refactor(acr): ACR Store 1 as a disposable memo, not a durable side-store (#305)

The shipped Store 1 persisted repo-file descriptors in a new SqliteObservationStore
— a second, unsynchronized SQLite writer on the event-log DB (no WAL/busy_timeout,
one shared Connection across coroutines) holding a fact the log already carries. That
violates "the event log is the only source of truth" (inv #1/#8): an Observation is
just FileWrittenEvent.path + postImageHash + describe(bytes).render(), and render is a
pure function of content-addressed CAS bytes.

Replace the whole durable apparatus with an in-process memo (descriptorMemo, keyed on
repoRoot+path+contentHash) on the orchestrator, next to the existing artifactContentCache.
Rebuilt from the log on restart, disposable, no external store. Preserves the perf win
(skip CAS read + regex on repeat describes) with none of the concurrency/lock risk.

Deletes: ObservationStore/InMemoryObservationStore/SqliteObservationStore (+ tests),
InfrastructureModule.createObservationStore, the FileReadTool priming path (which never
hit anyway — it keyed on a workspace-relative path while every consumer looks up the
absolute FileWrittenEvent.path), the FileReadConfig/SessionOrchestrator plumbing, and
five orphaned build.gradle deps.

Store 2 (soft-confidence hints) unchanged — it was correct.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HMbPmZZjcXhR2crU82zZ8S
This commit is contained in:
2026-07-21 18:42:47 +04:00
parent 8806de1628
commit 12775d56da
16 changed files with 9 additions and 267 deletions
@@ -7,7 +7,6 @@ import com.correx.core.inference.Tokenizer
import com.correx.core.journal.DefaultDecisionJournalRepository
import com.correx.core.artifacts.kind.ArtifactKindRegistry
import com.correx.core.artifactstore.ArtifactStore
import com.correx.core.context.observation.ObservationStore
import com.correx.core.events.events.ApprovalDecisionResolvedEvent
import com.correx.core.events.events.ArtifactContentStoredEvent
import com.correx.core.events.events.ClarificationAnswer
@@ -112,8 +111,7 @@ class DefaultSessionOrchestrator(
readyTaskCounter: ReadyTaskCounter? = null,
taskClaimCoordinator: TaskClaimCoordinator? = null,
tuning: OrchestrationTuning = OrchestrationTuning(),
observationStore: ObservationStore? = null,
) : SessionOrchestrator(repositories, engines, artifactStore, decisionJournalRepository, artifactKindRegistry = artifactKindRegistry, repoKnowledgeRetriever = repoKnowledgeRetriever, readyTaskCounter = readyTaskCounter, taskClaimCoordinator = taskClaimCoordinator, tuning = tuning, observationStore = observationStore), ApprovalGateway {
) : SessionOrchestrator(repositories, engines, artifactStore, decisionJournalRepository, artifactKindRegistry = artifactKindRegistry, repoKnowledgeRetriever = repoKnowledgeRetriever, readyTaskCounter = readyTaskCounter, taskClaimCoordinator = taskClaimCoordinator, tuning = tuning), ApprovalGateway {
override val tokenizer: Tokenizer? = tokenizer
override val cancellations: ConcurrentHashMap<SessionId, AtomicBoolean> =
ConcurrentHashMap<SessionId, AtomicBoolean>()
@@ -8,7 +8,6 @@ import com.correx.core.artifacts.kind.KindInference
import com.correx.core.artifactstore.ArtifactStore
import com.correx.core.context.builder.ContextPackBuilder
import com.correx.core.context.model.ContextPack
import com.correx.core.context.observation.ObservationStore
import com.correx.core.events.events.ClarificationAnswer
import com.correx.core.events.events.ClarificationAnsweredEvent
import com.correx.core.events.events.ClarificationRequestedEvent
@@ -203,7 +202,6 @@ abstract class SessionOrchestrator(
internal val readyTaskCounter: ReadyTaskCounter? = null,
internal val taskClaimCoordinator: TaskClaimCoordinator? = null,
internal val tuning: OrchestrationTuning = OrchestrationTuning(),
internal val observationStore: ObservationStore? = null,
) {
internal val log = LoggerFactory.getLogger(this::class.java)
internal val eventStore: EventStore = repositories.eventStore
@@ -250,6 +248,11 @@ abstract class SessionOrchestrator(
*/
internal val artifactContentCache: ConcurrentHashMap<String, String> = ConcurrentHashMap()
// ACR Store 1: in-process memo of describe().render(), keyed on (repoRoot, path, contentHash).
// Disposable — the source of truth is FileWrittenEvent + CAS; this only skips recomputing a pure
// function of content-addressed bytes. Empty string = a computed "no descriptor" (negative cache).
internal val descriptorMemo: ConcurrentHashMap<String, String> = ConcurrentHashMap()
/** Drops a terminated session's cached artifact contents (the heaviest per-session state — full
* file/JSON payloads). Safe: rehydrateArtifactContentCache rebuilds it from durable events if the
* session is ever resumed. Called on WorkflowCompleted/WorkflowFailed. */