mirror of
https://github.com/open-webui/docs.git
synced 2026-07-21 01:55:22 -04:00
refac
This commit is contained in:
Timothy Jaeryang Baek
@@ -1,17 +1,17 @@
|
||||
---
|
||||
title: Agent context and extensions
|
||||
title: Choose an agent extension
|
||||
sidebar_position: 5
|
||||
---
|
||||
|
||||
# Agent context and extensions
|
||||
# Choose an agent extension
|
||||
|
||||
The basic agent can work in a selected workspace. These additions answer four different questions: how do I make repeated work consistent, split independent research, use current web information, or give the workspace a voice, image, or memo?
|
||||
Use an extension only when the ordinary chat, workspace files, and terminal do not provide the missing capability. This is a chooser, not a second setup guide.
|
||||
|
||||
| What you need | Start here |
|
||||
| --- | --- |
|
||||
| Repeat a project rule or remember a durable preference | [Skills and memory](./skills-and-memory) |
|
||||
| Investigate independent questions in parallel | [Sub-agents](./subagents) |
|
||||
| Research or inspect a web page through the agent | [Web search and browser work](./web-and-browser) |
|
||||
| Speak a request, save a voice memo, or work with image output | [Voice, images, and memos](./voice-images-and-memos) |
|
||||
| Missing capability | Choose | Do not use it when |
|
||||
| --- | --- | --- |
|
||||
| A repeatable project method or a fact worth retaining | [Skills and memory](./skills-and-memory) | The instruction only matters for this one chat or already lives clearly in a project file |
|
||||
| Independent work that can return a concise result to the parent task | [Sub-agents](./subagents) | The task needs continuous steering, shares mutable work, or must survive a server restart |
|
||||
| Current information from a site, or browser-only inspection | [Web search and browser work](./web-and-browser) | A workspace file, supplied link, or stable local source already answers the question |
|
||||
| Spoken input, a voice memo, or image generation | [Voice, images, and memos](./voice-images-and-memos) | Typed text is simpler or the material should not leave the machine for a configured provider |
|
||||
|
||||
Choose only the capability that serves the task. Each one changes what data the agent can see or where it can send it.
|
||||
Each choice changes what data the agent can use or where it can send it. State the task, source, and desired result first, then add the smallest extension that closes the gap.
|
||||
|
||||
@@ -25,6 +25,6 @@ Start small. Choose a model, ask it to explain one real file, then decide whethe
|
||||
| Make a native agent available on the host | [Set up and detect a coding agent](./setup-and-detection) |
|
||||
| Review changes and control approval | [Supervise agent work safely](./supervise-work) |
|
||||
| Continue work across devices or recover after an interruption | [Plans, queues, resume, and context](./continuity) |
|
||||
| Add reusable guidance, parallel work, web research, or media input | [Agent context and extensions](./capabilities) |
|
||||
| Decide whether reusable guidance, parallel work, web research, or media input is actually needed | [Choose an agent extension](./capabilities) |
|
||||
|
||||
Not a fit: use a normal chat product when you only need conversation, knowledge, or model comparison. Use a disposable environment when an untrusted agent or user must not reach the real host. Running an unsupported agent CLI in a terminal remains useful, but it does not make that CLI a native Computer chat backend.
|
||||
|
||||
@@ -7,7 +7,7 @@ sidebar_position: 8
|
||||
|
||||
Use web tools when the answer is not on the machine: a current release note, a public page, or a site interaction that needs a real browser. This complements the workspace. It does not turn web output into a trusted source or make the browser's login state safe to share.
|
||||
|
||||
## Pick the smallest useful web capability
|
||||
## Choose the web capability you need
|
||||
|
||||
- Use search or URL reading for current facts and public pages.
|
||||
- Use browser work when the agent must navigate, click, type, or capture a page.
|
||||
@@ -21,7 +21,7 @@ Use web tools when the answer is not on the machine: a current release note, a p
|
||||
|
||||
## Check the result
|
||||
|
||||
The chat shows a search, read, or browser tool result with the expected page or answer. If it fails, check the configured endpoint or browser availability before retrying. If a page is wrong or stale, correct the source rather than treating the first result as proof.
|
||||
The chat shows a search, read, or browser tool result with the expected page or answer. If it fails, check the configured endpoint or browser availability before retrying. If a page is wrong or stale, correct the source rather than relying on the first result.
|
||||
|
||||
## What to trust
|
||||
|
||||
|
||||
@@ -17,7 +17,7 @@ You can stop here when the task does not need a shell, a project folder, or a pa
|
||||
|
||||
Open Terminal is an Open WebUI feature. It gives an Open WebUI chat a computing environment where the model can write files, run commands, install packages, and return the result in the conversation.
|
||||
|
||||
Use it for a self-contained job: analyze an uploaded dataset, build a small prototype, run a script, or give an agent a clean environment for a task. It is especially useful when the work does not depend on an existing machine's long-lived state. Run it in Docker when you want isolation, or configure bare-metal access when that is the deliberate choice.
|
||||
Use it for a self-contained job: analyze an uploaded dataset, build a small prototype, run a script, or give an agent a clean environment for a task. It is especially useful when the work does not depend on an existing machine's long-lived state. Run it in Docker for an isolated execution environment. Use bare-metal access only when direct host access is intentional.
|
||||
|
||||
[Learn about Open Terminal](/features/open-terminal)
|
||||
|
||||
@@ -51,14 +51,12 @@ This is common. Keep Open WebUI as the place to ask, compare, and organize. Use
|
||||
|
||||
The value is not forcing every job through one product. It is choosing the environment that matches the work in front of you.
|
||||
|
||||
## A simple way to decide
|
||||
## A quick way to decide
|
||||
|
||||
| If the first true sentence is… | Start with… |
|
||||
| --- | --- |
|
||||
| “I need to think through this, compare models, or work with shared AI knowledge.” | Open WebUI |
|
||||
| “I want an AI to run a self-contained task in a clean computing environment.” | Open WebUI with Open Terminal |
|
||||
| “The files, process, login, project, or agent I need already exist on my machine.” | Open WebUI Computer |
|
||||
| “I want an Open WebUI conversation to direct work in one existing workspace.” | Open WebUI with Computer |
|
||||
- Need to think, compare models, or use shared AI knowledge? Start in Open WebUI.
