Add Usage.cost.totalFromProvider to record whether cost.total is the
billed cost reported by the provider or an estimate from catalog rates.
Catalog-derived totals set it to false; a missing value also means
false. combineUsage marks a sum as provider-reported only when both
parts are.
Compaction summaries were limited to 0.8 * compaction.reserveTokens (13107 by default), turn-prefix summaries to 0.5 * reserveTokens, and branch summaries to a hardcoded 4096. On providers where reasoning shares the output limit (OpenAI Responses, adaptive Anthropic), high thinking levels used most of that limit and compaction failed with "generation hit the token cap", even though the model allows far more output.
Tying the limit to reserveTokens assumed the summary request is as large as the full context, so its output had to fit in the reserved space. It is usually much smaller: recent messages kept by keepRecentTokens are not sent, a split turn goes to a separate request, tool results are truncated, and tool definitions and the normal system prompt are not included. Manual compaction can also run long before the context is full. The provider layer already sizes output from the actual request, using the model's output limit reduced to fit the remaining context, as for normal turns.
Summary requests now set no maxTokens and rely on that. compaction.reserveTokens only controls when compaction starts.
generateSummary() and generateSummaryWithUsage() now ignore the reserveTokens parameter.
closes#9512, closes#9075, closes#10908
Separate conversation content from instructions and replace prefix/suffix wording with continuation-oriented summarization guidance.
Hopefully this avoids the refusals.
Fixes#9652
This change makes system prompt text and tool changes part of the transcript rather than silently rewriting its starting conditions. This lets Pi record when instructions changed or tools became available, restore that state after resuming or navigating branches, and preserve cached prompt prefixes where the upstream supports it.
Add provider-context prefix preparation and optional summary request settings needed by cache-friendly compaction. Existing AgentSession callers continue to pass undefined, so standalone summarization remains the default until the feature is wired in.
* add usage info to branch summary entries
* add usage to compaction entries
* allow custom tools to report llm usage in tool results
* allow observing and patching usage in tool_result hooks
* agent-harness: save usage in entries
for compaction, branch summaries and tool results
The root barrel is now core-only and side-effect free: types,
createModels/createProvider, auth substrate, lazyStream/lazyApi, faux,
utils. Generated catalogs, api-registry, env-api-keys, images, global
stream functions, and per-API lazy wrappers leave the root.
New @earendil-works/pi-ai/compat preserves the old surface verbatim as
a strict superset of the root: api-dispatch stream/complete with env
key injection, the builtin registration side effect (skip-if-present so
it cannot clobber earlier overrides), deprecated getModel/getModels/
getProviders aliases of the new getBuiltin* reads in providers/all,
lazy api wrappers + setBedrockProviderModule, and image generation.
Compat dies with the coding-agent ModelManager migration.
Packaging: exports map gains ./compat, ./providers/*, ./api/*;
sideEffects array lists only the effectful modules.
Old-global imports across agent/coding-agent/examples and pi-ai tests
switch to /compat (path-only; compat is a superset). The coding-agent
extension loader resolves the pi-ai ROOT specifier to compat, so
existing user extensions using the old global API keep working at
runtime until compat is removed. vitest configs alias /compat to src;
browser smoke imports old globals from /compat.
