nikivdev/code · error

{error}

Error message

{error}

What it means

This error is raised by `maple_call_tool` when the JSON-RPC call itself succeeded (no envelope-level error) but the tool's *result* payload embeds a domain-level error, detected by `maple_tool_result_error`. The library re-bails with that inner error string verbatim (`{error}`), so the message text comes from the Maple tool, not from this crate. It marks the difference between transport/protocol failures and tool-execution failures.

Source

Thrown at src/codex_telemetry.rs:464

        .filter(|value| !value.is_empty())
        .or_else(|| Some("Maple tool returned an unspecified error".to_string()))
}

fn maple_call_tool(
    config: &MapleReadConfig,
    name: &str,
    arguments: serde_json::Value,
) -> Result<serde_json::Value> {
    let result = maple_json_rpc_request(
        config,
        "tools/call",
        serde_json::json!({
            "name": name,
            "arguments": arguments,
        }),
    )?;
    if let Some(error) = maple_tool_result_error(&result) {
        anyhow::bail!("{error}");
    }
    Ok(result)
}

pub fn status() -> Result<CodexTelemetryStatus> {
    let config = parse_maple_exporter_config_from_env()?;
    let state = load_state()?;
    let state_path = telemetry_state_path()?;
    let events_path = codex_skill_eval::events_log_path()?;
    let outcomes_path = codex_skill_eval::outcomes_log_path()?;

    Ok(CodexTelemetryStatus {
        enabled: config.is_some(),
        configured_targets: config
            .as_ref()
            .map(|value| value.targets.len())
            .unwrap_or(0),
        service_name: config

View on GitHub (pinned to a747e741ae)

Solutions

1. Act on the inner message: if the trace is not found, flush first (`inspect_trace` with flush_first=true) and re-check the trace_id.
2. Verify the MAPLE_API_TOKEN has access to the requested trace/workspace.
3. Retry the tool call if the message indicates a transient internal error.
4. Check Maple server logs for the tool-side failure if the message is opaque.

Example fix

// before: inspecting a stale trace id
inspect_trace("abc123", false)?;
// after: flush pending telemetry first so the trace exists server-side
inspect_trace("abc123", true)?;

// ensure a trace id exists before inspecting it
let known = list_recent_trace_ids()?;
if !known.contains(&trace_id.to_string()) {
    eprintln!("trace {trace_id} unknown; flushing telemetry first");
    flush(64);
}

fn tool_result_ok(result: &serde_json::Value) -> bool {
    maple_tool_result_error(result).is_none()
}

match maple_call_tool("inspect_trace", args) {
    Err(e) if !e.to_string().contains("Maple MCP") => {
        // tool-level (domain) error: inspect the message for 'not found' / 'denied'
        eprintln!("Maple tool failed: {e}");
    }
    Err(e) => return Err(e),
    Ok(r) => r,
}

When it happens

Trigger: `maple_call_tool` (called by `inspect_trace`) obtains a result envelope where `maple_tool_result_error(&result)` returns `Some`, e.g. the tool reports 'trace not found', permission denied for the trace, or an internal tool failure, and the function bails with that string.

Common situations: Querying a trace_id that has been flushed/expired on the server, requesting a trace owned by another API token (RBAC denial), or the Maple tool hitting an internal error while aggregating telemetry.

Related errors

• Maple MCP request failed ({}): {}
• Maple MCP error {code}: {message}
• Gemini API error {}: {}
• FLOW_CODEX_MAPLE_LOCAL_ENDPOINT and FLOW_CODEX_MAPLE_LOCAL_I
• FLOW_CODEX_MAPLE_HOSTED_ENDPOINT and FLOW_CODEX_MAPLE_HOSTED

AI-assisted analysis of nikivdev/code@a747e741ae (2026-09-01). Data as JSON: /api/errors/833bf95a512ed663. Report an issue: GitHub.