|
||||
- Need an AI to run a self-contained task in an isolated environment? Add Open Terminal in Docker.
|
||||
- Need the files, process, login, project, or agent that already exists on your machine? Start in Computer.
|
||||
- Need an Open WebUI conversation to direct work in one existing workspace? Connect Open WebUI to Computer.
|
||||
|
||||
## If you already use an agent command-line tool
|
||||
|
||||
@@ -66,4 +64,4 @@ Computer can give an installed agent command a real terminal in the selected wor
|
||||
|
||||
## Keep the trust model in view
|
||||
|
||||
Open Terminal can provide a managed execution environment. Computer deliberately reaches the real host account and its files, shell, and processes. That is what makes it useful for a trusted owner and why it should stay private. Read [the security model](./remote-access/security-model) before exposing it remotely or connecting unattended integrations.
|
||||
Open Terminal can run in an isolated Docker environment or, when you choose it, on a host. Computer deliberately reaches the real host account and its files, shell, and processes. That is what makes it useful for a trusted owner and why it should stay private. Read [the security model](./remote-access/security-model) before exposing it remotely or connecting unattended integrations.
|
||||
|
||||
@@ -14,7 +14,7 @@ Build trust in one layer at a time. Each step has a result you can see before yo
|
||||
1. **Open a local folder.** Use [Local trial](./local-trial), then [open your first workspace](./first-workspace). Confirm a familiar file is the real file on the host.
|
||||
2. **Optionally add an agent.** Ask it to explain one real file, then decide whether to let it change anything. Start with [AI and coding agents](/ecosystem/computer/agents).
|
||||
3. **Optionally reach the host remotely.** Keep the service local by default; when you need another device, add a private route with [Remote access and security](/ecosystem/computer/remote-access).
|
||||
4. **Only then add unattended or external connections.** Schedules, bots, gateways, and external tool servers are high-trust paths. Start one at a time from [Automations and integrations](/ecosystem/computer/integrations), with a scoped workspace and a harmless proof.
|
||||
4. **Only then add unattended or external connections.** Schedules, bots, gateways, and external tool servers are high-trust paths. Start one at a time from [Automations and integrations](/ecosystem/computer/integrations), with a scoped workspace and a harmless read-only check.
|
||||
|
||||
The sequence is intentional, not a required feature checklist. Stop after any step that solves the problem you have today.
|
||||
|
||||
|
||||
@@ -5,7 +5,7 @@ sidebar_position: 2
|
||||
|
||||
# Local trial
|
||||
|
||||
Your first proof should be simple: choose a folder you recognize, open a familiar file, and confirm it is the same folder already on your machine. Nothing is copied to a cloud workspace.
|
||||
Start with a folder you recognize. Open a familiar file and see the same folder that is already on your machine. Nothing is copied to a cloud workspace.
|
||||
|
||||
## Use this when
|
||||
|
||||
|
||||
@@ -45,21 +45,27 @@ For more paths, including recurring checks, on-call triage, local-device preview
|
||||
|
||||
## Why this is different
|
||||
|
||||
| Without Computer | With Computer |
|
||||
| --- | --- |
|
||||
| Recreate context: find the right files, restart a service, locate the terminal, and explain the job again. | Reopen the machine where the useful context already exists. |
|
||||
| A chat AI can help, but it does not naturally have your local project or running application. | A configured agent can work in one selected real workspace beside the files and tools you already use. |
|
||||
| SSH and remote desktop can reach the host, but are clumsy for reviewing a diff, checking an app, or steering work from a phone. | A browser workspace brings together files, git, terminals, previews, chats, and agent activity. |
|
||||
### Keep the context you already built
|
||||
|
||||
Open the machine where the right files, project, process, and terminal already exist. There is no second project to create and no context to retell.
|
||||
|
||||
### See the work before you decide what to do
|
||||
|
||||
The workspace brings files, git, terminals, previews, chats, and agent activity together in a browser that works from another computer or phone.
|
||||
|
||||
### Add AI without moving the work away
|
||||
|
||||
When you configure a model or supported agent, it works in the selected real workspace beside the tools you already use. You can also use Computer with no AI at all.
|
||||
|
||||
## What Open WebUI Computer is
|
||||
|
||||
Computer is a private control surface for work on one machine you operate. It keeps a chosen workspace, its files, terminal and git state, supported agents, and their review history together. It is a self-hosted personal agent platform, not an adapter that turns every installed agent CLI into a first-class backend, and not a shared isolated computer service.
|
||||
Computer is a private control surface for work on one machine you operate. It keeps a chosen workspace, its files, terminal and git state, agent work, and review history together.
|
||||
|
||||
It is useful with no AI at all. Add a model, coding-agent subscription, browser tool, automation, bot, or external tool only when that improves a job you actually have. Each extra connection brings its own data and trust boundary.
|
||||
|
||||
## Your first proof should be small
|
||||
## Start with your own folder
|
||||
|
||||
Run it locally, open one familiar folder, and confirm that you can see the real files. No remote access, AI provider, bot, or automation is required for this first check.
|
||||
Run it locally and open one familiar folder. You will see the real files that are already on the machine. No remote access, AI provider, bot, or automation is required.
|
||||
|
||||
You need permission to run one install command on the machine that holds the work. If it is not your machine, ask its owner to do the local setup with you. You do not need to know git or use a terminal after the initial install.
|
||||
|
||||
@@ -67,13 +73,11 @@ You need permission to run one install command on the machine that holds the wor
|
||||
|
||||
## Choose the path that fits how you work
|
||||
|
||||
| You are here because… | Start here |
|
||||
| --- | --- |
|
||||
| You make or review visual work, school work, research, or independent projects | [Explore real jobs](./use-cases/) |
|
||||
| You use a coding agent or want an AI to work in a real project | [AI and coding agents](./agents/) |
|
||||
| You already use Open WebUI or Open Terminal | [Choose the right execution environment](./choose) |
|
||||
| You already use an agent command-line tool | [Can Computer manage the agent CLI I already use?](./choose#if-you-already-use-an-agent-command-line-tool) |
|
||||
| You need remote access, backups, or a full security model before adopting it | [Remote access and security](./remote-access/) |
|
||||
- Making or reviewing visual work, school work, research, or an independent project? [Explore real jobs](./use-cases/).
