continuedev/continue · error

Unable to connect to local Ollama instance. Ollama may not b

Error message

Unable to connect to local Ollama instance. Ollama may not be running.

What it means

Thrown in customFetch when a connection to http://127.0.0.1:11434 is refused and isOllamaInstalled() reports Ollama is installed. The message distinguishes 'installed but not running' from 'not installed', so this variant specifically means the binary/app exists but no server is listening on the default port.

Source

Thrown at core/llm/index.ts:500

        } else if (e.message.includes("/api/show")) {
          throw new Error(
            `HTTP ${e.response.status} ${e.response.statusText} from ${e.response.url}\n\n${e.response.body}`,
          );
        } else {
          if (!isAbortError(e)) {
            // Don't pollute console with abort errors. Check on name instead of instanceof, to avoid importing node-fetch here
            console.debug(
              `${e.message}\n\nCode: ${e.code}\nError number: ${e.errno}\nSyscall: ${e.erroredSysCall}\nType: ${e.type}\n\n${e.stack}`,
            );
          }
          if (
            e.code === "ECONNREFUSED" &&
            e.message.includes("http://127.0.0.1:11434")
          ) {
            const message = (await isOllamaInstalled())
              ? "Unable to connect to local Ollama instance. Ollama may not be running."
              : "Unable to connect to local Ollama instance. Ollama may not be installed or may not running.";
            throw new Error(message);
          }
          if (
            e.code === "ECONNREFUSED" &&
            e.message.includes("http://localhost:8000")
          ) {
            const isInstalled = await isLemonadeInstalled();
            let message: string;
            if (process.platform === "linux") {
              // On Linux, isLemonadeInstalled checks if it's running (via health endpoint)
              message =
                "Unable to connect to local Lemonade instance. Please ensure Lemonade is running. Visit http://lemonade-server.ai for setup instructions.";
            } else {
              // On Windows, we can check if it's installed
              message = isInstalled
                ? "Unable to connect to local Lemonade instance. Lemonade server may not be running."
                : "Unable to connect to local Lemonade instance. Lemonade may not be installed or may not be running.";
            }
            throw new Error(message);

View on GitHub (pinned to 5522c6f44c)

Solutions

1. Start the Ollama server: run `ollama serve` (or launch the desktop app)
2. Verify it's listening: curl http://127.0.0.1:11434/api/tags
3. If OLLAMA_HOST is customized, point the client's apiBase at that exact host:port
4. Enable autostart (systemd user unit / login item) if this recurs after reboots

Example fix

# before
# ECONNREFUSED http://127.0.0.1:11434
# after
ollama serve &
curl http://127.0.0.1:11434/api/tags  # 200 {"models":[...]}

const up = await fetch('http://127.0.0.1:11434/api/tags').then(r => r.ok).catch(() => false); if (!up) await spawn('ollama', ['serve']);

catch (e) { if (e.message.includes('Unable to connect to local Ollama')) { await startOllamaAndWaitForPort(11434); retry(); } else throw e; }

When it happens

Trigger: ECONNREFUSED to 127.0.0.1:11434 while Ollama is installed: the daemon isn't running, was stopped, or is bound to a different host/port (OLLAMA_HOST).

Common situations: Ollama desktop app quit or crashed; ollama serve not started on a headless box; OLLAMA_HOST=0.0.0.0 or a custom port so 127.0.0.1:11434 is not bound; system reboot without autostart.

Related errors

• Error fetching tags: ${e.message}
• HTTP ${e.response.status} ${e.response.statusText} from ${e.
• Failed to fetch Ollama library: ${response.status}
• j.error
• Error parsing Ollama response: ${e} ${chunk}

AI-assisted analysis of continuedev/continue@5522c6f44c (2026-08-27). Data as JSON: /api/errors/454d099bad108d20. Report an issue: GitHub.