This PR:
- Adds retainedTail to compaction entries in the new agent harness so we don't have to walk up the tree for the 2000 tokens before compaction,
- Changes getPathToRoot to getPathToRootOrCompaction to only load until last compaction, as unnecessary to access all nodes where it is called,
- Adds a SQLite storage backend, in a separate packages/session-backend-sqlite, with a migration system and schemas as per on-site discussions: sessions to match session header messages (except for metadata, which I couldn't understand what it's used for or where it gets written, so I omitted it), session_entries for shared entry types as columns plus payload as a json for what remains, session_sequences to represent the append-only, serialized nature of the jsonl files, branch_entries to attribute nodes to branches (relationship one-to-many), and session_materialized with the session info (see /session in TUI) to act as a "cache" or quick-access for costs, message count, token info, labels, session name, and model-thinking-level config (e.g. for fast resume).
- This is compatible with the new agent harness Session abstraction.
* 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
Allow callers to attach an opaque JSON object to the session header so
application context needed to rebuild a harness (for example an agent
profile reference) is readable from the first line alone, without
scanning session entries. The field is optional and ignored by readers
that do not use it; fork inherits the source metadata unless overridden.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* fix(ai,agent): stop salvaging malformed tool-call argument JSON
Finalized tool-call arguments must strict-parse (allowing only lossless
string-escape repair). When the streamed JSON is truncated or malformed,
the raw JSON is preserved on ToolCall.malformedArguments, arguments
become {}, and the agent loop refuses to execute the call with an error
tool result instead of running it with silently salvaged partial
arguments. Partial parsing is now only used for streaming previews.
* fix(agent): fail tool calls from length-truncated assistant messages
Reverts the malformed-arguments tracking approach in favor of handling
truncation in the agent loop: when an assistant message stops with
"length", all tool calls in it are potentially borked (streamed
arguments are salvage-parsed), so none are executed. Each gets an error
tool result telling the model to re-issue the call, and the loop
continues. No new fields on ToolCall and no provider changes.
fixes#6284
AgentHarnessOptions.models is required; the harness stream path,
compaction, and branch summarization go through models.streamSimple()/
completeSimple() instead of the compat globals. getApiKeyAndHeaders
stays and wins per-field over provider-resolved auth, but is no longer
required: without it, requests resolve through provider auth.
compact()/generateSummary()/generateBranchSummary() take a Models
parameter; explicit apiKey becomes optional. StreamFn is redefined
structurally (Models.streamSimple satisfies it), dropping the compat
type dependency from agent types.
Harness tests build per-file Models collections with fauxProvider()
and unique provider ids instead of mutating the global api-registry.
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.
Cristina Poncela Cubeiro