@jmtrin/opencode-kevin 0.4.0 → 0.6.0

package/dist/plugin/replay.d.ts

@@ -0,0 +1,36 @@
+import type { ReplayTranscript } from "./replay-types.js";
+/**
+ * v0.5.0 (K5-019 / plan §5.8, D5-12) — replay harness.
+ *
+ * Hermetic, deterministic driver: runs a recorded transcript through the
+ * plugin's components against an in-memory database with a frozen clock,
+ * then reports the outcome distribution. This is an artifact, not a
+ * release gate.
+ *
+ * The components are instantiated directly — NOT through `KevinPlugin`,
+ * which resolves a home-directory database path and registers tools.
+ * `Date.now()` must never be called in this path; every clock need is
+ * satisfied by the event's `at` timestamp (retrieval is frozen by the
+ * `deterministic_retrieval` setting; the Archiver clock is injected).
+ */
+export interface ReplayResult {
+    readonly transcript: string;
+    readonly memoriesCreated: number;
+    readonly injections: {
+        total: number;
+        effective: number;
+        ineffective: number;
+        inconclusive: number;
+        unmeasured: number;
+    };
+    readonly precisionRate: number;
+    readonly coverageRate: number;
+    readonly tokensInjected: {
+        prePrompt: number;
+        compacting: number;
+    };
+    readonly blocked: Record<string, number>;
+}
+export declare function replay(transcript: ReplayTranscript, opts?: {
+    dbPath?: string;
+}): Promise<ReplayResult>;

