa7e2e54cd1</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[langchain]]></category>
            <category><![CDATA[retrieval-augmented-gen]]></category>
            <dc:creator><![CDATA[Scott Regan]]></dc:creator>
            <pubDate>Mon, 23 Sep 2024 03:36:49 GMT</pubDate>
            <atom:updated>2024-09-23T03:36:49.197Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Skeleton-of-Thought</strong> is a technique for improving the performance of Retrieval-Augmented Generation (RAG) applications, especially for long-form responses.</p><p>In RAG, large language models (LLMs) generate text from relevant information retrieved from external knowledge bases, proprietary databases, or document repositories.</p><p><a href="https://www.google.com/url?sa=t&amp;source=web&amp;rct=j&amp;opi=89978449&amp;url=https://arxiv.org/abs/2307.15337&amp;ved=2ahUKEwjKopvn3MeIAxXaHzQIHbcVGy0QFnoECBUQAQ&amp;usg=AOvVaw2FqcvdE6RNoSnhmnhuv3_e">Skeleton-of-Thought</a> gives faster and more relevant results by prompting the LLM to first create an outline, much like a person might approach a writing task. The LLM then develops each point of the outline and these are combined in the final response.</p><p>As one of several techniques in <a href="https://medium.com/logspace/beyond-basic-rag-similarity-relevance-fdbd16373d4a">advanced RAG</a>, Skeleton-of-Thought improves relevance, response time and reduces hallucinations.</p><p>In this example, we’ll build a chatbot in Langflow with Skeleton-of-Thought by first prompting the LLM to outline the document with short bullet-point style sentences. Then, we will prompt the LLM to generate the details of each point — in parallel — before assembling the final document.</p><p>Ready? Let’s go!</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FyusIrre_27E%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DyusIrre_27E&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FyusIrre_27E%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="640" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/4729ff6e1a8ebce28d6e50c6b265f3cf/href">https://medium.com/media/4729ff6e1a8ebce28d6e50c6b265f3cf/href</a></iframe><h4>How does Skeleton-of-Thought Work?</h4><p>Skeleton of Thought works in a two-step process to improve the quality of responses in Retrieval-Augmented Generation (RAG) applications, especially for longer documents.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Bvwybqvh4LSVHPNxEzWSCA.png" /></figure><p>The process begins with a prompt instructing the LLM to generate a concise outline of 3–10 brief points, each only 3–5 words long.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_f_6ZLRp_cqUyqZez6YnWQ.png" /><figcaption>The first prompt asks the LLM to create a skeleton of a list of short points to answer the question.</figcaption></figure><p>Next, the LLM is prompted to develop each point separately, writing very short, 1–2 sentence explanations for each. This allows the LLM to focus on one idea at a time, expanding them in parallel rather than sequentially.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZUFwHcaVVlWF8jUR0ESPAg.png" /><figcaption>The next prompt instructs the LLM to expand on a single point.</figcaption></figure><p>Finally, the separate, completed points are assembled into the final document according to the outline. The final response is usually more structured, coherent, and reliable, particularly for long-form documents.</p><p>Have 15 minutes? Lets’s build it in Langflow!</p><h4>Step 1: Install the Template</h4><p>First, make sure you have Langflow <a href="https://docs.langflow.org/getting-started-installation">installed and running</a>. Once Langflow is running on your browser, go to the <a href="https://www.notion.so/Build-a-Calendar-Agent-with-Composio-and-CrewAI-in-Langflow-1b9cdba1bd1b48c1813c8ebbbca4d6d9?pvs=21">Langflow Store</a> and download the ‘Skeleton Thought Chain’ template.</p><p><em>💡 You’ll need an API key from the Langflow store to look up and install the template on your local Langflow.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/952/1*L7-rpf9t7faEZ5-Apnc2PA.gif" /><figcaption>Install the Skeleton Thought Chain template from the Langflow store</figcaption></figure><p>Once the template is installed from the store, return to your local Langflow and open the flow.</p><h4>Step 2: The Skeleton-of-Thought Component</h4><p>The Text Input Component on the left side of the flow feeds a question as into a Skeleton-of-Thought (SoT) component. Click on the Text field to examine or customize the question.</p><blockquote><em>💡</em> Tip: Enter your OpenAI API key in the “API Key” field for each of the four OpenAI components in the flow.