Module token_usage 

Per-file usage-summary cache on the redb token_usage table.

Keyed by canonical session-file path; the value is that file’s deduped FileUsageSummary. The cache is incremental: the scan reuses a file whose mtime and size still match the cached entry rather than re-parsing a transcript that can reach 100 MB.

Re-exports

use anyhow::Context; 🔒

use forge_agent::env::token_usage::FileUsageSummary; 🔒

use redb::TableDefinition; 🔒

use super::Db; 🔒

Constants

TOKEN_USAGE 🔒

Functions

decode 🔒

Decode a stored blob, tagging a failure with the owning path.

load

The cached summary for path, or None when absent (or the table doesn’t exist yet). A decode failure is surfaced so a corrupt row is diagnosable rather than silently treated as a cache miss.

store

Store summary for path, overwriting any prior entry.