Smart Chat FAQs
Core provider blocks and the Pro universal block use provider web interfaces. Pro API Chat uses a configured local or cloud model inside Obsidian. Choose the surface before following setup instructions.
What problem does Smart Chat solve?
#Smart Chat lets you use provider conversations from the notes they serve and, with Pro API Chat, ask configured models inside Obsidian using context you review.
Do I need an API key?
#Not for Core provider blocks or Pro universal provider threads; they use provider web interfaces and may require provider sign-in. Pro API Chat requires a working local model or the credentials required by its cloud provider.
What is the difference between Core, the Pro universal block, and Pro API Chat?
#| Surface | What it does | Model requirement |
|---|---|---|
| Core provider block | Renders one provider web interface such as ChatGPT or Claude in a note. | No API model. |
| Pro universal block | Adds one smart-chat block with a provider selector and note-owned provider bookmarks. |
No API model for provider web threads. |
| Pro API Chat | Runs configured local or cloud models in a native Obsidian thread with optional approved context. | One compatible working model. |
What counts as the first successful result?
#A visible response to a concrete prompt. Installing Chat, opening a view, inserting a block, selecting a model, or saving a URL is preparation. For a note-grounded Pro API request, the answer must also match the approved source.
Which providers do codeblocks support?
#Provider blocks support ChatGPT, Claude, Gemini, DeepSeek, Perplexity, Grok, Google AI Studio, Open WebUI, and Kimi. See Supported providers and exact insert commands.
Does Smart Chat work on mobile?
#Saved web thread URLs can remain useful as external bookmarks on mobile. The embedded provider surface, new-chat capture, automatic URL saving, and state actions require the desktop main Markdown pane.
How do I save or bookmark a thread?
#Add a provider block to the owning note and use the provider until it creates a recognized durable conversation URL. Smart Chat can then write that URL into Markdown. Reopen the saved entry and confirm it returns to the same conversation before relying on it.
Why did the provider answer but the block still says new chat?
#Response success and durable saving are separate. The provider returned an answer, but it has not produced a recognized durable conversation URL. Continue until a supported thread URL exists; do not treat a temporary or provider home-page URL as a saved conversation.
What do Active and Done mean?
#Active means the saved provider thread still needs attention. Done means you reviewed it and closed the loop for now. These are user-owned Markdown states.
Keep the thread Active until you can reopen it and review the result. Change it to Done only after that review.
Does Done mean the AI finished correctly?
#No. Done is a tracking state, not a claim about model correctness. Compare important claims with their sources before promoting them into trusted notes.
Is Chat Inbox the same as Chat Manager?
#No. Chat Inbox reads note-attached chat-active:: and chat-done:: provider bookmarks. Chat Manager manages Pro API Chat thread records. Run Open: Chat Manager view from the Command Palette under Smart Chat Pro.
Does the Dataview Chat Inbox work without Pro?
#Yes. Dataview can read note-attached chat-active:: and chat-done:: fields without Chat Pro. Run Insert Smart Chat thread Dataview blocks when Dataview is enabled, or add equivalent queries manually.
Should I delete done thread links?
#Not automatically. Done thread links can be useful history. Delete them only after reviewing the note and deciding the reference no longer helps.
Can I use local or API models?
#Yes. In Chat Pro settings, find Chat models, choose + New, and select an enabled provider. Cloud providers require their API Key. Ollama requires the Ollama host and a running Ollama app. Choose Refresh Models when needed, then select Chat Model. If several models exist, choose the intended row under Default chat model. Confirm that it is Current, then run Test or Test model. Current means selected; credentials, provider access, or a local server can still prevent an answer.
How do I configure an API key?
#Open model setup through Open settings on the Chat Pro Store row or through Smart Environment settings. Under Chat models, choose + New, select the cloud provider, enter the API Key, choose Refresh Models when needed, select Chat Model, choose it under Default chat model when several rows exist, and run Test or Test model. Do not send note content when the test returns an authorization error such as 401.
Does Smart Chat automatically use the active note?
#No. Provider codeblocks send only what you enter or upload through the provider interface. Pro API Chat uses the sources selected through Add context, @, Lookup context, drag and drop, or Context Builder. The active note and whole vault are not attached automatically.
Does attached context carry forward automatically?
#No. Prior non-excluded user and assistant messages can influence a later response, but attached source context is request-specific. Use Add context or Open context builder to select the required sources again.
What leaves my device?
#A provider web block sends what you type or upload through that provider's site. Pro API Chat sends the prompt, included prior messages, and context selected for that response to the chosen model provider. A local model can keep the request on the machine when its local runtime is active. Review the selected sources and provider terms before sending sensitive material.
How do I keep a thread from drifting?
#Review the current response's attached context, keep stable constraints in custom instructions, and exclude irrelevant prior exchanges from future conversation history. Reattach required sources on later requests instead of assuming they carried forward.
What is the difference between a failed generation and a bad response?
#A failed generation ends in an Error state and produces no completed answer. Correct the model, provider, credential, server, or request problem and retry the same bounded prompt. A bad response completes but is generic, unsupported, or contradictory. Keep it for comparison, correct the prompt or source set, and send a new request.
How do I reopen a recent Pro API thread?
#Choose Chat History from the thread toolbar. It searches thread names and displayed recent user-prompt previews. Press Enter to open the selected thread. It does not search every word in every message.
What if provider sign-in is difficult?
#Complete sign-in through Obsidian Web Viewer, return to the provider block, and choose Refresh. Clear Web Viewer browser data only after confirming that you can recreate the credentials and session. Otherwise, use the provider's own account-recovery guidance or open a saved thread externally.