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/916/1*hmfaOIwKRlp-ozPS-zmkXw.gif" /><figcaption>Add your OpenAI API key and customize the question</figcaption></figure><p>Next, see the two prompts in Python code for the component: one for generating the skeleton (a short outline of 3–5 words per point) and another for expanding each point into more detailed content.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/808/1*fM_HojsQV00Z9MIWDMoQYw.gif" /><figcaption>Examine the prompts in the Skeleton-of-Thought component Python code</figcaption></figure><p>Now, you’re ready to run the flow and observe how the entire flow works.</p><h4>Step 3: Run the complete flow</h4><p>Click the Play button on the rightmost Chat Output component to execute the complete flow.</p><p>Examine the expanded results for each of the outline points by clicking on the Text buttons of each of the 3 OpenAI components.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/808/1*we-HFb9E8DjxAy7r6Z2Ekg.gif" /><figcaption>Examine the outline expansion results in each OpenAI component.</figcaption></figure><p>Lastly, click into the Combine Text component for the final assembled response that combines each of the outline detailed responses into your final long-form response to the original question.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/808/1*YjNJ-u2tpI-pJltzjVb_4w.gif" /><figcaption>The Combine Text component assembles the 3 expanded points into a single LLM response</figcaption></figure><h4>That’s all?</h4><p>Yep! You just ran a RAG chatbot flow that uses Skeleton-of-Thought — another advanced RAG technique.</p><p>Try experimenting with different questions to compare a basic response with one that first creates an outline and then generates the different sections in parallel. Do you see improvements in speed and relevance?</p><p>We hope this inspires you to keep learning about more <a href="https://medium.com/logspace/beyond-basic-rag-similarity-relevance-fdbd16373d4a">advanced RAG</a>, there’s a lot more we cover in other tutorials and blogs! 🚀</p><p>Thanks for supporting Langflow! Join the community on <a href="https://discord.com/invite/EqksyE2EX9">Discord</a> and <a href="https://github.com/langflow-ai/langflow">GitHub</a>!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6aa7e2e54cd1" width="1" height="1" alt=""><hr><p><a href="https://medium.com/logspace/build-an-llm-app-using-skeleton-of-thought-6aa7e2e54cd1">Build an LLM app using Skeleton-of-Thought</a> was originally published in <a href="https://medium.com/logspace">Langflow</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Build a Multi-Query RAG pipeline in Langflow ]]></title>
            <link>https://medium.com/logspace/build-a-multi-query-rag-pipeline-in-langflow-67d7d43b0699?source=rss-df3fd9625068------2</link>
            <guid isPermaLink="false">https://medium.com/p/67d7d43b0699</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[langchain]]></category>
            <category><![CDATA[retrieval-augmented-gen]]></category>
            <category><![CDATA[llm]]></category>
            <dc:creator><![CDATA[Scott Regan]]></dc:creator>
            <pubDate>Mon, 02 Sep 2024 16:22:02 GMT</pubDate>
            <atom:updated>2024-09-02T16:22:02.367Z</atom:updated>
            <content:encoded><![CDATA[<p>Retrieval-Augmented Generation (RAG) is an AI app development technique to use external content with large language models (LLMs) in order to generate relevant responses about data that is not part of it’s training.</p><p>For example, you can use RAG to build a chatbot that can answer questions from real-time customer or transaction data, or from internal documents or data sources.</p><p>But basic RAG applications can “hallucinate”, where the the LLM generates convincing, but incorrect or fabricated information. RAG apps with high hallucination rates are risky to put in front of real users!</p><p>To address this, <a href="https://medium.com/logspace/beyond-basic-rag-similarity-relevance-fdbd16373d4a">advanced RAG techniques</a> like <a href="https://arxiv.org/abs/2305.06983">FLARE</a>, <a href="https://arxiv.org/abs/2210.03629">ReAct</a>, and <a href="https://hackernoon.com/achieving-relevant-llm-responses-by-addressing-common-retrieval-augmented-generation-challenges">others</a> are evolving to reduce hallucinations. These methods typically use the LLM to generate multiple variations of the user question or multiple LLM responses in order to create more relevant answers.</p><p>In this example, we’ll walk through a Multi-Query RAG pipeline in Langflow. This fairly straightforward advanced RAG technique improves relevance by generating multiple query variations in order to create a richer context in the LLM prompt.