#543: Deep Agents: LangChain's SDK for Agents That Plan and Delegate

When you type a question into ChatGPT, the model only has what you typed to work with. But tools like Claude Code can plan, iterate, test, and recover from mistakes. They work more like we do. The difference is the agent harness: Planning tools, file system access, sub-agents, and carefully crafted system prompts that turn a raw LLM into something genuinely capable. Sydney Runkle is back on Talk Python representing LangChain and their new open source library, Deep Agents: A framework for building your own deep agents with plain Python functions, middleware hooks, and MCP support. This is how the magic works under the hood. Episode sponsors Sentry Error Monitoring, Code talkpython26 Temporal Talk Python Courses Links from the show Guest Sydney Runkle: github.com Claude Code uses: x.com Deep Research: openai.com Manus: manus.im Blog post announcement: blog.langchain.com Claudes system prompt: github.com sub agents: docs.anthropic.com the quick start: docs.langchain.com CLIs: github.com Talk Python's CLI: talkpython.fm custom tools: docs.langchain.com DeepAgents Examples: github.com Custom Middleware: docs.langchain.com Built in middleware: docs.langchain.com Improving Deep Agents with harness engineering: blog.langchain.com Prebuilt middleware: docs.langchain.com Watch this episode on YouTube: youtube.com Episode #543 deep-dive: talkpython.fm/543 Episode transcripts: talkpython.fm Theme Song: Developer Rap 🥁 Served in a Flask 🎸: talkpython.fm/flasksong ---== Don't be a stranger ==--- YouTube: youtube.com/@talkpython Bluesky: @talkpython.fm Mastodon: @[email protected] X.com: @talkpython Michael on Bluesky: @mkennedy.codes Michael on Mastodon: @[email protected] Michael on X.com: @mkennedy

If you've built documentation in the Python ecosystem, chances are you've used Martin Donath's work. His Material for MKDocs powers docs for FastAPI, uv, AWS, OpenAI, and tens of thousands of other projects. But when MKDocs 2.0 took a direction that would break Material and 300 ecosystem plugins, Martin went back to the drawing board. The result is Zensical: A new static site generator with a Rust core, differential builds in milliseconds instead of minutes, and a migration path designed to bring the whole community along. Episode sponsors Sentry Error Monitoring, Code talkpython26 Talk Python Courses Links from the show Guest Martin Donath: github.com Zensical: zensical.org Material for MkDocs: squidfunk.github.io Getting Started: zensical.org Github pages: docs.github.com Cloudflare pages: pages.cloudflare.com Michaels Example: gist.github.com Material for MkDocs: zensical.org gohugo.io/content-management/shortcodes: gohugo.io a sense of size of the project: blobs.talkpython.fm Zensical Spark: zensical.org Watch this episode on YouTube: youtube.com Episode #542 deep-dive: talkpython.fm/542 Episode transcripts: talkpython.fm Theme Song: Developer Rap 🥁 Served in a Flask 🎸: talkpython.fm/flasksong ---== Don't be a stranger ==--- YouTube: youtube.com/@talkpython Bluesky: @talkpython.fm Mastodon: @[email protected] X.com: @talkpython Michael on Bluesky: @mkennedy.codes Michael on Mastodon: @[email protected] Michael on X.com: @mkennedy

When LLMs write code to accomplish a task, that code has to actually run somewhere. And right now, the options aren't great. Spin up a sandboxed container and you're paying a full second of cold start overhead plus the complexity of another service. Let the LLM loose on your actual machine and... well, you'd better be watching. On this episode, I sit down with Samuel Colvin, creator of Pydantic, now at 10 billion downloads, to explore Monty, a Python interpreter written from scratch in Rust, purpose-built to run LLM-generated code. It starts in microseconds, is completely sandboxed by design, and can even serialize its entire state to a database and resume later. We dig into why this deliberately limited interpreter might be exactly what the AI agent era needs. Episode sponsors Talk Python Courses Python in Production Links from the show Guest Samuel Colvin: github.com CPython: github.com IronPython: ironpython.net Jython: www.jython.org Pyodide: pyodide.com monty: github.com Pydantic AI: pydantic.dev Python AI conference: pyai.events bashkit: github.com just-bash: github.com Narwhals: narwhals-dev.github.io Polars: pola.rs Strands Agents: aws.amazon.com Subscribe Running Pydantic’s Monty Rust sandboxed Python subset in WebAssembly: simonwillison.net Rust Python: github.com Valgrind: valgrind.org Cod Speed: codspeed.io Watch this episode on YouTube: youtube.com Episode #541 deep-dive: talkpython.fm/541 Episode transcripts: talkpython.fm Theme Song: Developer Rap 🥁 Served in a Flask 🎸: talkpython.fm/flasksong ---== Don't be a stranger ==--- YouTube: youtube.com/@talkpython Bluesky: @talkpython.fm Mastodon: @[email protected] X.com: @talkpython Michael on Bluesky: @mkennedy.codes Michael on Mastodon: @[email protected] Michael on X.com: @mkennedy

