chenhg5/cc-connect · error

%s read response: %w

Error message

%s read response: %w

What it means

rpcRequestOverIO reads newline-delimited JSON-RPC responses from the codex app-server; when ReadBytes fails (EOF because the process exited, or a pipe I/O error) before a matching response arrives, it returns an error prefixed with the RPC method name. This means the app-server died or closed its stdout before answering the request, not that the response was malformed.

Source

Thrown at agent/codex/session.go:736

	return strings.TrimSpace(resp.Config.Model), normalizeRuntimeReasoningEffort(stringValue(resp.Config.ModelReasoningEffort)), nil
}

func rpcRequestOverIO(stdin io.Writer, reader *bufio.Reader, id int64, method string, params any, out any) error {
	payload := map[string]any{
		"jsonrpc": "2.0",
		"id":      id,
		"method":  method,
		"params":  params,
	}
	if err := writeRPCMessage(stdin, payload); err != nil {
		return err
	}

	for {
		line, err := reader.ReadBytes('\n')
		if err != nil {
			return fmt.Errorf("%s read response: %w", method, err)
		}

		var probe map[string]json.RawMessage
		if err := json.Unmarshal(bytes.TrimSpace(line), &probe); err != nil {
			continue
		}
		if _, ok := probe["id"]; !ok {
			continue
		}

		var resp rpcResponseEnvelope
		if err := json.Unmarshal(bytes.TrimSpace(line), &resp); err != nil {
			continue
		}
		respID, ok := rpcIDToInt64(resp.ID)
		if !ok || respID != id {
			continue
		}

View on GitHub (pinned to 4000b2338a)

Solutions

  1. Run the codex binary manually (`codex app-server` or the configured subcommand) to see why it exits — startup crashes are the usual cause.
  2. Update the Codex CLI to a version that supports app-server JSON-RPC mode.
  3. Check `command` and args in the codex agent config for typos or flags the installed version rejects.
  4. Check authentication (`codex login`) — some versions exit when credentials are missing.
  5. Capture the app-server's stderr output to identify the exit reason; retry once transient resource pressure is resolved.

Example fix

// before: old codex build exits on the app-server subcommand
$ codex --version
codex-cli 0.1.0  // unsupported
// after
$ npm i -g @openai/codex && codex --version
codex-cli 0.9.0
Defensive patterns

Strategy: retry

Validate before calling

// verify the app-server subcommand starts and answers before RPC use
cmd := exec.Command(cfg.Command, "app-server", "--help")
if err := cmd.Run(); err != nil {
    return fmt.Errorf("codex app-server unsupported or failing: %w", err)
}

Type guard

func isRPCReadError(err error, method string) bool {
    return err != nil && strings.HasPrefix(err.Error(), method+" read response:")
}

Try / catch

resp, err := rpcRequestOverIO(stdin, stdout, ctx, "initialize", params, &out)
if isRPCReadError(err, "initialize") {
    // app-server exited early; capture stderr and retry with a fresh process
    slog.Error("app-server died during RPC", "err", err, "stderr", stderr.String())
}

Prevention

When it happens

Trigger: reader.ReadBytes('\n') returns io.EOF or another read error inside rpcRequestOverIO while waiting for a reply to a method such as an initialization or config RPC issued by loadCodexRuntimeConfig.

Common situations: The codex app-server crashes or exits during startup (bad args, missing auth); the binary version does not support app-server mode and exits immediately; a request timeout in the caller kills the process while the RPC is pending; pipe broken by process kill.

Related errors

AI-assisted analysis of chenhg5/cc-connect@4000b2338a (2026-09-06). Data as JSON: /api/errors/ffb7a76c0bb9a716. Report an issue: GitHub.