</p><p>You can get a fast overview of this Multi-Query RAG example in the video below, and then let’s build it in about 20 minutes in Langflow!</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FJP_TRi-cowk%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DJP_TRi-cowk&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FJP_TRi-cowk%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="640" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/9b35d0952850ca9196cc128e7cb3430a/href">https://medium.com/media/9b35d0952850ca9196cc128e7cb3430a/href</a></iframe><p>Ready? Let’s do it!</p><h4>Simple RAG</h4><p>In a simple RAG pipeline, text from external documents, transactions or other sources are extracted and split into ‘chunks’ or text fragments of a predetermined length. These chunks are fed to an LLM, which ‘embeds’, or generates vectors, for each chunk. These vectors (called ‘embeddings’) are stored in a vector database along with the original text data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*O8kTQoOgvAVIjweu0qxSyA.png" /></figure><p>When the user asks a question, their query is also embedded into a vector which is used for a similarity search to find ‘nearest neighbor’ vectors in a database, representing text that is semantically similar to the users question.</p><p>These query results are converted back from vectors into text and then sent to an LLM as context in the prompt to generate the response.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HyyKXfmhRM7aRvtfbKZ8Zw.png" /></figure><h4>Multi-Query RAG</h4><p>In Multi-Query RAG, we take the process a step further. The LLM generates multiple queries or multiple variations of the user’s question.</p><p>These queries are then embedded and used to search the vector database. The multiple, different responses are combined to create more context in LLM prompt, likely giving a better response.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*077ge_pCiL95EdQIm9BVdQ.png" /></figure><h4>Step 1: Install the Template</h4><p>First, make sure you have Langflow <a href="https://docs.langflow.org/getting-started-installation">installed and running</a>. Once Langflow is running on your browser, go to the <a href="https://www.notion.so/Build-a-Calendar-Agent-with-Composio-and-CrewAI-in-Langflow-1b9cdba1bd1b48c1813c8ebbbca4d6d9?pvs=21">Langflow Store</a> and download the RAG Multi-Query template.</p><p><em>💡 You’ll need an API key from the Langflow store to look up and install the template on your local Langflow.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bO92LuJZRADhjyiPK0YS5Q.gif" /><figcaption>Install the Mulit-Query RAG template from the Langflow Store</figcaption></figure><p>Of course, you can also build the flow from an empty canvas by creating a new project and adding and connecting all the components, but starting with a template project will save you some time 🙂</p><h4>Step 2: Data Ingestion</h4><p>First, add this PDF of <a href="https://www.dropbox.com/scl/fi/lm3rbjnglaog1jr3wih8v/tipsforsurvivingcollege.pdf?rlkey=82rddsp5riqwhgqkw3y7h81iv&amp;dl=0">College Survival Tips</a> into the File Loader component as an external data source to vectorize.</p><p>Adjust the Split Text Component parameters to a chunk size of 50 and overlap of 20 characters. <strong><em>Important</em></strong>: add a single space character in the Separator field - the component will add it between chunks.</p><p>This flow uses the Cohere LLM to create the embeddings, so create a free account, add your API key.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BAwU0HwBRh18y4pVxG0MJg.gif" /><figcaption>Add the PDF to the File Component, Set the Chunking parameters and add your Cohere API key</figcaption></figure><p>Run the injestion pipeline by hitting Play on the Chroma DB Component. This builds and runs all the components to the right, creating the collection of vector embeddings in Chroma DB. Then click the Vector Store icon to see your vector embeddings (see below).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*VZgQqFrSS2vzi2AkBwxLWQ.gif" /></figure><h4>Step 3: Multi-Query Generation</h4><p>Next, let’s head up to the query flow in Langflow.</p><p>Modify the user question in the Chat Input component to something that is related to our external PDF we ingested above.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pM3SJnzL3uDkjMmBfEGFdQ.gif" /><figcaption>Modify the user question to be something related to the injested external document</figcaption></figure><p>Click the Template field of the Prompt Component and see the prompt asks the LLM to create three different versions of this question.