Pagenova documentation

Ask better questions about the page in front of you.

Pagenova is a Chrome and Brave side-panel assistant. Ground a conversation in selected text, tabs, files, screenshots, and optional web search, then choose a local or cloud model.

Start here Learn about context

Ollamallama3.2:1b

Selected text · 1 source

What does this paragraph mean?

It explains the context you selected.
Pagenova sends only the current selection with this answer.

+Ask a question...

01Use a local modelConfigure Ollama once, then select a model. 02Give it contextHighlight text, select tabs, or attach a file. 03Ask about an imageAttach, paste, or capture a normal webpage.

Install Pagenova

Install Pagenova from the Chrome Web Store, then open it from your browser toolbar.

Pagenova currently requires Chrome 116 or later. Brave is supported as a Chromium browser. Firefox is not currently supported.

1
Add Pagenova.
Open the Pagenova listing in the Chrome Web Store and choose Add to Chrome. Brave users can install it from the same listing.
2
Pin Pagenova.
Open the browser extensions menu and pin Pagenova to your toolbar.
3
Open the side panel.
Open a normal webpage, then select the Pagenova toolbar icon.

Set up Ollama

Ollama keeps model inference on your machine. Pagenova connects to http://localhost:11434 by default.

1. Install and pull a model

Install Ollama, then pull a text model. Pull a vision model only if you plan to ask about images.

ollama pull llama3.2:1b
ollama pull llava

2. Allow Pagenova’s origin

Ollama must allow the extension origin for pulls and embeddings. Copy Pagenova’s exact origin from Settings.

setx OLLAMA_ORIGINS "chrome-extension://<extension-id>"

Windows: restart the Ollama app after setting the variable. macOS/Linux: run OLLAMA_ORIGINS="chrome-extension://<extension-id>" ollama serve.

Use the exact origin shown in Settings. It limits Ollama access to Pagenova rather than every browser extension.

Chat and history

Choose a provider and model in the header, write a question, then send it. Responses stream as they are generated.

Send and stop

Press Enter to send or Shift + Enter for a new line. While a response streams, Send changes to Stop.

Follow-up questions

Pagenova sends recent conversation turns so questions such as “expand the second point” retain the chat’s subject.

History

The clock opens saved conversations. Reopen, delete, or export the current conversation as Markdown.

Give Pagenova the right context

Context is included only when you ask. It is data for the model, not an instruction that overrides your question.

Fastest

Selected text

Highlight text on the active normal webpage. When you send, Pagenova reads that selection only and adds it as Selected text for that answer.

It does not monitor selections as you browse. Select the paragraph first, then type and send your question.

Across pages

Open tabs

Open the + menu, consent once to let Pagenova read open-tab titles and URLs, then choose one or more normal webpages.

Tabs are read on demand. Pagenova does not run a persistent content script on every page you visit.

Your files

Documents

Attach PDF, DOCX, TXT, Markdown, CSV, JSON, or common code files from the + menu. Attachments remain usable in follow-up turns in the same conversation.

Legacy binary .doc files are not supported.

What the context status means

Using full page context: The selected pages and files fit in the model prompt, so they were included directly.
Retrieved excerpts: The content was large. Pagenova chunked it, embedded it, and supplied excerpts most related to your question.
Indexing context: Pagenova is preparing longer content for retrieval. Wait for it to finish before asking a question that depends on it.

Images and screenshots

Use a vision-capable model for image questions. Pagenova warns you before sending an image to a model that does not support images.

Attach an image

Choose an image in the + menu.
Paste an image from the clipboard.
Drag an image onto the panel.
Use From this page or right-click an image and choose Ask AI about this image.

Capture a screenshot

Open the normal webpage you want to capture.
Open Pagenova, then choose Screenshot.
Allow optional screenshot access when the browser asks.

Pagenova cannot capture browser-internal pages, extension pages, or other protected tabs. It never captures a tab automatically.

For Ollama, Pagenova checks model capabilities through Ollama’s model details and falls back to a conservative model-name check only when capability metadata is unavailable.

Web search

Use the globe button when your question needs information beyond your current context.

DuckDuckGo

The default. It is keyless and returns web results that Pagenova adds as context for the answer.

Tavily

Optional. Add a Tavily API key in Settings for LLM-oriented results and an extracted answer.

Works with local models

An Ollama model can stay local while an enabled search sends the question to the selected search provider.

Sources stay visible

Expand the Sources line under an answer to see the tabs, documents, selected text, or web results supplied for that turn.

Providers and models

Open Settings from the gear icon to save credentials. Pagenova only shows a cloud provider in the chat menu after its key has been saved.

Provider	What you need	Notes
Ollama	Local server and model	No API key. Default address: `http://localhost:11434`.
OpenAI	API key	Select a chat model after saving the key.
Anthropic	API key	Claude models are supported directly.
Google Gemini	API key	Gemini models are supported directly.
Other compatible	HTTPS OpenAI-compatible URL and API key	Use for DeepSeek, OpenRouter, Groq, self-hosted vLLM, LM Studio, and similar services. HTTP is allowed only for localhost.

API keys stay only for the current browser session and are sent only to the provider selected for a request.

Privacy and permissions

Pagenova does not add analytics or telemetry. It uses browser permissions to perform the action you request.

Stored locally

Settings, conversation history, and retrieval data stay in browser storage. API keys stay only for the current browser session.

Sent only on request

Chosen context and your question go to the selected model provider when you send a message. Web search sends your query only while the globe toggle is on.

Page access

Tab text and page images are extracted on demand for tabs you choose. Pagenova does not maintain a persistent content script.

Screenshot permission

Optional screenshot access is requested when you first choose Screenshot. It lets Pagenova capture a normal active webpage after a tab switch.

Read the complete privacy policy.

Troubleshooting

Start with the error text in Pagenova. The fixes below cover the common setup failures.

Ollama pull or embeddings return 403

Copy the exact OLLAMA_ORIGINS command from Settings, then restart Ollama. See Set up Ollama.

A cloud provider returns 401

Save the correct API key in Settings. For Other compatible, also verify the HTTPS base URL. The provider remains hidden until its key is saved.

“Pagenova cannot capture this tab”

Switch to a normal http or https webpage. Browser pages and extension pages are protected and cannot be captured. Choose Screenshot again and accept the permission prompt if it appears.

Context is missing or shows zero sources

For tab context, choose the page through the + menu. For selected text, highlight it on the active normal webpage immediately before sending. A browser-internal page cannot provide either kind of context.

Retrieved answers became weak after switching embedding backends

Existing chunks may have a different embedding dimension. Re-upload the documents or clear stored data to re-index them with the current backend.

An image cannot be sent to the selected model

Choose a vision-capable model. Pagenova checks Ollama model capabilities when it can; other providers use their supported model families.

Still need help?

Use the technical reference for implementation details.

It documents the architecture, message protocol, RAG pipeline, testing, release checks, and contribution workflow.

Open technical reference