https://pythonflow.com/tags/lcel/ Recent content in Lcel on Pythonflow Hugo en-us alicandonmez90@gmail.com (flowfelis) alicandonmez90@gmail.com (flowfelis) Sat, 27 Jun 2026 07:00:00 +0100 https://pythonflow.com/posts/langbot-day-10-agents/ Sat, 27 Jun 2026 07:00:00 +0100alicandonmez90@gmail.com (flowfelis) https://pythonflow.com/posts/langbot-day-10-agents/ <p><strong>Recap:</strong> On Day 9, LangBot became a full RAG pipeline — every chat message retrieves relevant documents and the LLM answers from them. The <code>/search</code> command, list-mode routing, and regular chat all coexist. But there is a problem hiding in plain sight: the routing logic is hand-coded in <code>if</code>/<code>elif</code> blocks. LangBot doesn&rsquo;t <em>decide</em> anything — it follows a script. Today we replace that script with an <strong>agent</strong>: an LLM that reasons about which tool to use, calls it, and responds — autonomously.</p> https://pythonflow.com/posts/langbot-day-9-rag-pipeline/ Fri, 26 Jun 2026 07:00:00 +0100alicandonmez90@gmail.com (flowfelis) https://pythonflow.com/posts/langbot-day-9-rag-pipeline/ <p><strong>Recap:</strong> On Day 8, LangBot gained a long-term memory — a vector store (Pinecone) filled with document embeddings. You can ask <code>/search home office stipend</code> and get relevant chunks back. But that is half the picture. You still have to type <code>/search</code>, read raw chunks, and synthesize the answer yourself. Today we close the loop: the retriever feeds into the chatbot automatically, so you ask <em>&ldquo;How much is the home office stipend?&rdquo;</em> and LangBot answers from your documents — no slash command, no manual lookup.</p> https://pythonflow.com/posts/langbot-day-7-streaming/ Tue, 23 Jun 2026 07:00:00 +0100alicandonmez90@gmail.com (flowfelis) https://pythonflow.com/posts/langbot-day-7-streaming/ <p><strong>Recap:</strong> On Day 6, LangBot gained output parsers and now responds in two modes — chat mode returns typed <code>LangBotResponse</code> objects with mood and confidence, and list mode returns clean <code>list[str]</code> via <code>CommaSeparatedListOutputParser</code>. Both modes use <code>.invoke()</code>, which means the entire response arrives at once after the model finishes generating. Today we fix that — LangBot learns to type its replies one token at a time.</p> <hr> <h2 id="the-problem-waiting-for-the-whole-answer">The problem: waiting for the whole answer</h2> <p>Every interaction so far has the same rhythm:</p> https://pythonflow.com/posts/langbot-day-6-output-parsers/ Mon, 22 Jun 2026 07:00:00 +0100alicandonmez90@gmail.com (flowfelis) https://pythonflow.com/posts/langbot-day-6-output-parsers/ <p><strong>Recap:</strong> On Day 5, LangBot started returning typed objects — every response is now a <code>LangBotResponse</code> with <code>.message</code>, <code>.mood</code>, and <code>.confidence</code>. That is great when you need structured metadata. But structured output is not always the right tool. Sometimes you just want a clean string. Sometimes you want a parsed list. Today we add output parsers — lightweight LCEL components that transform raw <code>AIMessage</code> output into exactly the shape you need.</p> https://pythonflow.com/posts/langbot-day-4-lcel/ Sat, 20 Jun 2026 07:00:00 +0100alicandonmez90@gmail.com (flowfelis) https://pythonflow.com/posts/langbot-day-4-lcel/ <p><strong>Recap:</strong> On Day 1 we built a basic CLI chatbot. Day 2 gave it memory so it remembers the conversation. Day 3 added a prompt template so LangBot has a personality. But every turn still requires two separate steps — format the template, then call the model. Today we stitch them together into one fluid pipeline.</p> <hr> <h2 id="the-problem-two-steps-where-one-should-do">The problem: two steps where one should do</h2> <p>Here is how LangBot works on Day 3. Every single turn:</p>