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hulihi7LBzCWI0oryJzNwg.gif" /><figcaption>The first prompt asks the LLM to create multiple questions based on the input question field</figcaption></figure><h4>Step 4: Run Each Query and Combine Responses into Context</h4><p>Add your Cohere API key to each of the three Group Components, which were created by grouping the vector search flow above (see GIF below). Each Group runs a vector search for each of the questions.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xdetcjSTyGaZPz8O9NyhPg.gif" /><figcaption>Each question is embedded and submitted as a vector search (see the grouped flow above the main flow)</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gx-iJhOb9FycvhtUks24Ww.gif" /><figcaption>Run the full flow and examine each of the different responses resulting from the three vector search queries</figcaption></figure><p>Select the Template field in the Prompt Component to see the assembled context that will be used to generate the final response.</p><h4>Step 5: Submit the combined context for the final LLM response</h4><p>Run the complete pipeline from the OpenAI component at the end of the flow (this will run every component to the left) and select the Text field to see the LLM response. You can also run the Playground to try your workflow in the Chat Output interface as below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3W2mfScg5e8CYxF9y45e1g.gif" /><figcaption>Run the final flow in the Playground for the LLM response to the original user question.</figcaption></figure><h3>Congrats!</h3><p>You just built a Multi-Query RAG pipeline — an advanced RAG technique that uses the LLM to create multiple varieties of the original question to create a more accurate LLM response.</p><p>We hope this inspires you to keep learning about more <a href="https://medium.com/logspace/beyond-basic-rag-similarity-relevance-fdbd16373d4a">advanced RAG techniques</a>, there’s a lot more we will cover soon! 🚀</p><p>Thanks for supporting Langflow! Join the community on <a href="https://discord.com/invite/EqksyE2EX9">Discord</a> and <a href="https://github.com/langflow-ai/langflow">GitHub</a>!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=67d7d43b0699" width="1" height="1" alt=""><hr><p><a href="https://medium.com/logspace/build-a-multi-query-rag-pipeline-in-langflow-67d7d43b0699">Build a Multi-Query RAG pipeline in Langflow 🚀</a> was originally published in <a href="https://medium.com/logspace">Langflow</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Build a Calendar Agent with Composio and CrewAI in Langflow]]></title>
            <link>https://medium.com/logspace/build-a-calendar-agent-with-composio-and-crewai-in-langflow-bc8a3244eea3?source=rss-df3fd9625068------2</link>
            <guid isPermaLink="false">https://medium.com/p/bc8a3244eea3</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[flowise]]></category>
            <category><![CDATA[langchain]]></category>
            <category><![CDATA[langchain-agents]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Scott Regan]]></dc:creator>
            <pubDate>Wed, 21 Aug 2024 20:44:18 GMT</pubDate>
            <atom:updated>2024-08-23T00:06:31.241Z</atom:updated>
            <content:encoded><![CDATA[<h3>Agent-Driven Schedule</h3><h4>Build a calendar scheduling agent with Composio and CrewAI in Langflow</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2rXsOT0c0AT8uvGHGpUDeQ.png" /></figure><p>Agents are AI applications that can execute complex tasks, combining LLMs with functionality such as planning, memory, and external API access.</p><p>Agents take their name from their “agency”, or ability to make decisions and take actions (for example, choosing to use an API over another, reading a specific page of a document, or accessing a website).</p><p>Agents achieve their tasks by using tools. Tools are capabilities, like a function or an API operation for a common service. Agents use these tools based on triggers, like a user request or an event that prompts the agent to take action.</p><p>A single agent operates like a decision-maker, similar to a person handling a task alone while multi-agent systems involve a series of agents working together to solve complex problems, often modeled like a team of people that ‘divide and conquer’ for more accurate results.</p><p>In this guide, we’ll walk you through a simple agent built on Langflow that can read from your calendar and create new appointments automatically.</p><p>This example uses a <a href="http://composio.dev/">Composio</a> integration with Google Calendar as a tool for <a href="http://crewai.com/">CrewAI</a> agents to manage your calendar based on a chat conversation.