|
||||
- Want an AI to work in a real project? [Set up an agent](./agents/).
|
||||
- Already use Open WebUI or Open Terminal? [Choose the right execution environment](./choose).
|
||||
- Already use an agent command-line tool? [See how it works with Computer](./choose#if-you-already-use-an-agent-command-line-tool).
|
||||
- Need remote access, backups, or a full security model first? [Read remote access and security](./remote-access/).
|
||||
|
||||
## Choose the computer that matches the work
|
||||
|
||||
@@ -81,10 +85,6 @@ Open WebUI is the AI interface and workflow platform. Open Terminal gives an Ope
|
||||
|
||||
[Choose the right execution environment](./choose)
|
||||
|
||||
## Open WebUI Computer vs Open Terminal
|
||||
|
||||
Open Terminal gives an Open WebUI chat model a computing environment to drive. Open WebUI Computer gives you the existing computer you already operate, including its files, terminal, editor, git state, and running sessions. Choose Open Terminal when a chat needs somewhere to execute. Choose Computer when the machine with the useful context already exists.
|
||||
|
||||
## Keep it private
|
||||
|
||||
Computer runs on a machine you control. Its core workspace needs no managed cloud account, cloud control plane, or AI provider. An authenticated user can access the host filesystem and shell, so treat remote access like private SSH, not a public web app. It is a strong fit for one trusted owner, not for giving untrusted people separate isolated machines.
|
||||
|
||||
@@ -53,7 +53,7 @@ For a test check, add a [notification target](./notifications-and-webhooks) only
|
||||
|
||||
## Read the first run like a rehearsal
|
||||
|
||||
Open the new run-history entry and its linked chat. This is the proof that matters: you should be able to see the selected workspace, the task activity, and the terminal or tool output that supports the result.
|
||||
Open the new run-history entry and its linked chat. You should be able to see the selected workspace, the task activity, and the terminal or tool output that supports the result.
|
||||
|
||||
If the report says the test passed, but the linked chat does not show the test command or its output, improve the prompt or model setup before you schedule it. A successful schedule is not enough. You need a result you can judge later in seconds.
|
||||
|
||||
|
||||
@@ -5,7 +5,7 @@ sidebar_position: 1
|
||||
|
||||
# Automations and integrations
|
||||
|
||||
Integrations make a real Computer workspace reachable from another product or let it act without the browser open. That is valuable only when the workspace, sender, credentials, and side effects are all deliberate. Start with a read-only proof and add one connection at a time.
|
||||
Integrations make a real Computer workspace reachable from another product or let it act without the browser open. That is valuable only when the workspace, sender, credentials, and side effects are all deliberate. Start with a read-only check and add one connection at a time.
|
||||
|
||||
| What you want | Read this |
|
||||
| --- | --- |
|
||||
|
||||
@@ -7,7 +7,7 @@ sidebar_position: 7
|
||||
|
||||
A messaging bot can be convenient when your phone is the only screen you have and the answer is in a trusted Computer workspace. It is not a general customer-support bot: an inbound message can start host-capable agent work.
|
||||
|
||||
## The first useful proof
|
||||
## Start with one harmless question
|
||||
|
||||
If your client repositories live on a workstation and you already check Slack while commuting, the first useful bot question is not "fix whatever is wrong." It is: "Which branch is checked out in the selected client workspace?" You get a real answer from the machine without opening a laptop, while keeping the first request read-only.
|
||||
|
||||
|
||||
@@ -9,7 +9,7 @@ Open WebUI is the place to organize AI conversations, models, knowledge, and sha
|
||||
|
||||
It is valuable when you want Open WebUI as the front door without copying a repository or pretending a remote agent has your local machine context. It is not a general two-way sync product: it does not import Computer’s standalone chats into Open WebUI and does not make Open WebUI knowledge bases automatically available to Computer.
|
||||
|
||||
## The first useful proof
|
||||
## Start with a read-only request
|
||||
|
||||
You may already use Open WebUI to organize research conversations while your docs checkout, local scripts, and browser login live on a Mac. You do not need another repository or a generic answer. You need one Open WebUI chat to inspect the real checkout.
|
||||
|
||||
@@ -19,7 +19,7 @@ Connect one deliberately bounded workspace, then ask a harmless question such as
|
||||
Read the README in this workspace. Do not change files. Tell me the command it gives for the local documentation preview.
|
||||
```
|
||||
|
||||
If that answer is grounded in the file on the Computer host, the combination is useful. Save edits, deploys, and broad tasks for after this proof works.
|
||||
If that answer is grounded in the file on the Computer host, you are ready to use the combination for a narrow task. Save edits, deploys, and broad tasks for later.
|
||||
|
||||
:::danger Gateway requests are unattended
|
||||
Gateway requests run with **full tool approval**. Open WebUI cannot pause a file edit, command, or external tool call for a per-tool confirmation round trip. Connect only trusted Open WebUI clients to deliberately bounded workspaces and models. For interactive approval, open the same workspace directly in Open WebUI Computer.
|
||||
|
||||
@@ -11,7 +11,7 @@ You need a quick health signal, a durable diagnostic record, or a controlled loc
|
||||
|
||||
## Before you start
|
||||
|
||||
The unauthenticated health endpoint is `/api/health`. Environment variables are read when the process starts. `config.toml` lives in the data directory and mirrors server and app configuration; avoid editing it while the app is running.
|
||||
The unauthenticated health endpoint is `/api/health`. Environment variables are read when the process starts. `config.toml` lives in the data directory and mirrors server and app configuration; avoid editing it while the app is running. CORS is a browser-origin policy, not an authentication or network boundary.
