janhq/jan · error · Error

model-errors:noRunningSession

model-errors:noRunningSession

Error message

model-errors:noRunningSession

What it means

Thrown by createLlamaCppModel when the Tauri invoke('plugin:llamacpp|find_session_by_model', { modelId }) returns null after the start phase. The model either was never started (provider was undefined so start was skipped) or started but no session registered for that modelId (model-factory.ts:935). Message is i18n key 'model-errors:noRunningSession'.

Source

Thrown at web-app/src/lib/model-factory.ts:935

        console.error('Failed to start llamacpp model:', error)
        // A serialized engine error is a plain object, so the previous
        // `instanceof Error` path stringified it into raw JSON for the user.
        throw new Error(
          i18n.t('model-errors:startModelFailed', {
            reason: describeEngineError(error),
          })
        )
      }
    }

    // Get session info which includes port and api_key
    const sessionInfo = await invoke<SessionInfo | null>(
      'plugin:llamacpp|find_session_by_model',
      { modelId }
    )

    if (!sessionInfo) {
      throw new Error(
        i18n.t('model-errors:noRunningSession', { model: modelId })
      )
    }

    const onLlamacppServerError = provider
      ? () => {
          void (async () => {
            try {
              const { useServiceStore } = await import('@/hooks/useServiceHub')
              const hub = useServiceStore.getState().serviceHub
              await hub?.models().reloadModel(provider, modelId)
            } catch (e) {
              console.warn('[llamacpp] reload after crash failed:', e)
            }
          })()
        }
      : undefined
    // The global toggle can strip reasoning_content from resent assistant turns,

View on GitHub (pinned to fad3f12a14)

Solutions

1. Ensure provider is passed into createLlamaCppModel so startModel actually runs.
2. Avoid concurrent model swaps / page reloads during boot; let startModel settle before issuing inference.
3. If reproducible, check the llama.cpp extension logs for a crashed or never-registered session.
4. Verify the modelId used for find_session_by_model matches the one startModel registered (no prefix/suffix drift).

Example fix

// before
if (!sessionInfo) {
  throw new Error(i18n.t('model-errors:noRunningSession', { model: modelId }))
}
// after: retry the lookup once after a short delay to absorb start/register race
if (!sessionInfo) {
  await new Promise(r => setTimeout(r, 500))
  const retried = await invoke<SessionInfo | null>('plugin:llamacpp|find_session_by_model', { modelId })
  if (!retried) throw new Error(i18n.t('model-errors:noRunningSession', { model: modelId }))
  return buildLlamaCppModel(modelId, retried, ...)
}

// ensure provider is passed so the start block runs
function assertProviderForLocal(provider: ProviderObject | undefined, modelId: string) {
  if (!provider) throw new Error(`Cannot start ${modelId}: provider object is required to boot a local session`)
}

function isSessionInfo(x: unknown): x is SessionInfo {
  return typeof x === 'object' && x !== null && typeof (x as SessionInfo).port === 'number'
}

let sessionInfo = await invoke<SessionInfo | null>('plugin:llamacpp|find_session_by_model', { modelId })
if (!sessionInfo) {
  // one bounded retry to absorb the start/register race
  await new Promise(r => setTimeout(r, 500))
  sessionInfo = await invoke<SessionInfo | null>('plugin:llamacpp|find_session_by_model', { modelId })
}
if (!sessionInfo) throw new Error(i18n.t('model-errors:noRunningSession', { model: modelId }))

When it happens

Trigger: provider is falsy so the startModel block is skipped entirely; startModel resolved but the sidecar crashed between start and find_session_by_model; a different modelId string is used for the lookup than was used for start (id drift); the session table was cleared (KILL_SIDECAR event) mid-flight.

Common situations: Reloading the page or cancelling inference right as the model boots; switching models rapidly so the prior session is torn down; the sidecar was killed by a data-folder relocation or settings change; modelId includes a provider prefix in one call but not the other.

Related errors

• No running MLX session found for model: ${modelId}
• model-errors:startModelFailed
• llamacpp extension not found
• Failed to decompress archive: ${String(e)}
• ${String(e)}

AI-assisted analysis of janhq/jan@fad3f12a14 (2026-08-12). Data as JSON: /api/errors/df18198f414f34cf. Report an issue: GitHub.