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2Fxt-C3jy2YTU%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3Dxt-C3jy2YTU&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2Fxt-C3jy2YTU%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="640" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/c0f29538fe7fe7960c81a69625670f5d/href">https://medium.com/media/c0f29538fe7fe7960c81a69625670f5d/href</a></iframe><p>Ready? Let’s go!</p><h3>Step 1: Set up a Google Calendar integration</h3><p>First, create a free Composio account and select the Google Calendar integration from the Composio Tools Gallery.</p><blockquote>💡 <a href="http://compsio.dev/">Composio</a> is an agentic integration platform that enables LLMs/AI agents to interact with external services via function calling. Using Composio, you can connect AI agents to tools via APIs and manage user-level authentication.</blockquote><p>Set up a Google Calendar integration by signing into your Google Account (tip: don’t forget to check the boxes for access before selecting ‘enable’).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*v0bsp1dqPc_gVqOQBS20cA.gif" /><figcaption>Create a Google Calendar integration in Composio</figcaption></figure><p>Make sure you get a confirmation that the integration is successful, then head over to the Settings tab and copy your API key for the next step.</p><h3>Step 2: Install Langflow and download the template</h3><p>Next, make sure you have Langflow version 1.0.15 or later <a href="https://docs.langflow.org/getting-started-installation">installed and running</a>. Once Langflow is running on your browser, go to the <a href="https://www.notion.so/Build-a-Calendar-Agent-with-Composio-and-CrewAI-in-Langflow-1b9cdba1bd1b48c1813c8ebbbca4d6d9?pvs=21">Langflow Store</a> and download the <strong>Gmail Calendar Scheduling Agent</strong> template.</p><blockquote>💡 You’ll need an API key from the Langflow store to look up and install the template on your local Langflow.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xgm0-8CWspS2fF-uCEQKng.gif" /><figcaption>Install the template of this agent example from the Langflow Store</figcaption></figure><p>You could also build the flow from an empty canvas by creating a new project and adding and connecting all the components, but starting with a template project will save you time!</p><h3>Step 3: Equip the CrewAI agent with the Google Calendar tool</h3><p>Open the template workflow, find the Composio Google Calendar component, and add your Composio API key. On the actions drop-down, select the actions for (1) find calendar events, (2) create calendar events, and (3) get the current date and time.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1oy-BvpgShNN8lcYRMZdEw.gif" /><figcaption>Give the Composio component your API key and select the find, create, and date/time actions</figcaption></figure><h3>Step 4: Give the agent a goal and a task</h3><p>You’ve now equipped the CrewAI Agent component with the Google Calendar tools and actions. Notice that we’ve already given the agent a role description and backstory in their respective component fields.</p><blockquote>💡 <a href="https://www.notion.so/Build-a-Calendar-Agent-with-Composio-and-CrewAI-in-Langflow-1b9cdba1bd1b48c1813c8ebbbca4d6d9?pvs=21">CrewAI</a> is a framework for orchestrating role-playing, autonomous AI agents that work together to tackle complex tasks. Langflow includes components for CrewAI agents and tasks.</blockquote><p>The agent connects to a CrewAI Sequential task component, which defines the task to be performed. Sequential tasks require the agent to complete one task before beginning the next one.</p><p>We give the agent a task description through the text field of a Chat Input component. The Chat Input sends this to the Prompt component as a variable, which in turn prompts the Sequential Task component to execute the task.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zeIEOiSjaMNcS-a4nCjbew.gif" /><figcaption>Give the agent a specific task in the Chat Input prompt</figcaption></figure><p>You can modify the task you want the agent to perform by changing the Chat Input text field. For example, try asking the agent to search for a certain event on your calendar or to create a new calendar appointment with a title, date, and time.</p><h3>Step 5: Pick a model and run the agent!</h3><p>We’re almost there! Add your OpenAI API key and select a model such as GPT-4o.</p><p>Run the entire flow by clicking the Play button on the Chat Output component and wait for the project to build and save. Inspect the output text and see the results of your agent’s task.