|
||||
|
||||
## Do it
|
||||
|
||||
@@ -23,6 +23,14 @@ curl http://127.0.0.1:8000/api/health
|
||||
|
||||
Set `CPTR_LOG_LEVEL=DEBUG` for more server detail or `CPTR_LOG_FORMAT=json` for structured stdout, then restart. Set `CPTR_AUDIT_LOG_LEVEL=METADATA` for a low-content mutation trail. `REQUEST` adds redacted request bodies and `REQUEST_RESPONSE` adds redacted response bodies; the default audit file is `~/.cptr/logs/audit.jsonl`, and `CPTR_AUDIT_LOG_PATH` changes it. By default, chat API paths are excluded, and terminal WebSocket activity is not an HTTP audit record, so use the linked chat and terminal output when you need evidence of a task. Use `CPTR_DATA_DIR=/path/to/data` only before start, and make the selected directory persistent and writable.
|
||||
|
||||
For a browser frontend deliberately hosted at another origin, set an explicit comma-separated allowlist before starting Computer. The default `CPTR_CORS_ALLOWED_ORIGINS` value is `*`, which allows all origins. Do not leave that default for deliberate cross-origin use:
|
||||
|
||||
```bash
|
||||
CPTR_CORS_ALLOWED_ORIGINS=https://computer.example.test,https://admin.example.test cptr run
|
||||
```
|
||||
|
||||
Do not add origins merely because a device reaches Computer through Tailscale Serve. The browser uses the private Computer URL there, so a separate frontend origin is not required.
|
||||
|
||||
## Verify it worked
|
||||
|
||||
Health returns JSON with `status`, `uptime_seconds`, and `pid`. Restart with the chosen log setting and observe the requested output format in the process log. After a harmless mutation, the selected audit file contains a redacted JSON entry. When changing the data directory, confirm it contains `app.db` and `config.toml` after a clean start.
|
||||
@@ -33,4 +41,4 @@ If health fails locally, inspect the server's stdout and verify the chosen port.
|
||||
|
||||
## Trust boundary
|
||||
|
||||
`/api/health` intentionally does not require login, so do not place secrets or sensitive diagnostics in it. Debug and audit logs may contain operational data and need protected storage. Audit logging is a redacted HTTP mutation trail, not a complete forensic record of every agent, terminal, or external action.
|
||||
`/api/health` intentionally does not require login, so do not place secrets or sensitive diagnostics in it. Debug and audit logs may contain operational data and need protected storage. Audit logging is a redacted HTTP mutation trail, not a complete forensic record of every agent, terminal, or external action. An explicit CORS allowlist only limits browser origins; it does not replace private networking, sign-in, or host permissions.
|
||||
|
||||
@@ -1,19 +1,20 @@
|
||||
---
|
||||
title: Private remote access with Tailscale or a tunnel
|
||||
title: Private remote access with Tailscale
|
||||
sidebar_position: 4
|
||||
---
|
||||
|
||||
# Reach a machine away from home
|
||||
# Private remote access with Tailscale
|
||||
|
||||
## Use this when
|
||||
|
||||
Use this when you need the real branch and terminal from a phone outside the home network, without opening a public router port.
|
||||
Use this when you need the real branch and terminal from a phone outside the home network, without opening a public router port. Tailscale Serve is the supported private-access route in this guide.
|
||||
|
||||
## Before you start
|
||||
|
||||
- Complete the [local trial](/ecosystem/computer/getting-started/local-trial) and [first workspace](/ecosystem/computer/getting-started/first-workspace) on the host first.
|
||||
- Keep Open WebUI Computer on its default localhost binding (`127.0.0.1`).
|
||||
- Enrol only devices you control in a private Tailscale network. Tailscale Serve may prompt an administrator to enable HTTPS certificates for that network.
|
||||
- Do not use Tailscale Funnel or another public tunnel. Public-tunnel deployment is unsupported for Open WebUI Computer.
|
||||
|
||||
## Do it
|
||||
|
||||
@@ -40,8 +41,8 @@ Turn off Wi-Fi on the remote phone so it uses a different network. Open the URL
|
||||
|
||||
## If it did not
|
||||
|
||||
If Serve cannot reach Computer, first verify `http://127.0.0.1:8000/api/health` on the host. Then rerun `tailscale serve status` and confirm both devices are enrolled and allowed by the tailnet policy. Do not solve a private-access failure by opening a router port. To remove the proxy, run `tailscale serve reset`.
|
||||
If Serve cannot reach Computer, first verify `http://127.0.0.1:8000/api/health` on the host. Then rerun `tailscale serve status` and confirm both devices are enrolled and allowed by the tailnet policy. Do not solve a private-access failure by opening a router port or a public tunnel. To remove the proxy, run `tailscale serve reset`.
|
||||
|
||||
## Trust boundary
|
||||
|
||||
Tailscale Serve limits reachability to enrolled devices and policy, but it does not reduce what a signed-in Computer user can do. A public tunnel URL has a different threat model: do not publish one without an independent access-control design. See [the Tailscale Serve documentation](https://tailscale.com/docs/reference/tailscale-cli/serve) for current command behavior.
|
||||
Tailscale Serve limits reachability to enrolled devices and tailnet policy, but it does not reduce what a signed-in Computer user can do. Public tunnel deployment, including Tailscale Funnel, is unsupported. Do not expose this service to the public internet. See [the Tailscale Serve documentation](https://tailscale.com/docs/reference/tailscale-cli/serve) for current command behavior.
|
||||
|
||||
@@ -15,7 +15,7 @@ Decide whether the failure is local, LAN, or private-remote. `localhost` on a ph
|
||||
|
||||
## Do it
|
||||
|
||||
On the host, run `curl http://127.0.0.1:8000/api/health`. For private remote use, keep the existing localhost-bound server running and follow [Tailscale or tunnels](/ecosystem/computer/remote-access/tailscale-and-tunnels).
|
||||
On the host, run `curl http://127.0.0.1:8000/api/health`. For private remote use, keep the existing localhost-bound server running and follow [private remote access with Tailscale](/ecosystem/computer/remote-access/tailscale-and-tunnels).
|
||||
|
||||
For intentional LAN-only testing, stop the existing server first, restart it with `cptr run --host 0.0.0.0`, and use the host's private IP from the second device. Verify the host firewall limits the port to the intended private subnet, then restore the default localhost binding when the test is complete.
