Tips & Troubleshooting

Tips

Starting new chat conversations

You may want to start a new conversation when:

  • Beginning a new topic or task
  • Your chat conversation has become lengthy and the context is less relevant
  • You want to reset the model’s context to avoid confusion from previous messages
  • You feel the model is “stuck” or not responding as expected

Starting a new conversation helps ensure that the model focuses on the current context and provides more accurate and relevant responses.

To create a new chat conversation:

  1. Open the Chat pane.
  2. Click the button in the top right of the Chat pane or use the keyboard shortcut in the Chat pane.

Viewing previous chat conversations

To view your chat conversations:

  1. Click the clock icon in the top right of the Chat pane, or run the Chat: Show Chats… command in the Command Palette.
  2. You can then select a previous conversation to switch back to it.

Understanding token usage

Your language model provider bills you for each chat message and response that consumes tokens. Contact your language model provider for detailed token usage and pricing information.

Positron Assistant can display token usage for each chat message and response pair. To view token usage:

  1. Open the Command Palette and run the command Preferences: Open User Settings to open the settings.
  2. Search for the setting positron.assistant.showTokenUsage.enable.
  3. Check the box to enable token usage display.
    • Note that token usage may not be supported for all language model providers.
  4. Return to the Chat pane. Token usage will now be displayed below each message and response pair.

Visibility into tool calls

The inputs and outputs of tool calls can help you understand how the model is using your session context to generate responses.

To enable tool call details:

  1. Open the Command Palette and run the command Preferences: Open User Settings to open the settings.
  2. Search for the setting positron.assistant.toolDetails.enable.
  3. Check the box to enable tool call details in the Chat pane.
  4. Return to the Chat pane. Tool call details will now be displayed when tools are used.

Troubleshooting

I am getting unexpected or incorrect responses

If you receive unexpected or incorrect responses from the model, consider the following steps:

  • Provide more context: Ensure that your messages include sufficient context about your code, data, or the task at hand.
  • Rephrase your query: Try rewording your question or request to be more specific or clear.
  • Start a new conversation: If the current conversation has become lengthy or confusing, starting a new chat conversation may help reset the context.
  • Reach out for support: If you continue to experience issues, consider reaching out for support to provide feedback or report the problem.

I am getting an error message

If you encounter an error message while using Positron Assistant, try the following:

  • Try again: Sometimes, transient issues can cause errors. Retry your request to see if the issue persists.
  • Start a new conversation: If the error is related to the current chat context, starting a new chat conversation may resolve it.
  • Check provider status: Check if your language model provider has any service outages or issues.
  • Reach out for support: If the error persists, consider reaching out for support to provide feedback or report the problem.

Feedback & Issues

We welcome your feedback to help us improve Positron Assistant!

Please provide as much detail as possible when reporting issues, including:

  • Steps to reproduce the problem,
  • Screenshots or screen recordings if possible,
  • Any relevant logs, error messages and diagnostics.

If you encounter any issues or have suggestions, please let us know by:

Logs

View output logs

  1. Open the Command Palette
  2. Run the command Output: Show Output Channels…
  3. Enter “Assistant” in the input box to filter the output channels
  4. Press to open the Assistant output logs

For additional detail in the logs:

  1. Open the Command Palette
  2. Run the command Developer: Set Log Level…
  3. Select Debug for debug-level logging, or Trace to enable the most detailed logging
  4. Reproduce the issue you are experiencing
  5. Return to the Assistant output logs to view the detailed logs

Export chat log

To export the log of a chat conversation:

  1. Focus your cursor in the chat interface of the conversation you want to export, either in the Chat pane or inline chat
  2. Open the Command Palette
  3. Run one of the commands that start with Positron Assistant: Export the current chat
  4. Click the button in the notification that appears to open the exported chat log

Diagnostics

For detailed diagnostics about Positron Assistant, including information about your language model providers, version information, Positron Assistant settings, and more:

  1. Open the Command Palette
  2. Run the command Positron Assistant: Collect Diagnostics
  3. A diagnostic report will be generated and opened in a new editor tab

Glossary

BYOK

Bring Your Own (API) Key. This means you need to obtain and provide your own API key from the language model provider in order to use their models with Positron Assistant.

Context

The information available to the language model when generating responses. In Positron Assistant, context may include the current file content, session state (such as loaded data and plots), and any relevant metadata about your Positron environment. Providing rich context helps the model generate more accurate and relevant responses.

Model

A specific version or configuration of a language model provided by a language model provider, such as Anthropic Claude or GitHub Copilot. Different models may have different capabilities, performance characteristics, and pricing. Examples include “Claude Sonnet 4.5” or “GPT-5”.

Prompt

The input text or instructions given to a language model to generate a response. Prompts can include questions, commands, or context to guide the model’s output. Positron Assistant has predefined prompts for various tasks, and you can also create custom prompts. These prompts are provided to the model along with your messages in chat or inline chat sessions.

Provider

A company or service that offers access to language models via APIs. Also known as a language model provider or a vendor. Examples include Anthropic and GitHub Copilot.

Tokens

Units of text that language models process. Tokens can be as short as one character or as long as one word. Language models have limits on the number of tokens they can process in a single request, which includes both the input prompt and the generated response.

Tool

A specialized function or capability that a language model can use to enhance its responses. Positron Assistant provides tools that give language models access to the Variables pane, Plots pane, and Console, among other session and workspace context.