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wEjDtb_D8vtKe6OUrq9UOQ.gif" /></figure><h3>Congrats!</h3><p>You just built a Scheduling Agent that can read, analyze, and update your calendar — all with a few simple steps in Langflow!</p><p>We hope this inspires you to keep building and thinking of creative agent applications to automate tasks and free up time for what really matters. 🚀</p><p>Thanks for supporting Langflow! Join the community on <a href="https://discord.com/invite/EqksyE2EX9">Discord</a> and <a href="https://github.com/langflow-ai/langflow">GitHub</a>!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bc8a3244eea3" width="1" height="1" alt=""><hr><p><a href="https://medium.com/logspace/build-a-calendar-agent-with-composio-and-crewai-in-langflow-bc8a3244eea3">Build a Calendar Agent with Composio and CrewAI in Langflow</a> was originally published in <a href="https://medium.com/logspace">Langflow</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Langflow Pre-Release v1.0.0rc0]]></title>
            <link>https://medium.com/logspace/langflow-pre-release-v1-0-0rc0-9566b7ac6694?source=rss-df3fd9625068------2</link>
            <guid isPermaLink="false">https://medium.com/p/9566b7ac6694</guid>
            <category><![CDATA[llm-applications]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[langchain]]></category>
            <category><![CDATA[developer-tools]]></category>
            <dc:creator><![CDATA[Scott Regan]]></dc:creator>
            <pubDate>Sun, 23 Jun 2024 03:11:35 GMT</pubDate>
            <atom:updated>2024-06-23T03:11:35.578Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Feature complete, stability, and control… check! ✅. Ready for Takeoff</strong> 🚀</p><p>We are almost there! Langflow v1.0.0rc0 is our final pre-release. Today we’re releasing a feature-complete update with all the new components, features, and flows we’ve been working towards, with much improved stability. It is a big step towards our vision of creating the most powerful visual framework for building multi-agent and RAG applications out there — open-source, Python-powered, and fully agnostic. The official 1.0 release is very close!</p><p><strong>The Vision</strong></p><p>Langflow 1.0 isn’t just a showroom of pre-built AI components; it’s the manufacturing plant! This release delivers the fundamental building blocks required to create, customize, and share your own reusable AI components. It brings <strong>control</strong> over advanced AI pipelines with a variety of models, data sources, and APIs.</p><p>Let’s dive into what’s new and why it matters! 🤿</p><p><strong>Same Component, Multiple Outputs</strong></p><p>Components can now have more than a single output, allowing for unique flexibility in creating complex flows. The game-changer is output routing — it allows for so many new capabilities it’s almost silly to describe!</p><ol><li>Branch to one or more subsequent components;</li><li>Apply logic operations like if/else and exploit decision-making;</li><li>Create classification models that choose between paths;</li><li>Enable the development of agent architectures from scratch;</li><li>Build an orchestrator that routes between agents.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YiAtiXXL6oM31wucDRikkg.png" /><figcaption>Components can now have multiple outputs</figcaption></figure><p><strong>Flow + Composition</strong></p><p>Some components run in real-time, while others take their time… Langflow brings the best of both flow and composition, allowing components to work with both immediate execution and modular building. This means that easy flows are <em>really</em> easy and the complex ones are still possible!</p><ul><li><strong>Flow:</strong> Data is processed by one component and then passed to the next component in line for immediate execution.</li><li><strong>Composition</strong>: Allows components to not only forward data but also share states for modular building.</li></ul><p>For example, a flow can sequentially process text, and after a few steps, trigger an agent. It can access functions that wait to be called or to respond. This blend of flow and composition brings an unprecedented level of flexibility and control to data workflows in LLM-based apps and agents that use multiple models and APIs working together to achieve tasks.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*V6QdyeKEtyJYvJR8vjEjhg.png" /><figcaption>Components can forward data, state, and access functions</figcaption></figure><p><strong>Output Preview</strong></p><p>Each component now includes an output visualizer that opens a pop-up screen, allowing you to easily inspect and monitor transmissions between components. It provides instant feedback on your workflows, letting you see results as they are processed. 