|
||||
|
||||
|
||||
@@ -2,7 +2,7 @@
|
||||
"label": "Use cases",
|
||||
"position": 4,
|
||||
"collapsible": true,
|
||||
"collapsed": false,
|
||||
"collapsed": true,
|
||||
"link": {
|
||||
"type": "doc",
|
||||
"id": "ecosystem/computer/use-cases/index"
|
||||
|
||||
@@ -1,6 +1,6 @@
|
||||
---
|
||||
title: "Supervise a coding agent away from your desk"
|
||||
sidebar_position: 4
|
||||
sidebar_position: 8
|
||||
---
|
||||
|
||||
# Supervise a coding agent away from your desk
|
||||
|
||||
@@ -1,11 +1,13 @@
|
||||
---
|
||||
title: "Keep a creative, school, or maker project together"
|
||||
sidebar_position: 8
|
||||
sidebar_position: 3
|
||||
---
|
||||
|
||||
# Keep a creative, school, or maker project together
|
||||
|
||||
Use this when your project is spread between a computer, a phone, reference images, a draft, and ideas you captured while away from your desk. Open WebUI Computer gives those materials one real folder you can reopen from another device. It is not a replacement for a drawing, video, or presentation app. It is the place that keeps the files, notes, and next steps from falling apart between them.
|
||||
Put a project note and its reference files in one folder, then open that same folder from the computer or device you have with you. That is the first useful result: the next step, draft, and materials stay together instead of falling into separate apps and message threads.
|
||||
|
||||
Use this when your project is spread between a computer, a phone, reference images, a draft, and ideas captured away from your desk. Open WebUI Computer gives those materials one real folder you can reopen from another device. It is not a replacement for a drawing, video, or presentation app.
|
||||
|
||||
## Use this when
|
||||
|
||||
@@ -15,15 +17,15 @@ You are making a school project, a portfolio piece, a club site, a zine, a small
|
||||
|
||||
- Complete the [local trial](/ecosystem/computer/getting-started/local-trial) and choose a dedicated project folder as the workspace.
|
||||
- Use copies for photos, recordings, or references that must remain untouched in their original location.
|
||||
- If you plan to use the installed web app's Share action, confirm that your device and browser support it. You can always add files and notes directly in the workspace instead.
|
||||
- Optional: if you plan to use the installed web app's Share action, confirm that your device and browser support it. You can always add files and notes directly in the workspace instead.
|
||||
|
||||
## Do it
|
||||
|
||||
1. In the file browser, create a few simple folders for the project, such as `references`, `drafts`, and `notes`.
|
||||
2. Add reference images, PDFs, text, or other project files by opening existing local files or uploading chosen copies into the relevant folder.
|
||||
3. Create a short Markdown note with the project goal and the next action. Reopen it whenever you return so the project explains itself.
|
||||
4. On a supported installed web app, share a link, text, or file from the device into the workspace and confirm the destination. If sharing is unavailable, create the note or upload the item from the workspace instead.
|
||||
5. If Voice Memos are enabled for the instance, record a quick idea and save it with a clear filename. When transcription is configured, review the saved transcript before treating it as the final wording.
|
||||
4. Optional: on a supported installed web app, share a link, text, or file from the device into the workspace and confirm the destination. If sharing is unavailable, create the note or upload the item from the workspace instead.
|
||||
5. Optional: if Voice Memos are enabled for the instance, record a quick idea and save it with a clear filename. When transcription is configured, review the saved transcript before treating it as the final wording.
|
||||
|
||||
## Verify it worked
|
||||
|
||||
|
||||
@@ -11,7 +11,7 @@ You do not need to be a developer or connect AI to try the useful part. Start lo
|
||||
|
||||
## I have a folder I care about, but I do not use coding tools
|
||||
|
||||
[Open it locally first](/ecosystem/computer/getting-started/local-trial). The first proof is simply that the folder stays in its real location and opens in Computer, rather than being copied into a new cloud workspace.
|
||||
[Open it locally first](/ecosystem/computer/getting-started/local-trial). The folder stays in its real location and opens in Computer, rather than being copied into a new cloud workspace.
|
||||
|
||||
## I need to keep feedback with a prototype or project
|
||||
|
||||
|
||||
@@ -15,7 +15,8 @@ You want Open WebUI as the AI front door and Open WebUI Computer as the place th
|
||||
|
||||
- Complete the [local trial](/ecosystem/computer/getting-started/local-trial) and [first workspace](/ecosystem/computer/getting-started/first-workspace), then run both services where Open WebUI can reach Computer’s gateway over a private, trusted network.
|
||||
- Create the target workspace and configure its default agent or model in Computer.
|
||||
- Create a gateway key and follow the [combined setup guide](/ecosystem/computer/integrations/open-webui-gateway) for the exact connection and continuity headers.
|
||||
- Gateway requests use **full tool approval**. Computer does not pause for an interactive confirmation before each tool action, so connect only a trusted Open WebUI user and a workspace you are willing to let that request act in. Read the [detailed gateway guide](/ecosystem/computer/integrations/open-webui-gateway) before creating a key.
|
||||
- Create a gateway key and follow the detailed guide for the exact connection and continuity headers.
|
||||
|
||||
## Do it
|
||||
|
||||
|
||||
@@ -1,6 +1,6 @@
|
||||
---
|
||||
title: "Test a local app on a real device"
|
||||
sidebar_position: 5
|
||||
sidebar_position: 6
|
||||
---
|
||||
|
||||
# Test a local app on a real device
|
||||
@@ -11,7 +11,7 @@ This is valuable because a phone or tablet can reveal problems a desktop browser
|
||||
|
||||
## Use this when
|
||||
|
||||
Use this when a process in a workspace has opened a local port and you want to view that process through Open WebUI Computer from another signed-in device. This is especially useful for checking a real browser, touch layout, or a quick internal demo.
|
||||
Use this when a process in a workspace has opened a local port and you want to view that process through Open WebUI Computer from another signed-in device. This is especially useful for checking a real browser, touch layout, or a quick private preview.
