Missed the live demo? You can now watch Matt Wallace, Senior Solutions Advisor at Posit, walk through how Posit’s modern platform streamlines the entire data science lifecycle. See how to go from a manager’s request to final delivery, all on your own schedule, then check out the questions that were asked live.
Explore how Positron, Positron Assistant, and Databot (AI) help data scientists quickly analyze, document and refactor legacy code, then deploy the revised app in a click to Posit Connect. With reproducibility via Package Manager and governance via granular access controls, Posit delivers a rapid, traceable deployment cycle that closes the loop with stakeholders, elegantly and efficiently.
What you’ll see in action:
- Integrated Development Environment: A Positron session in Posit Workbench, designed to support reproducible analyses, scalable models, and collaborative workflows.
- Legacy Refactoring: Using Positron and AI-powered Databot to document and refactor old code in minutes.
- Traceable Deployment: Deploy the app from Positron in Workbench to Posit Connect with granular access controls and one-click publishing.
- Guaranteed Reproducibility: Leveraging Posit Package Manager to ensure your models are auditable and stable for years to come.
Live Q&A during the session:
Q: I am very curious about the decision to place chat in the left panel. This prevents the file explore and chat to be visible at the same time. Was this deliberate? If so, what was the intent behind this decision?
A: Good question, and this can be a bit of friction for folks coming from RStudio.
By default, Positron Assistant lives in the left sidebar, which means it competes with the file explorer for the same panel. The quickest fix is a setting called "Move Chat into Editor Area or New Window" that pops the chat out as an editor tab (or a separate window), leaving the file explorer free in the sidebar. Most people find the editor tab option works well since you can position it wherever makes sense for your screen.
Q: Will Databot continue to be available even with the release of Posit Assistant (Posit AI)?
A: Databot is being deprecated in favor of Posit Assistant, which is the new unified AI experience in Positron. The good news is that this is not a subtraction story. Posit Assistant picks up everything Databot did for exploratory data analysis and extends it into a broader, more capable coding agent. So the EDA workflows you relied on Databot for are still there, just rolled into a single tool that also handles code generation, inline completions, debugging, and more.
Databot still works today if you have it installed, but the path forward is Posit Assistant. If you have not tried it yet, it is worth enabling.
Q: Is there a way to export the full Positron Assistant chat to a PDF record? How is the chat saved for historical purposes?
A: There is a built-in export command for this. There is a button that you can click in Posit Assistant to export your chat contents.
A couple of things worth knowing on the history side: chat history is available within your current session, but it is not persisted server-side between sessions. Posit does not store your prompts or conversations at all. Everything runs directly between your local client and your model provider (Anthropic, Copilot, etc.), so the exported log is the record.
If your team has audit or compliance requirements around AI-assisted work, the practical approach is to build a habit around that export command and archive the logs through your normal document management process.
Q: What about data dictionaries?
A: If you have an existing data dictionary, whether that's a CSV, a markdown file, a Word doc, whatever format you keep it in, you can attach it directly as context in the chat pane. So when you're asking Assistant to write code against your data, it's not guessing at column names or what a field means. It actually knows your schema and your definitions. That's a pretty big deal for teams where naming conventions are non-obvious or where the same concept shows up differently across systems.
The other direction is generating one. If you load a dataset and ask Assistant to document it, it can look at your data in the Variables pane, infer types and distributions, and help you produce a starting point for a data dictionary that your team can then refine. Not a replacement for the institutional knowledge your team brings, but a solid first draft in minutes rather than hours.
Q: With the Positron Assistant or Databot, can you provide custom organizational instructions to implement certain best practices or company standards and certain guardrails?
A: Positron Assistant supports both automatic workspace instructions and custom instructions files to tailor AI behavior to your project’s needs. For full details: https://positron.posit.co/assistant-chat-instructions.html
Q: Are there any tutorials that teach Positron to beginners?
A: Yes! We have a lot of great resources for Positron beginners. First, I encourage you to check out our Positron webpage. We also have a YouTube playlist dedicated to new users. If you are looking for structured, self-paced learning, Posit Academy offers instructor-guided courses in R and Python that are a great complement to getting started in Positron.For more consistent updates, feature launches, tips & tricks, you can follow us on LinkedIn or sign up for our mailing list here.
Q: How do tokens work for Posit Assistant?
A: There are two ways to handle tokens with Posit Assistant, and which one applies depends on your setup.
For enterprise teams on Posit Workbench, the standard path is Bring Your Own Key. You connect Posit Assistant to your organization's existing API credentials for Anthropic, AWS Bedrock, Azure, or other supported providers. Token costs flow directly to your model provider and Posit is not in the billing loop.
For individual users or smaller teams, Posit AI is a $20/month subscription that includes $15 in monthly API credits plus unlimited Next Edit Suggestions. That covers most typical usage, with an auto-recharge option if you need more.
Either way, usage is not a flat per-question rate. It scales with task complexity, model choice, and response length.
Q: Are there any scientific domains that Positron Assistant is particularly strong on, or not so strong?
A: The honest answer is that Posit Assistant's strength is not domain knowledge in the traditional sense. What makes it different from a general-purpose chatbot is IDE integration. It sees your live session, including your loaded data, console history, variables, and plots, and responds based on what is actually in your environment rather than generic assumptions. That benefit applies regardless of scientific domain.
Positron Assistant supports project-level instruction files where you can encode your team's preferred packages, naming conventions, analytical standards, and domain context. A pharma team might document their pharmaverse conventions and clinical trial SOPs. A finance team might specify their preferred modeling frameworks and data standards. Whatever you invest there, the assistant carries forward automatically in every session.
So the practical answer is: it is strong wherever R and Python are strong, and it gets stronger in any domain where you take the time to encode your team's context. That investment pays off quickly and compounds over time as more of your team adopts it.
Q: You mentioned "human in the loop" and approving actions up front — can the admin configure which actions require approval vs. run automatically, and is there an audit trail of what DataBot ran?
A: At the user level, Posit Assistant prompts for approval before executing anything that touches your files or environment. You see that in the demo where it was going to write or modify something, it paused and asked first. Plan mode takes that further and lets you review the full strategy before any code runs at all.
On the admin side, Posit Workbench gives administrators control over who has access to Posit Assistant in the first place, down to the user or group level. There is also a granular permissions system that controls what the assistant can read, edit, or run. So you can tighten that envelope based on your organization's comfort level.
On audit trails, there is a built-in export command that captures the full conversation including the code and context from your session. The honest caveat is that this is session-based rather than a centrally managed audit log, so if your team needs a formal record for compliance purposes, the practical workflow is to make that export part of your process and archive it through your normal document management. That is not automatic today, but it is a real artifact and it does capture everything that ran.
Q: Which LLM providers does Positron Assistant support, and can we bring our own (Bedrock, Azure OpenAI, on-prem)?
A: Positron Assistant supports a wide range of LLM providers, and yes, you can bring your own. Out of the box, Positron Assistant supports direct API access to Anthropic, OpenAI, GitHub Copilot, Amazon Bedrock, Snowflake Cortex, including custom providers that are considered OpenAI compatible.