🔍</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KqnoDiH_mbBffGBGLPcSiw.gif" /><figcaption>Preview the output of a component to inspect results at each step</figcaption></figure><p><strong>Inputs and Outputs Handling</strong></p><p>Inputs and outputs finally make more sense to us, and hopefully to you too.</p><p>We’re proposing Langflow-native types to keep things consistent, but not limited to use any Python type. For instance, a Chat Input component sends out what we call a Message object (text + metadata like date, time, and sender), but maybe you want to introduce external types from your favorite Python package? Go wild. Each native type will have their own visualization modes and will evolve according to new integrations added.</p><p>And there’s so much more — an entirely redesigned code structure for custom components, multi-modal capabilities and even editable memory sessions! We’re working 24/7 to finalize everything and we’re beyond excited to have you on this journey with us. Keep sending your feedback on <a href="https://github.com/langflow-ai/langflow">GitHub</a> and <a href="https://bit.ly/langflow-discord">Discord</a>, and thanks for being a part of this! 🌟</p><p>Sincerely,</p><p><strong>The Langflow Team 🚀</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9566b7ac6694" width="1" height="1" alt=""><hr><p><a href="https://medium.com/logspace/langflow-pre-release-v1-0-0rc0-9566b7ac6694">Langflow Pre-Release v1.0.0rc0</a> was originally published in <a href="https://medium.com/logspace">Langflow</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Langflow + Notion ]]></title>
            <link>https://medium.com/logspace/langflow-notion-efd56b9896e6?source=rss-df3fd9625068------2</link>
            <guid isPermaLink="false">https://medium.com/p/efd56b9896e6</guid>
            <dc:creator><![CDATA[Scott Regan]]></dc:creator>
            <pubDate>Mon, 27 May 2024 22:14:22 GMT</pubDate>
            <atom:updated>2024-05-28T17:40:49.037Z</atom:updated>
            <content:encoded><![CDATA[<h4><strong>Streamline RAG &amp; Agents with Notion components for Langflow</strong></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4zLbQBZXJGmfG8j8h9E9ug.jpeg" /></figure><p>Like many software development teams, we ❤️‍ <a href="https://www.notion.so/">Notion</a>!</p><p>We use it to collaborate and organize documents, tasks, transcripts, and pretty much everything.</p><p>So we’re excited to have our first Notion component bundle for Langflow, thanks to our friends at <a href="https://x.com/NamastexLabs/status/1793771289330716953">NamastexLabs</a>. You can find and download it from the <a href="https://www.langflow.store/store/f891405e-cd80-408a-bbc1-34c1facbafda/">Langflow Store</a> along with <a href="https://pre-release.langflow.org/integrations/notion/intro">documentation</a>.</p><h3>Level Up with the Notion Component Bundle</h3><p>Imagine using LLMs to search, create and update Notion content from a Langflow data flow or application.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FPH3nw9zCPcc%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DPH3nw9zCPcc&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FPH3nw9zCPcc%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/bbe56eda4e1f12398a90e76fe9c02a1a/href">https://medium.com/media/bbe56eda4e1f12398a90e76fe9c02a1a/href</a></iframe><h3>So what can I do with Notion in Langflow?</h3><p>You’ll find a Langflow component for several Notion operations including:</p><ul><li>🔍 <strong>Searches:</strong> Search pages and data as well databases to determine the best one for a task.</li><li>➕ <strong>Create Page</strong>: Create a new Notion page.</li><li>📚 <strong>List Pages:</strong> Retrieve a list of pages and access data in your Notion workspace.</li><li>👥 <strong>List Users:</strong> Retrieve a list of users with access to a Notion workspace.</li><li>✍️ <strong>Add Page Content:</strong> Make it easy to update pages programmatically.</li><li>🧮 <strong>List Database Properties:</strong> Query properties of a database structure and metadata.</li><li>📃 <strong>Page Content Viewer:</strong> Retrieve page content based on a page id.</li></ul><h3>Build New Use Cases with Notion and Langflow</h3><p>Here are just a few ideas of what you could build:</p><ul><li><strong>🚀 Automate Task Creation:</strong> Use AI to create and update tasks from simple descriptions.</li><li>🎧 <strong>Extract Meeting Context:</strong> Analyze meeting details and update pages with key points.</li><li>📊 <strong>Generate Reports:</strong> Compile and update reports in Notion based on real-time data.