|
||||
|
||||
## Before you start
|
||||
|
||||
|
||||
@@ -1,6 +1,6 @@
|
||||
---
|
||||
title: "Fix an issue from another device"
|
||||
sidebar_position: 2
|
||||
sidebar_position: 7
|
||||
---
|
||||
|
||||
# Fix an issue from another device
|
||||
|
||||
@@ -1,11 +1,13 @@
|
||||
---
|
||||
title: "Work with local documents and data"
|
||||
sidebar_position: 7
|
||||
sidebar_position: 4
|
||||
---
|
||||
|
||||
# Work with local documents and data
|
||||
|
||||
Use this when the report, spreadsheet, PDF, or source material you need is already on your computer and you want to understand it without scattering copies across chat uploads and download folders. Open WebUI Computer keeps the source files, a working note, and optional AI help in the same real project folder. A generic chat can summarize an attachment, but it does not naturally keep your original documents, your local data files, and your conclusion together.
|
||||
Open the report and data file you already have, then save one findings note beside them. That is the first useful result: the source material and your conclusion stay together in the real folder instead of becoming scattered uploads and downloads.
|
||||
|
||||
Use this when the report, spreadsheet, PDF, or source material you need is already on your computer. Open WebUI Computer keeps the source files and working note together. A generic chat can summarize an attachment, but it does not naturally keep your original documents, local data files, and conclusion together.
|
||||
|
||||
## Use this when
|
||||
|
||||
@@ -14,7 +16,7 @@ You are reading reports, comparing a CSV or spreadsheet, organizing sources for
|
||||
## Before you start
|
||||
|
||||
- Complete the [local trial](/ecosystem/computer/getting-started/local-trial) and create a workspace for the folder that holds the material.
|
||||
- Decide whether the documents are safe to place in that workspace and safe to send to an AI provider. Keep sensitive files local if you do not have a provider and data-handling arrangement you trust.
|
||||
- Decide whether the documents are safe to place in that workspace. Optional AI help sends relevant prompt and file context to its configured provider, so keep sensitive files local if you do not have a provider and data-handling arrangement you trust.
|
||||
- Use copies when the source material must remain unchanged. Open WebUI Computer shows the real files in the selected folder.
|
||||
|
||||
## Do it
|
||||
@@ -22,7 +24,7 @@ You are reading reports, comparing a CSV or spreadsheet, organizing sources for
|
||||
1. Open the workspace and use the file browser to open existing files or upload deliberate copies into a project subfolder.
|
||||
2. Open a report and a related data file in tabs or split panes. The editor can preview PDFs, CSV and TSV data, and common Office files such as DOCX and XLSX.
|
||||
3. Create a short Markdown note in the workspace, such as `findings.md`, and record the question, source filenames, and what you observed.
|
||||
4. If you have configured an AI model you trust, ask a bounded question in the workspace chat, such as “Read these two files and list differences I should check. Do not change files.” Review its answer against the open source files.
|
||||
4. Optional: if you have configured an AI model you trust, ask a bounded question in the workspace chat, such as “Read these two files and list differences I should check. Do not change files.” Review its answer against the open source files.
|
||||
5. Save the note beside the sources so the next session starts with the material and conclusion together.
|
||||
|
||||
## Verify it worked
|
||||
|
||||
@@ -1,6 +1,6 @@
|
||||
---
|
||||
title: "Resume a workspace from another device"
|
||||
sidebar_position: 3
|
||||
sidebar_position: 2
|
||||
---
|
||||
|
||||
# Resume a workspace from another device
|
||||
|
||||
@@ -1,6 +1,6 @@
|
||||
---
|
||||
title: "Review a prototype and capture feedback"
|
||||
sidebar_position: 6
|
||||
sidebar_position: 5
|
||||
---
|
||||
|
||||
# Review a prototype and capture feedback
|
||||
|
||||
@@ -5,35 +5,19 @@ sidebar_position: 6
|
||||
|
||||
# Browser sessions and local-port previews
|
||||
|
||||
Open WebUI Computer can show a local service that is already running on the host through a workspace preview tab. This lets you check an app from a phone or tablet while keeping the original terminal, logs, and source files in reach.
|
||||
Browser tabs and local-port previews look similar, but they answer different questions. Use a browser tab to visit a website. Use a local-port preview to inspect an app that is already listening from a terminal in this workspace.
|
||||
|
||||
## Use this when
|
||||
| You need to | Use |
|
||||
| --- | --- |
|
||||
| Open an external or internal URL without leaving Computer | [Browser sessions](./browser-sessions) |
|
||||
| Check a development server beside its source and terminal logs | [Local-port previews](./local-port-previews) |
|
||||
|
||||
Preview a development server, inspect a local web UI, or pair a browser view with terminal output. Use a normal deployment path for a public demo or an application that must be reached by people outside the trusted Computer instance.
|
||||
Neither is a deployment mechanism. Use a normal deployment path for a public demo or an application that must be reached by people outside the trusted Computer instance.
|
||||
|
||||
## Before you start
|
||||
## Choose deliberately
|
||||
|
||||
- Start the service in the intended workspace and confirm the listening port from its output.
|
||||
- Sign in to the same private Open WebUI Computer instance from the browser or device you will use for testing.
|
||||
- Know whether the local app assumes `localhost`, a particular origin, or its own authentication. The preview does not rewrite the app's security model.
|
||||
Browser sessions are temporary runtime sessions. Local-port previews depend on a process that is already running on the host. If Computer, its browser service, or the local process restarts, recover from the focused guide instead of assuming the tab preserved the working state.
|
||||
|
||||
## Do it
|
||||
## Safety boundary
|
||||
|
||||
1. Open the file browser in the workspace. When a process opens a port, its `:port` appears in the **Ports** row.
|
||||
2. Select the port to open a preview tab. Switch back to the terminal or arrange the preview beside source files as needed.
|
||||
3. Navigate through the app and inspect terminal logs while testing. Save source edits and let the development server reload, or refresh the preview when its toolchain requires it.
|
||||
4. Close the preview when finished; stop the server separately from the terminal if it should no longer run.
|
||||
|
||||
## Verify it worked
|
||||
|
||||
The preview displays content from the expected local process, and requests or logs appear in that process's terminal. A change made in the selected workspace is reflected after the normal reload path. Reopening the preview shows the same host service while it remains running.
