What causes "learning_update_facet: upsert failed: {e:#}" in tinyhumansai/openhuman?

Retry the update once the concurrent writer (rebuild/reset/reflection) finishes

learning_update_facet: upsert failed: {e:#}

Error message

learning_update_facet: upsert failed: {e:#}

What it means

Thrown by learning_update_facet when the facet was read and mutated in memory but FacetCache::upsert() (ProfileStore::upsert_full) failed to persist it. It is the write-path twin of the get error: the value change and the UserState::Pinned transition were NOT saved. Causes are SQLite write failures — busy database, disk full, constraint/I/O errors.

Source

Thrown at src/openhuman/agent/learning/tools.rs:254

        PermissionLevel::Write
    }

    async fn execute(&self, args: serde_json::Value) -> anyhow::Result<ToolResult> {
        log::debug!("[tool][learning] update_facet invoked");
        let class_str = read_required_str(&args, "class")?;
        let key_suffix = read_required_str(&args, "key")?;
        let value = read_required_str(&args, "value")?;
        let fk = full_key(&class_str, &key_suffix);
        let cache = get_cache()?;
        let mut facet = cache
            .get(&fk)
            .map_err(|e| anyhow::anyhow!("learning_update_facet: {e:#}"))?
            .ok_or_else(|| anyhow::anyhow!("learning_update_facet: facet not found: {fk}"))?;
        facet.value = value;
        facet.user_state = UserState::Pinned;
        cache
            .upsert(&facet)
            .map_err(|e| anyhow::anyhow!("learning_update_facet: upsert failed: {e:#}"))?;
        Ok(ToolResult::success(serde_json::to_string(&json!({
            "facet": facet_to_json(&facet),
        }))?))
    }
}

/// Set/clear a facet's pin via `set_user_state`. Shared by pin/unpin.
async fn set_pin(
    args: serde_json::Value,
    tool: &str,
    state: UserState,
) -> anyhow::Result<ToolResult> {
    let class_str = read_required_str(&args, "class")?;
    let key_suffix = read_required_str(&args, "key")?;
    let fk = full_key(&class_str, &key_suffix);
    let cache = get_cache()?;
    let updated = cache
        .set_user_state(&fk, state)

View on GitHub (pinned to a221052e0d)

Solutions

Retry the update once the concurrent writer (rebuild/reset/reflection) finishes
Verify the facet's persisted state afterwards with learning_get_facet — if the old value is still there the upsert truly did not land
Check disk space and workspace_dir DB permissions
Ensure only one core process uses the workspace (OPENHUMAN_CORE_REUSE_EXISTING debugging leftovers are a classic double-open)

Defensive patterns

Strategy: retry

Try / catch

let res = tool_exec("learning_update_facet", args).await;
if let Err(e) = &res {
    if e.to_string().contains("upsert failed") && e.to_string().contains("locked") {
        tokio::time::sleep(Duration::from_millis(500)).await;
        return tool_exec("learning_update_facet", args).await; // idempotent value write
    }
}
res

Prevention

After any reported upsert failure, verify with learning_get_facet whether the value landed before re-sending
Do not interleave learning_rebuild_cache and facet updates in the same turn
Watch disk space — full disks turn transient lock errors into persistent I/O errors

When it happens

Trigger: learning_update_facet executes its UPDATE while a stability rebuild or learning_reset_cache is writing the same table; disk full; DB opened read-only or corrupted. The tool had already fetched the facet successfully, so the read path was healthy at that moment.

Common situations: Update racing learning_rebuild_cache or learning_forget_facet on the same key; low-disk laptops; a second core instance pointed at the same workspace_dir.

Related errors

AI-assisted analysis of tinyhumansai/openhuman@a221052e0d (2026-08-16). Data as JSON: /api/errors/8ae967f02bb9392b. Report an issue: GitHub.