</li><li>🔬 <strong>Database Insights:</strong> Retrieve database properties to understand Notion structures.</li><li>👤 <strong>User Management:</strong> List and manage users with access to your Notion workspace.</li></ul><p><strong>Ready to Get Started?</strong></p><p>You can find the Notion Component <a href="https://www.langflow.store/store/f891405e-cd80-408a-bbc1-34c1facbafda/">bundle here</a> and <a href="https://pre-release.langflow.org/integrations/notion/intro">docs here</a>. We would love to hear your feedback and suggestions for new components and use cases — Get started with <a href="https://bit.ly/langflow_repo">Langflow on GitHub</a> and join our <a href="https://bit.ly/langflow-discord">Discord</a> server! 🌟</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=efd56b9896e6" width="1" height="1" alt=""><hr><p><a href="https://medium.com/logspace/langflow-notion-efd56b9896e6">Langflow + Notion 🎈</a> was originally published in <a href="https://medium.com/logspace">Langflow</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[New Release — Langflow 1.0.35a]]></title>
            <link>https://medium.com/logspace/new-release-langflow-1-0-35a-b1cdf3c8181f?source=rss-df3fd9625068------2</link>
            <guid isPermaLink="false">https://medium.com/p/b1cdf3c8181f</guid>
            <category><![CDATA[generative-ai-tools]]></category>
            <category><![CDATA[ai-agent-development]]></category>
            <category><![CDATA[llm-agent]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[retrieval-augmented-gen]]></category>
            <dc:creator><![CDATA[Scott Regan]]></dc:creator>
            <pubDate>Wed, 22 May 2024 00:44:08 GMT</pubDate>
            <atom:updated>2024-05-22T00:44:08.775Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cAnfTaQXJv3PNMCpiMD-ZA.png" /></figure><h3>New Release — Langflow 1.0.35a</h3><p>As we get closer to our 1.0 release, Langflow is starting to resemble a true “product” with the addition of new features, components, and capabilities.</p><p>Our goal is to provide greater control over Language Models. For example, how can a developer determine the input and output details within the nuances of LLM pipelines, and visualize what is happening behind the scenes?</p><p>We think this will drastically change the way we interact with AI, and we are excited to roll out some of these changes today with our new alpha release — <a href="https://github.com/langflow-ai/langflow/releases"><strong>Langflow 1.0.35a</strong></a></p><h3>📃 Logs &amp; Monitoring</h3><p>A new logs page has been added! Now, both component executions and message history from the chat can be inspected in an interactive table. This will make it easier to debug, inspect, and manage messages passing through components.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*UyMAd4-vpu-yjtmmQdmJYA.png" /><figcaption>Logs enable you to inspect messages and executions.</figcaption></figure><h3>📁 Folders</h3><p>We introduced folders on the home page to help categorize flows and components. Create, remove and edit them to keep your work organized. 🗄️</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*onsysBNGmDceNY6UYDrNqw.png" /><figcaption>Organize your flows and components with folders</figcaption></figure><h3>📋 Table View</h3><p>A <strong>Table</strong> output view has been added to Langflow. Records can now be visualized as a table in the <strong>Playground</strong>, perfect for inspecting structured information intuitively and live.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*s2l8tM_l63AF4PK_AJsFMw.png" /><figcaption>Visualize data in the Langflow Playground</figcaption></figure><p>See our<strong> </strong><a href="https://github.com/langflow-ai/langflow/releases"><strong>Release Notes</strong></a> for more new features and capabilities, including:</p><ul><li>Python 3.12 compatibility</li><li>Select Box for bulk actions</li><li>Support for OpenAI GPT4o</li><li>Selectable flows and components</li><li>New inputs and outputs</li><li>New experimental components</li></ul><p>Langflow continues to bridge the gap between code and no-code, inviting you to redefine what’s possible with AI!</p><p>If you want to support Langflow, please <a href="https://github.com/langflow-ai/langflow">star our GitHub repo!</a> 🌟</p><p>Happy Building,</p><p>The Langflow Team</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b1cdf3c8181f" width="1" height="1" alt=""><hr><p><a href="https://medium.com/logspace/new-release-langflow-1-0-35a-b1cdf3c8181f">New Release — Langflow 1.0.35a</a> was originally published in <a href="https://medium.com/logspace">Langflow</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>