|
||||
|
||||
## If it did not
|
||||
|
||||
- **There is no Ports row:** inspect the terminal for a failed start or a different port. Ensure the process is listening rather than merely compiling.
|
||||
- **The page fails to load:** try the displayed port again and check server logs. Some apps need their configured host or origin adjusted for proxying; fix the app deliberately rather than exposing the service publicly.
|
||||
- **A page works on desktop but not the test device:** check the device browser, app authentication, and responsive behavior. The preview proves the real-device request path, not every platform-specific behavior.
|
||||
|
||||
## Trust boundary
|
||||
|
||||
The preview is a private convenience for services on the host. It can make a development server visible to anyone able to use this instance. Do not use it as a public URL, and do not assume it protects a service that lacks its own authentication.
|
||||
Both surfaces can expose real, valuable state. A browser session may use a managed browser or, when configured, a personal Chrome source with signed-in websites. Treat it as access to the host browser: only use trusted Computer instances and operators, and do not share a tab or instance that can reach sensitive accounts. A port preview is a private convenience, not a public URL or an authentication layer for the local service.
|
||||
|
||||
@@ -0,0 +1,33 @@
|
||||
---
|
||||
title: Browser sessions
|
||||
sidebar_position: 7
|
||||
---
|
||||
|
||||
# Browser sessions
|
||||
|
||||
Use a Browser tab when you need to visit an external or internal site while keeping the related chat, files, and terminal visible in Computer. It is useful for checking a staging site, following a research link, or reproducing a browser-only problem alongside the project.
|
||||
|
||||
## Open and use a browser tab
|
||||
|
||||
1. In the workspace tab bar, create a **Browser** tab.
|
||||
2. Enter the full URL in its address field and press Enter.
|
||||
3. Use the browser tab's Back, Forward, and Reload controls while you inspect the site. Its label updates as the page title becomes available.
|
||||
4. Keep the browser beside the relevant chat or terminal when you need to compare a result with the work that produced it.
|
||||
|
||||
This is a browsing surface, not a durable browser profile. Its configured browser source may be a managed browser or a personal Chrome source, depending on how the Computer instance is set up.
|
||||
|
||||
## What survives and what does not
|
||||
|
||||
The workspace can remember that a Browser tab belongs in the layout, but the live browser session and its proxy cookie state exist only while the Computer server is running. A tab that has expired or is reopened after a Computer restart may need a fresh session, URL, and sign-in.
|
||||
|
||||
Agent-driven browser work has the same practical limit: it is a live session and can expire when idle or close when the server stops.
|
||||
|
||||
## Recover a browser tab
|
||||
|
||||
- **The tab says it expired or no longer loads:** create a new Browser tab and enter the URL again.
|
||||
- **A signed-in site is logged out:** sign in again only if this is a trusted instance and browser source. Do not assume cookies or personal-browser state were retained.
|
||||
- **The expected personal browser is unavailable:** use the configured managed browser or ask the instance operator to check the browser source. Do not work around it by exposing the site or sharing credentials in chat.
|
||||
|
||||
## Signed-in browser safety
|
||||
|
||||
A signed-in website can perform real actions, including actions the website permits for that account. Treat browser access as host access. Keep Computer private, limit access to trusted people, and do not use shared or public deployments for sensitive accounts. Close the Browser tab when the work is complete.
|
||||
@@ -12,11 +12,11 @@ This is the part that makes Open WebUI Computer more than a chat window. A works
|
||||
| Need | Guide |
|
||||
| --- | --- |
|
||||
| Confirm you opened the real project, not a copy | [Workspace paths](./workspace-paths) |
|
||||
| Review a prototype, document, image, or shared file and keep feedback with it | [Review files and keep feedback with the project](./review-files-and-feedback) |
|
||||
| Save a note, shared item, or import in the selected workspace | [Save notes, shares, and imports](./review-files-and-feedback) |
|
||||
| Read, edit, or preview files | [Files and editor](./files-editor) |
|
||||
| Resume a command or start a local service | [Terminals](./terminals) |
|
||||
| Review a branch or diff | [Git](./git) |
|
||||
| Open a running local app from another device | [Browser and port preview](./browser-port-preview) |
|
||||
| Open a website or inspect a running local app from another device | [Browser sessions and local-port previews](./browser-port-preview) |
|
||||
| Find an earlier chat, message, or file | [Search and chats](./search-chats) |
|
||||
| Make the workspace easy to reach from a phone | [PWA, sharing, and shortcuts](./pwa-share-shortcuts) |
|
||||
|
||||
|
||||
@@ -0,0 +1,32 @@
|
||||
---
|
||||
title: Local-port previews
|
||||
sidebar_position: 8
|
||||
---
|
||||
|
||||
# Local-port previews
|
||||
|
||||
Use a local-port preview to inspect a development server that is already running from a terminal in the selected workspace. This keeps the app, source files, and the terminal producing its logs in one place, including when you check the Computer instance from another device.
|
||||
|
||||
## Open a running local app
|
||||
|
||||
1. Start the app in a terminal in the intended workspace and confirm from its output that it is listening on a port.
|
||||
2. Open the workspace file browser. When Computer detects a listening workspace terminal port, it appears in the **Ports** row.
|
||||
3. Select the displayed `:port` to create a preview tab.
|
||||
4. Make a source change, follow the app's usual reload path, and check the preview and terminal logs together.
|
||||
|
||||
The preview connects to the real process already listening on the host. It does not start, deploy, or secure that app for you.
|
||||
|
||||
## What survives and what does not
|
||||
|
||||
The preview can be reopened while the same Computer server and local process remain alive. If the terminal process exits, the host restarts, or Computer restarts, the app and preview connection do not resume on their own. Start the service again, then reopen the port.
|
||||
|
||||
## Recover a preview
|
||||
|
||||
- **No port appears:** inspect the terminal for a failed start, a different port, or a process that is compiling but not listening.
|
||||
- **The preview cannot load:** check the server logs, refresh the tab, and verify the app's host, origin, and authentication settings. Fix those settings in the app rather than exposing the port publicly.
|
||||
- **It worked before a restart:** reopen the workspace terminal, restart the development server, wait for its port to appear, and select it again.
