Continue autocomplete gone quiet in your Swift files? Work through the most common causes — model config, extension state, file type, and conflicts — with clear fixes.
You are typing Swift in VS Code and Continue's inline suggestions never appear. The chat panel might still work, but the gray ghost text you expect while typing is missing.
This is one of the most common Continue issues, and it usually comes down to configuration rather than a real bug. Autocomplete and chat are separate features that can be configured independently.
Before changing anything, confirm what still works. If chat responds but autocomplete does not, the problem is almost certainly in the autocomplete model configuration.
Work through the checks below in order. Each one rules out a common cause, and stopping at the first fix that works saves time.
Continue lets you assign a specific model to autocomplete, separate from chat. If no autocomplete model is set, you will get no inline suggestions even though chat is fine.
Open your Continue configuration and look for the autocomplete or tab-completion model setting. Confirm a valid model is assigned there, not just a chat model.
The official docs describe the exact configuration key and the recommended models for completion. Follow that guidance, since completion models are often different from chat models.
If the field is empty or points at a model your provider does not serve, that is your fix. Set a valid autocomplete model, save, and reload.
Confirm Continue is actually enabled for this workspace. Open the Extensions view and verify it is installed, enabled, and not disabled for the current folder.
Update to the latest version. Autocomplete behavior changes across releases, and an old build can misbehave in ways a newer one has already fixed.
Reload the VS Code window after any change. A stale extension host is a surprisingly common reason features silently stop responding.
If Continue's sidebar icon is missing entirely, the extension is not loaded. Reinstalling and reloading resolves that class of problem quickly.
VS Code has its own editor settings that govern inline suggestions. If those are disabled, no extension's ghost text will appear, Continue included.
Open VS Code settings and confirm inline suggestions are turned on. This is an editor-level toggle independent of Continue.
Also confirm the accept key is what you expect. If Tab is bound to something else in your keybindings, suggestions may be generating but not accepting.
After enabling inline suggestions, test again in a Swift file. Sometimes this single editor toggle is the entire fix.
Autocomplete needs the model to respond quickly. If your provider is unreachable or your API key is invalid, completions silently fail while you keep typing.
Open the Continue chat and send a test request using the same provider. If chat also errors, the problem is your provider connection, not autocomplete specifically.
For local models, make sure your runner is actually running. A stopped local service produces exactly this symptom of no suggestions with no obvious error.
Check Continue's logs or output panel for connection or authentication errors. Those messages usually name the real cause directly.
Confirm VS Code recognizes the file as Swift. If the language mode in the status bar shows plain text, extensions may not engage as expected.
Set the language mode to Swift explicitly if needed, and consider installing the official Swift extension for proper recognition. Correct file typing helps the whole toolchain.
Give completions a moment. On slower models, especially large local ones, suggestions can lag enough that they feel absent when they are merely delayed.
If latency is the issue, switch autocomplete to a smaller, faster model. Completion benefits far more from speed than from raw model size.
Other AI or completion extensions can compete for the same inline suggestion slot. When two tools both try to provide ghost text, one may win and the other appears broken.
Temporarily disable other AI completion extensions and test Continue alone. If suggestions return, you have found a conflict.
Decide which tool owns autocomplete and disable the rest, or configure them so they do not overlap. Running one inline provider at a time is the cleanest setup.
Re-enable other extensions one at a time if you need them, checking autocomplete after each. This isolates exactly which one interferes.
When several checks pass but suggestions are still missing, reduce the problem to its simplest form. Close and reopen VS Code entirely, since a full restart clears a stuck extension host more reliably than a soft reload.
Create a tiny throwaway Swift file with an obvious completion, like a half-typed function signature, and test there. A minimal file removes workspace-specific noise and tells you whether autocomplete works at all.
If the minimal case succeeds but your real project does not, the difference is something in that project or its settings. Look at workspace-level configuration and any per-folder overrides that might disable inline suggestions.
If even the minimal case fails, the issue is your Continue or model configuration rather than the project. That distinction alone often points you straight at the fix and saves a long, unfocused search.
A lot of confusion around this problem disappears once you internalize that autocomplete and chat are independent. They can use different models, and one can be perfectly healthy while the other is misconfigured.
That is why a working chat panel is not proof that autocomplete is set up. The two features are wired separately in Continue, so each needs its own valid model assignment.
Keeping this distinction in mind reshapes your debugging. When suggestions vanish but chat still answers, you can skip provider-wide theories and go straight to the autocomplete model configuration.
It also guides prevention. When you first set up Continue, deliberately assign and test a completion model rather than assuming a chat model covers both roles, and this whole class of problem becomes far less likely.
If none of the checks worked, reset to a known-good state. Reinstall Continue, reload VS Code, and reconfigure a single simple provider for both chat and autocomplete.
Test with a well-supported hosted model before returning to a local or exotic setup. Narrowing to one variable at a time is the fastest path to the culprit.
Read Continue's output logs carefully, since they often state the failure plainly. Copy any error text before searching for it.
If the issue persists, check Continue's GitHub issues for similar reports or open a new one with your config and logs. As an open-source project, that is where fixes and answers happen.
Chat and autocomplete use separate model assignments in Continue. If chat works but autocomplete is silent, the most likely cause is that no valid autocomplete model is configured.
Autocomplete works best with fast completion-oriented models, which are often different from chat models. Check Continue's docs for recommended autocomplete models and set one explicitly.
Yes. Other AI completion extensions can compete for the inline suggestion slot. Temporarily disable them and test Continue alone to check for a conflict.
Large models, especially local ones, can lag enough that suggestions feel missing. Switch autocomplete to a smaller, faster model to improve responsiveness.
Open Continue's output or log panel in VS Code. Connection, authentication, and model errors are usually reported there in plain text.