Monorepos -- you've heard the talks, you've read the blog posts, maybe you've seen a few tantalizing glimpses into how Google or Meta organize their massive codebases. But it's often in the abstract and behind closed doors. What if you could crack open a real, production monorepo, one with over a million lines of Python and over 100 of sub-packages, and actually see how it's built, step by step, using modern tools and standards? That's exactly what Apache Airflow gives us. On this episode, I sit down with Jarek Potiuk and Amogh Desai, two of Airflow's top contributors, to go inside one of the largest open-source Python monorepos in the world and learn how they manage it with uv, pyproject.toml, and the latest packaging standards, so you can apply those same patterns to your own projects. Episode sponsors Agentic AI Course Python in Production Talk Python Courses Links from the show Guests Amogh Desai: github.com Jarek's GitHub: github.com definition of a monorepo: monorepo.tools airflow: airflow.apache.org Activity: github.com OpenAI: airflowsummit.org Part 1. Pains of big modular Python projects: medium.com Part 2. Modern Python packaging standards and tools for monorepos: medium.com Part 3. Monorepo on steroids - modular prek hooks: medium.com Part 4. Shared “static” libraries in Airflow monorepo: medium.com PEP-440: peps.python.org PEP-517: peps.python.org PEP-518: peps.python.org PEP-566: peps.python.org PEP-561: peps.python.org PEP-660: peps.python.org PEP-621: peps.python.org PEP-685: peps.python.org PEP-723: peps.python.org PEP-735: peps.python.org uv: docs.astral.sh uv workspaces: blobs.talkpython.fm prek.j178.dev: prek.j178.dev your presentation at FOSDEM26: fosdem.org Tallyman: github.com Watch this episode on YouTube: youtube.com Episode #540 deep-dive: talkpython.fm/540 Episode transcripts: talkpython.fm Theme Song: Developer Rap 🥁 Served in a Flask 🎸: talkpython.fm/flasksong ---== Don't be a stranger ==--- YouTube: youtube.com/@talkpython Bluesky: @talkpython.fm Mastodon: @[email protected] X.com: @talkpython Michael on Bluesky: @mkennedy.codes Michael on Mastodon: @[email protected] Michael on X.com: @mkennedy

You're adding type hints to your Python code, your editor is happy, autocomplete is working great. But then you switch tools and suddenly there are red squiggles everywhere. Who decides what a float annotation actually means? Or whether passing None where an int is expected should be an error? It turns out there's a five-person council dedicated to exactly these questions -- and two brand-new Rust-based type checkers are raising the bar. On this episode, I sit down with three members of the Python Typing Council -- Jelle Zijlstra, Rebecca Chen, and Carl Meyer -- to learn how the type system is governed, where the spec and the type checkers agree and disagree, and get the council's official advice on how much typing is just enough. Episode sponsors Sentry Error Monitoring, Code talkpython26 Agentic AI Course Talk Python Courses Links from the show Guests Carl Meyer: github.com Jelle Zijlstra: jellezijlstra.github.io Rebecca Chen: github.com Typing Council: github.com typing.python.org: typing.python.org details here: github.com ty: docs.astral.sh pyrefly: pyrefly.org conformance test suite project: github.com typeshed: github.com Stub files: mypy.readthedocs.io Pydantic: pydantic.dev Beartype: github.com TOAD AI: github.com PEP 747 – Annotating Type Forms: peps.python.org PEP 724 – Stricter Type Guards: peps.python.org Python Typing Repo (PRs and Issues): github.com Watch this episode on YouTube: youtube.com Episode #539 deep-dive: talkpython.fm/539 Episode transcripts: talkpython.fm Theme Song: Developer Rap 🥁 Served in a Flask 🎸: talkpython.fm/flasksong ---== Don't be a stranger ==--- YouTube: youtube.com/@talkpython Bluesky: @talkpython.fm Mastodon: @[email protected] X.com: @talkpython Michael on Bluesky: @mkennedy.codes Michael on Mastodon: @[email protected] Michael on X.com: @mkennedy

