Compaction: preserve opaque identifiers in summaries (openclaw#25553) thanks @rodrigouroz

Verified:
- pnpm install --frozen-lockfile
- pnpm build
- pnpm check
- pnpm test:macmini

Co-authored-by: rodrigouroz <384037+rodrigouroz@users.noreply.github.com>
Co-authored-by: Tak Hoffman <781889+Takhoffman@users.noreply.github.com>
This commit is contained in:
Rodrigo Uroz
2026-02-27 11:14:05 -03:00
committed by GitHub
parent 84a88b2ace
commit 0fe6cf06b2
15 changed files with 344 additions and 4 deletions

View File

@@ -920,6 +920,10 @@ export const FIELD_HELP: Record<string, string> = {
"Minimum floor enforced for reserveTokens in Pi compaction paths (0 disables the floor guard). Use a non-zero floor to avoid over-aggressive compression under fluctuating token estimates.",
"agents.defaults.compaction.maxHistoryShare":
"Maximum fraction of total context budget allowed for retained history after compaction (range 0.1-0.9). Use lower shares for more generation headroom or higher shares for deeper historical continuity.",
"agents.defaults.compaction.identifierPolicy":
'Identifier-preservation policy for compaction summaries: "strict" prepends built-in opaque-identifier retention guidance (default), "off" disables this prefix, and "custom" uses identifierInstructions. Keep "strict" unless you have a specific compatibility need.',
"agents.defaults.compaction.identifierInstructions":
'Custom identifier-preservation instruction text used when identifierPolicy="custom". Keep this explicit and safety-focused so compaction summaries do not rewrite opaque IDs, URLs, hosts, or ports.',
"agents.defaults.compaction.memoryFlush":
"Pre-compaction memory flush settings that run an agentic memory write before heavy compaction. Keep enabled for long sessions so salient context is persisted before aggressive trimming.",
"agents.defaults.compaction.memoryFlush.enabled":