Skip to content
WhoseDev logoWhoseDev Docs

Tab completions

WhoseDev Tab provides ghost-text inline completions while you edit. Suggestions appear after a short debounce and can come from a local Ollama model, WhoseDev Cloud, or an automatic route that picks the best available option.

Tab completions are separate from Ask, Plan, and Act agent sessions. Agent modes never silently fall back to cloud when you chose local; Tab has its own routing settings.

Enable Tab

Section titled “Enable Tab”

Setting whosedev.com.inlineCompletions.enabled controls whether Tab suggestions run. When enabled, typing triggers completion requests according to your routing configuration.

Choose a route

Section titled “Choose a route”

Setting whosedev.com.inlineCompletions.routing:

Value Behavior
local Use Ollama on this machine
cloud Use WhoseDev-managed inline completion backend
auto Prefer local when available, otherwise cloud (subject to fallback settings)

Configure local models in WhoseDev AI Settings or with:

  • whosedev.com.inlineCompletions.localModel — member and single-line completions
  • whosedev.com.inlineCompletions.localMultilineModel — short multiline suggestions
  • whosedev.com.inlineCompletions.localBlockModel — multiline and block completions
  • whosedev.com.inlineCompletions.ollamaBaseUrl — Ollama daemon URL (shared with local agent routes)

Cloud fallback (Tab only)

Section titled “Cloud fallback (Tab only)”

whosedev.com.inlineCompletions.cloudFallback allows cloud inline completions when local Ollama is unavailable. This applies only to Tab—not to Ask, Plan, or Act sessions.

Accept and refine

Section titled “Accept and refine”
  • Press Tab to accept a suggestion when ghost text is visible.
  • After accepting, Tab can suggest a logical next edit on the same line or block.
  • Immediate triggers (such as member access after a dot) can bypass the normal debounce.

Status and feedback

Section titled “Status and feedback”

Optional indicators help you see when Tab is working:

Setting Purpose
whosedev.com.inlineCompletions.showPendingIndicator Subtle indicator while a suggestion is loading
whosedev.com.inlineCompletions.showProviderStatus Status bar readiness for local vs cloud Tab

Tuning

Section titled “Tuning”
Setting Purpose
whosedev.com.inlineCompletions.debounceMs Delay before a typing-triggered request
whosedev.com.inlineCompletions.contextTimeoutMs Max wait for local retrieval enrichment
whosedev.com.inlineCompletions.keepAlive Ollama keep_alive duration for loaded models

Troubleshooting

Section titled “Troubleshooting”

If Tab never appears:

  1. Confirm whosedev.com.inlineCompletions.enabled is on.
  2. For local routing, verify Ollama is running and models are installed (WhoseDev: Manage Offline Models).
  3. For cloud routing, sign in and confirm your plan includes managed AI.
  4. See Troubleshooting for Ollama connectivity issues.
Section titled “Related topics”