|
||||
- **It differs on another device:** check the device browser, app authentication, and responsive behavior. The preview reaches the live local app, but it does not remove platform-specific differences.
|
||||
|
||||
## Trust boundary
|
||||
|
||||
A port preview is a private convenience for people who can use this Computer instance. Do not treat it as a public URL, and do not assume it protects a local service that lacks its own authentication.
|
||||
@@ -1,6 +1,6 @@
|
||||
---
|
||||
title: "Install the app, share into a workspace, and use shortcuts"
|
||||
sidebar_position: 8
|
||||
sidebar_position: 10
|
||||
---
|
||||
|
||||
# Install the app, share into a workspace, and use shortcuts
|
||||
|
||||
@@ -1,30 +1,29 @@
|
||||
---
|
||||
title: Review files and keep feedback with the project
|
||||
title: Save notes, shares, and imports in a workspace
|
||||
sidebar_position: 2
|
||||
---
|
||||
|
||||
# Review files and keep feedback with the project
|
||||
# Save notes, shares, and imports in a workspace
|
||||
|
||||
When a prototype, document, image, or recording already belongs to a project on your machine, the useful feedback should stay there too. Computer lets you open the real file or local preview, then keep a note, shared link, uploaded file, or voice memo in the same workspace instead of losing the review in a separate chat app. The feedback remains connected to the real project, not a recreated attachment collection.
|
||||
Use the selected workspace as the destination for project material you want Computer to keep: a note, an imported file, a shared link or text, or a chat message that records a decision. This page covers where each kind of material goes and what to check before you save it. For the end-to-end review workflow, see [Review a prototype and capture feedback](/ecosystem/computer/use-cases/review-and-share-feedback).
|
||||
|
||||
## Start with the review you actually have
|
||||
## Choose the right destination
|
||||
|
||||
- Open a file from the workspace when you need to inspect its source or a supported preview.
|
||||
- Open a local port preview when the thing to review is a running web prototype.
|
||||
- Use a workspace chat or note for written feedback.
|
||||
- Use the installed app's share flow when your browser and device support it, or upload/copy the item through the workspace when they do not.
|
||||
| Material | Keep it as | Check before saving |
|
||||
| --- | --- | --- |
|
||||
| A decision, question, or written feedback | A workspace chat message or note | The selected workspace path and the file or screen it refers to |
|
||||
| A file you want the project to own | An import in the chosen workspace folder | The destination folder and whether the file contains sensitive material |
|
||||
| A link or shared text from a supported installed web app | A chat draft or note, depending on the selected share action | The workspace chosen by the share flow and the resulting draft before you send it |
|
||||
| A voice or image input | The relevant chat or configured media workflow | The configured provider and what data it may receive |
|
||||
|
||||
## Keep one piece of feedback with the right project
|
||||
Open real project files with the [file editor and browser](./files-editor). Open a running app with a [local-port preview](./local-port-previews). For voice, images, and memos, see [Voice, images, and memos](/ecosystem/computer/agents/voice-images-and-memos).
|
||||
|
||||
1. Select the workspace and confirm its path before opening or sharing anything.
|
||||
2. Open the file, image, supported preview, or local app you want to review.
|
||||
3. Add a concise workspace chat message or note that names the file, screen, or decision. On a supported installed web app, share a link, file, or text into Computer and choose the workspace destination.
|
||||
4. If audio is easier, record a short voice memo after audio features are enabled. Keep the recording and, when configured, its transcript with the workspace.
|
||||
## Use the installed app's share flow
|
||||
|
||||
## Check the result
|
||||
When the browser and device support the installed web app, you can share supported text, links, images, PDFs, or code into Computer. The share flow asks whether to create a chat draft or workspace note, or to import a file into a destination folder. It opens Computer with that pending item; review the destination and content before completing the action.
|
||||
|
||||
Reopen the workspace from another device. The original file or local preview is still associated with the expected project, and the feedback appears as the intended chat, note, imported file, or memo. If an item arrived in the wrong place, stop before acting on it and use the workspace picker to locate it rather than resending it repeatedly.
|
||||
If sharing is unavailable on the device, copy the text into the workspace chat or note, or import the file through Computer instead. The share payload is held locally by the browser until it opens Computer; it is not a background sync service.
|
||||
|
||||
## Limits and trust
|
||||
## Limits and recovery
|
||||
|
||||
Computer can only preview formats and share behaviors supported by the current browser and device. A local app must already be running before it can appear as a port preview. Sharing or uploading writes real data into the selected workspace, and hosted speech or image features can send media to their configured provider. Keep the instance private and do not use a review note as a place for secrets.
|
||||
Computer can only preview formats and share behaviors supported by the current browser and device. Sharing or importing writes real data to the selected workspace. If an item appears in the wrong project, stop before acting on it, switch to the correct workspace, and move or import it deliberately rather than repeatedly resending it. Hosted speech or image features can send media to their configured provider, so keep the instance private and do not use notes or media input for secrets.
|
||||
|
||||
@@ -1,6 +1,6 @@
|
||||
---
|
||||
title: "Search and chats"
|
||||
sidebar_position: 7
|
||||
sidebar_position: 9
|
||||
---
|
||||
|
||||
# Search and chats
|
||||
|
||||
@@ -26,7 +26,7 @@ This is where ideas become working software. Ask a question, get a running scrip
|
||||
|
||||
|
||||
:::tip Looking for "your computer in a browser" instead?
|
||||
Open Terminal is a tool the **chat AI** drives. If you instead want to operate your own machine from any browser (files, editor, terminal and git, mobile-first), with AI as an optional helper, see [**Open WebUI Computer**](/ecosystem/computer), a separate Open WebUI project. [Compare the two](/ecosystem/computer#open-webui-computer-vs-open-terminal).
|
||||
Open Terminal is a tool the **chat AI** drives. If you instead want to operate your own machine from any browser (files, editor, terminal and git, mobile-first), with AI as an optional helper, see [**Open WebUI Computer**](/ecosystem/computer), a separate Open WebUI project. [Choose the right execution environment](/ecosystem/computer/choose).
|
||||
:::
|
||||
|
||||
---
|
||||
|
||||
Reference in New Issue
Block a user