Digital humanities sounds niche, until you realize it can mean a searchable archive of U.S. amendment proposals, Irish folklore, or pigment science in ancient art. Today I’m talking with David Flood from Harvard’s DARTH team about an unglamorous problem: What happens when the grant ends but the website can’t. His answer, static sites, client-side search, and sneaky Python. Let’s dive in. Episode sponsors Sentry Error Monitoring, Code talkpython26 Command Book Talk Python Courses Links from the show Guest David Flood: davidaflood.com DARTH: digitalhumanities.fas.harvard.edu Amendments Project: digitalhumanities.fas.harvard.edu Fionn Folklore Database: fionnfolklore.org Mapping Color in History: iiif.harvard.edu Apatosaurus: apatosaurus.io Criticus: github.com github.com/palewire/django-bakery: github.com sigsim.acm.org/conf/pads/2026/blog/artifact-evaluation: sigsim.acm.org Hugo: gohugo.io Water Stories: waterstories.fas.harvard.edu Tsumeb Mine Notebook: tmn.fas.harvard.edu Dharma and Punya: dharmapunya2019.org Pagefind library: pagefind.app django_webassembly: github.com Astro Static Site Generator: astro.build PageFind Python Lib: pypi.org Frozen-Flask: frozen-flask.readthedocs.io Watch this episode on YouTube: youtube.com Episode #538 deep-dive: talkpython.fm/538 Episode transcripts: talkpython.fm Theme Song: Developer Rap 🥁 Served in a Flask 🎸: talkpython.fm/flasksong ---== Don't be a stranger ==--- YouTube: youtube.com/@talkpython Bluesky: @talkpython.fm Mastodon: @[email protected] X.com: @talkpython Michael on Bluesky: @mkennedy.codes Michael on Mastodon: @[email protected] Michael on X.com: @mkennedy

You love building web apps with Python, and HTMX got you excited about the hypermedia approach -- let the server drive the HTML, skip the JavaScript build step, keep things simple. But then you hit that last 10%: You need Alpine.js for interactivity, your state gets out of sync, and suddenly you're juggling two unrelated libraries that weren't designed to work together. What if there was a single 11-kilobyte framework that gave you everything HTMX and Alpine do, and more, with real-time updates, multiplayer collaboration out of the box, and performance so fast you're actually bottlenecked by the monitor's refresh rate? That's Datastar. On this episode, I sit down with its creator Delaney Gillilan, core maintainer Ben Croker, and Datastar convert Chris May to explore how this backend-driven, server-sent-events-first framework is changing the way full-stack developers think about the modern web. Episode sponsors Sentry Error Monitoring, Code talkpython26 Command Book Talk Python Courses Links from the show Guests Delaney Gillilan: linkedin.com Ben Croker: x.com Chris May: everydaysuperpowers.dev Datastar: data-star.dev HTMX: htmx.org AlpineJS: alpinejs.dev Core Attribute Tour: data-star.dev data-star.dev/examples: data-star.dev github.com/starfederation/datastar-python: github.com VSCode: marketplace.visualstudio.com OpenVSX: open-vsx.org PyCharm/Intellij plugin: plugins.jetbrains.com data-star.dev/datastar_pro: data-star.dev gg: discord.gg HTML-ivating your Django web app's experience with HTMX, AlpineJS, and streaming HTML - Chris May: www.youtube.com Senior Engineer tries Vibe Coding: www.youtube.com 1 Billion Checkboxes: checkboxes.andersmurphy.com Game of life example: example.andersmurphy.com Watch this episode on YouTube: youtube.com Episode #537 deep-dive: talkpython.fm/537 Episode transcripts: talkpython.fm Theme Song: Developer Rap 🥁 Served in a Flask 🎸: talkpython.fm/flasksong ---== Don't be a stranger ==--- YouTube: youtube.com/@talkpython Bluesky: @talkpython.fm Mastodon: @[email protected] X.com: @talkpython Michael on Bluesky: @mkennedy.codes Michael on Mastodon: @[email protected] Michael on X.com: @mkennedy

