What causes "RPC envelope contains undefined data" in tinyhumansai/openhuman?

Read the text after 'config load failed:' — it names the actual cause and, for TOML errors, the file and line to fix

RPC envelope contains undefined data

Error message

RPC envelope contains undefined data

What it means

Startup error from `openhuman subconscious tick`: `Config::load_or_init()` failed and the underlying error is appended verbatim after the colon. load_or_init reads (or creates) the workspace config and applies OPENHUMAN_* env overrides, so the appended cause is typically a TOML parse error with file/line, a filesystem/permission error on the workspace dir, or an invalid env-override value. The wrapper uses map_err with display formatting, so the original error chain is flattened into the message text.

Source

Thrown at app/src/services/api/agentProfilesApi.ts:12

import type { AgentProfile, AgentProfilesResponse } from '../../types/agentProfile';
import { callCoreRpc } from '../coreRpcClient';

interface Envelope<T> {
  data?: T;
}

function unwrapEnvelope<T>(response: Envelope<T> | T): T {
  if (response && typeof response === 'object' && 'data' in response) {
    const envelope = response as Envelope<T>;
    if (envelope.data === undefined) {
      throw new Error('RPC envelope contains undefined data');
    }
    return envelope.data;
  }
  return response as T;
}

export const agentProfilesApi = {
  list: async (): Promise<AgentProfilesResponse> => {
    const response = await callCoreRpc<Envelope<AgentProfilesResponse>>({
      method: 'openhuman.profiles_list',
    });
    return unwrapEnvelope(response);
  },

  select: async (profileId: string): Promise<AgentProfilesResponse> => {
    const response = await callCoreRpc<Envelope<AgentProfilesResponse>>({
      method: 'openhuman.profiles_select',
      params: { profile_id: profileId },

View on GitHub (pinned to a221052e0d)

Solutions

Read the text after 'config load failed:' — it names the actual cause and, for TOML errors, the file and line to fix
Check the workspace dir is readable+writable by the process user (default ~/.openhuman; or the path passed via --workspace / OPENHUMAN_WORKSPACE)
Validate the config file parses as TOML before retrying (e.g. `python3 -c "import tomllib,sys; tomllib.load(open(sys.argv[1],'rb'))" <file>`)
Bisect a corrupt workspace from an environment problem: `openhuman subconscious tick --workspace /tmp/fresh-ws`

Example fix

# before: config load failed: expected equals, found newline at line 12
#   (broken hand edit in the workspace config.toml)
# after: fix line 12 quoting, or start clean:
openhuman subconscious tick --workspace /tmp/fresh-ws

Defensive patterns

Strategy: try-catch

Validate before calling

# bash: pre-flight the workspace before ticking
ws="${OPENHUMAN_WORKSPACE:-$HOME/.openhuman}"
[ -d "$ws" ] || mkdir -p "$ws" 2>/dev/null || { echo "workspace not writable: $ws" >&2; exit 2; }
[ -w "$ws" ] || { echo "workspace not writable: $ws" >&2; exit 2; }
cfg=$(find "$ws" -maxdepth 2 -name '*.toml' | head -1)
[ -z "$cfg" ] || python3 -c "import tomllib,sys; tomllib.load(open(sys.argv[1],'rb'))" "$cfg" 2>/dev/null || { echo "config TOML invalid: $cfg" >&2; exit 2; }

Try / catch

# Rust caller: the anyhow context flattens the chain, so match on the message prefix
match run_subconscious_command(&args) {
    Err(e) if e.to_string().starts_with("config load failed:") => {
        eprintln!("config/workspace problem — fix the cause after the colon, or tick a fresh --workspace");
        std::process::exit(78); // EX_CONFIG
    }
    other => other,
}

Prevention

Treat the text after 'config load failed:' as the real diagnosis — it carries the underlying IO/TOML error verbatim
Keep workspace dirs writable by the service user; in containers set HOME or OPENHUMAN_WORKSPACE to a writable volume
Validate config edits with a TOML parser before restarting the daemon
Use a scratch --workspace to bisect corrupt-config vs environment failures

When it happens

Trigger: Hand-edited config.toml with broken quoting/syntax (cause shows the line); workspace directory (default ~/.openhuman, or your --workspace) unwritable — read-only volume, wrong owner, disk full; a malformed OPENHUMAN_* env override that fails schema parsing; partial/corrupt write from a crash during config migration.

Common situations: First run inside a container with no writable HOME; editing config with an editor that changed quoting; a version upgrade that partially migrated config schema; --workspace pointing at a path that cannot be created.

Related errors

AI-assisted analysis of tinyhumansai/openhuman@a221052e0d (2026-08-16). Data as JSON: /api/errors/e5a110a8916314e1. Report an issue: GitHub.