You've built your FastAPI app, it's running great locally, and now you want to share it with the world. But then reality hits -- containers, load balancers, HTTPS certificates, cloud consoles with 200 options. What if deploying was just one command? That's exactly what Sebastian Ramirez and the FastAPI Cloud team are building. On this episode, I sit down with Sebastian, Patrick Arminio, Savannah Ostrowski, and Jonathan Ehwald to go inside FastAPI Cloud, explore what it means to build a "Pythonic" cloud, and dig into how this commercial venture is actually making FastAPI the open-source project stronger than ever. Episode sponsors Command Book Python in Production Talk Python Courses Links from the show Guests Sebastián Ramírez: github.com Savannah Ostrowski: github.com Patrick Arminio: github.com Jonathan Ehwald: github.com FastAPI labs: fastapilabs.com quickstart: fastapicloud.com an episode on diskcache: talkpython.fm Fastar: github.com FastAPI: The Documentary: www.youtube.com Tailwind CSS Situation: adams-morning-walk.transistor.fm FastAPI Job Meme: fastapi.meme Migrate an Existing Project: fastapicloud.com Join the waitlist: fastapicloud.com Talk Python CLI Talk Python CLI Announcement: talkpython.fm Talk Python CLI GitHub: github.com Command Book Download Command Book: commandbookapp.com Announcement post: mkennedy.codes Watch this episode on YouTube: youtube.com Episode #536 deep-dive: talkpython.fm/536 Episode transcripts: talkpython.fm Theme Song: Developer Rap 🥁 Served in a Flask 🎸: talkpython.fm/flasksong ---== Don't be a stranger ==--- YouTube: youtube.com/@talkpython Bluesky: @talkpython.fm Mastodon: @[email protected] X.com: @talkpython Michael on Bluesky: @mkennedy.codes Michael on Mastodon: @[email protected] Michael on X.com: @mkennedy

Building on the web is like working with the perfect clay. It’s malleable and can become almost anything. But too often, frameworks try to hide the web’s best parts away from us. Today, we’re looking at PyView, a project that brings the real-time power of Phoenix LiveView directly into the Python world. I'm joined by Larry Ogrodnek to dive into PyView. Episode sponsors Talk Python Courses Python in Production Links from the show Guest Larry Ogrodnek: hachyderm.io pyview.rocks: pyview.rocks Phoenix LiveView: github.com this section: pyview.rocks Core Concepts: pyview.rocks Socket and Context: pyview.rocks Event Handling: pyview.rocks LiveComponents: pyview.rocks Routing: pyview.rocks Templating: pyview.rocks HTML Templates: pyview.rocks T-String Templates: pyview.rocks File Uploads: pyview.rocks Streams: pyview.rocks Sessions & Authentication: pyview.rocks Single-File Apps: pyview.rocks starlette: starlette.dev wsproto: github.com apscheduler: github.com t-dom project: github.com Watch this episode on YouTube: youtube.com Episode #535 deep-dive: talkpython.fm/535 Episode transcripts: talkpython.fm Theme Song: Developer Rap 🥁 Served in a Flask 🎸: talkpython.fm/flasksong ---== Don't be a stranger ==--- YouTube: youtube.com/@talkpython Bluesky: @talkpython.fm Mastodon: @[email protected] X.com: @talkpython Michael on Bluesky: @mkennedy.codes Michael on Mastodon: @[email protected] Michael on X.com: @mkennedy

Your cloud SSD is sitting there, bored, and it would like a job. Today we’re putting it to work with DiskCache, a simple, practical cache built on SQLite that can speed things up without spinning up Redis or extra services. Once you start to see what it can do, a universe of possibilities opens up. We're joined by Vincent Warmerdam to dive into DiskCache. Episode sponsors Talk Python Courses Python in Production Links from the show diskcache docs: grantjenks.com LLM Building Blocks for Python course: training.talkpython.fm JSONDisk: grantjenks.com Git Code Archaeology Charts: koaning.github.io Talk Python Cache Admin UI: blobs.talkpython.fm Litestream SQLite streaming: litestream.io Plash hosting: pla.sh Watch this episode on YouTube: youtube.com Episode #534 deep-dive: talkpython.fm/534 Episode transcripts: talkpython.fm Theme Song: Developer Rap 🥁 Served in a Flask 🎸: talkpython.fm/flasksong ---== Don't be a stranger ==--- YouTube: youtube.com/@talkpython Bluesky: @talkpython.fm Mastodon: @[email protected] X.com: @talkpython Michael on Bluesky: @mkennedy.codes Michael on Mastodon: @[email protected] Michael on X.com: @mkennedy