package/dist/plugin/replay.js

@@ -0,0 +1,203 @@
+import { dirname, join } from "node:path";
+import { fileURLToPath } from "node:url";
+import { Archiver } from "./Archiver.js";
+import { CausalChain } from "./CausalChain.js";
+import { ContextInjector } from "./ContextInjector.js";
+import { InjectionLedger } from "./InjectionLedger.js";
+import { MemoryService } from "./MemoryService.js";
+import { Migrate } from "./Migrate.js";
+import { Reflector } from "./Reflector.js";
+import { Store } from "./Store.js";
+import { ToolCallObserver } from "./ToolCallObserver.js";
+import { Metrics } from "./metrics.js";
+const MIGRATIONS_DIR = join(dirname(fileURLToPath(import.meta.url)), "..", "migrations");
+export async function replay(transcript, opts) {
+    const store = new Store({ path: opts?.dbPath ?? ":memory:" });
+    await new Migrate(store, MIGRATIONS_DIR).run();
+    // Freeze retrieval BEFORE any query happens (K5-008): deterministic
+    // mode uses the DATE_NOW sentinel, ignores recency and never bumps.
+    store
+        .prepare("INSERT OR IGNORE INTO kevin_settings (key, value) VALUES ('deterministic_retrieval', '1')")
+        .run();
+    store
+        .prepare("UPDATE kevin_settings SET value = '1' WHERE key = 'deterministic_retrieval'")
+        .run();
+    // v0.6.0 (K6-022) — the release floor (0.6) blocks every
+    // single-observation memory the v0.5-era fixture exercises; the replay
+    // harness preserves the v0.5 push semantics it was recorded under, the
+    // same opt-out the legacy plugin harnesses apply.
+    store
+        .prepare("INSERT OR IGNORE INTO kevin_settings (key, value) VALUES ('injection_confidence_floor', '0')")
+        .run();
+    store
+        .prepare("UPDATE kevin_settings SET value = '0' WHERE key = 'injection_confidence_floor'")
+        .run();
+    const metrics = new Metrics(store);
+    const memoryService = new MemoryService(store, metrics);
+    const observer = new ToolCallObserver(store, metrics);
+    // v0.3.0 fix (mirrors index.ts) — the Reflector stamps
+    // tool_calls.error_fingerprint with the stderr-based fingerprint the
+    // error memory uses; without it the recurrence queries in settle and
+    // the feedback loop never match (tool_calls.fingerprint is keyed on
+    // `${tool}|${args}|${success}`, memories on a hash of stderr/stdout).
+    const linkErrorStmt = store.prepare("UPDATE tool_calls SET error_fingerprint = ? WHERE id = ?");
+    const reflector = new Reflector(memoryService, {
+        throttleMs: 0,
+        onLinkError: (callID, fp) => {
+            linkErrorStmt.run(fp, callID);
+        },
+    }, metrics);
+    const ledger = new InjectionLedger(store, metrics);
+    const injector = new ContextInjector(memoryService, metrics, ledger);
+    const causalChain = new CausalChain(store, memoryService, metrics);
+    // The Archiver reads the clock through this accessor, which is updated
+    // to the current event's `at` before each event is processed.
+    let currentAt = new Date(transcript.events[0]?.at ?? Date.now());
+    const archiver = new Archiver(store, memoryService, metrics, () => currentAt);
+    const lastUserQueryBySession = new Map();
+    const callCache = new Map();
+    function resolveQuery(sessionId) {
+        return lastUserQueryBySession.get(sessionId) ?? "";
+    }
+    for (const event of transcript.events) {
+        currentAt = new Date(event.at);
+        await handleEvent(event);
+    }
+    metrics.flush();
+    const outcomeRows = store
+        .prepare("SELECT outcome, COUNT(*) AS n FROM kevin_injections GROUP BY outcome")
+        .all();
+    const counts = {
+        unmeasured: 0,
+        effective: 0,
+        ineffective: 0,
+        inconclusive: 0,
+    };
+    for (const r of outcomeRows) {
+        counts[r.outcome] = r.n;
+    }
+    const total = outcomeRows.reduce((acc, r) => acc + r.n, 0);
+    const tokenValue = (key) => {
+        const row = store
+            .prepare("SELECT value FROM kevin_metrics WHERE key = ?")
+            .get(key);
+        return row?.value ?? 0;
+    };
+    return {
+        transcript: transcript.name,
+        memoriesCreated: totalMemories(store),
+        injections: {
+            total,
+            effective: counts.effective,
+            ineffective: counts.ineffective,
+            inconclusive: counts.inconclusive,
+            unmeasured: counts.unmeasured,
+        },
+        precisionRate: metrics.precisionRate(),
+        coverageRate: metrics.coverageRate(),
+        tokensInjected: {
+            prePrompt: tokenValue("tokens_injected_pre_prompt"),
+            compacting: tokenValue("tokens_injected_compacting"),
+        },
+        blocked: metrics.blockedSnapshot(),
+    };
+    async function handleEvent(event) {
+        switch (event.kind) {
+            case "session.created":
+                // Mirror of the live wiring (index.ts): a fresh session must
+                // not inherit the previous session's injection seen-set
+                // (plan §5.1 rule 3). Without this, a multi-session
+                // transcript would block every memory in session 2 with
+                // `seen_this_session` while the live plugin would admit it.
+                injector.onSessionCreated(event.sessionId);
+                return;
+            case "chat.message": {
+                const derived = injector.deriveQuery([
+                    { role: "user", content: event.text },
+                ]);
+                if (derived.length > 0) {
+                    lastUserQueryBySession.set(event.sessionId, derived);
+                }
+                return;
+            }
+            case "tool.before": {
+                callCache.set(event.callId, {
+                    tool: event.tool,
+                    args: event.args,
+                });
+                observer.onBefore({
+                    tool: event.tool,
+                    args: event.args,
+                    sessionId: event.sessionId,
+                    callID: event.callId,
+                    projectId: null,
+                }, {});
+                return;
+            }
+            case "tool.after": {
+                const cached = callCache.get(event.callId);
+                const tool = cached?.tool ?? "unknown";
+                const args = cached?.args ?? {};
+                const stderr = event.stderr ?? "";
+                const stdout = event.stdout ?? "";
+                observer.onAfter({
+                    tool,
+                    args,
+                    sessionId: event.sessionId,
+                    callID: event.callId,
+                    projectId: null,
+                }, { success: event.success, stdout, stderr, exitCode: event.exitCode });
+                if (!event.success) {
+                    await reflector.invoke({
+                        toolName: tool,
+                        argsSummary: observer.summarizeArgs(args),
+                        stderr,
+                        stdout,
+                        exitCode: event.exitCode,
+                        errorType: observer.inferErrorType(stderr, stdout, event.exitCode),
+                        sessionId: event.sessionId,
+                        callID: event.callId,
+                        projectId: null,
+                    });
+                }
+                else {
+                    causalChain.onSuccess(tool, args, null, event.sessionId);
+                }
+                return;
+            }
+            case "system.transform": {
+                const query = resolveQuery(event.sessionId);
+                if (!query)
+                    return;
+                const output = { system: [] };
+                injector.onSystemTransform({
+                    sessionID: event.sessionId,
+                    messages: [{ role: "user", content: query }],
+                }, output);
+                return;
+            }
+            case "compacting": {
+                const query = resolveQuery(event.sessionId);
+                if (!query)
+                    return;
+                const output = { context: [] };
+                injector.onCompacting({
+                    sessionID: event.sessionId,
+                    messages: [{ role: "user", content: query }],
+                }, output);
+                return;
+            }
+            case "session.idle": {
+                ledger.settle(event.sessionId);
+                archiver.run();
+                metrics.flush();
+                return;
+            }
+        }
+    }
+}
+function totalMemories(store) {
+    const row = store.prepare("SELECT COUNT(*) AS n FROM memories").get();
+    return row.n ?? 0;
+}
+//# sourceMappingURL=replay.js.map

package/dist/plugin/replay.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"replay.js","sourceRoot":"","sources":["../../plugin/replay.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,MAAM,WAAW,CAAC;AAC1C,OAAO,EAAE,aAAa,EAAE,MAAM,UAAU,CAAC;AACzC,OAAO,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AACzC,OAAO,EAAE,WAAW,EAAE,MAAM,kBAAkB,CAAC;AAC/C,OAAO,EAAE,eAAe,EAAE,MAAM,sBAAsB,CAAC;AACvD,OAAO,EAAE,eAAe,EAAE,MAAM,sBAAsB,CAAC;AACvD,OAAO,EAAE,aAAa,EAAE,MAAM,oBAAoB,CAAC;AACnD,OAAO,EAAE,OAAO,EAAE,MAAM,cAAc,CAAC;AACvC,OAAO,EAAE,SAAS,EAAE,MAAM,gBAAgB,CAAC;AAC3C,OAAO,EAAE,KAAK,EAAE,MAAM,YAAY,CAAC;AACnC,OAAO,EAAE,gBAAgB,EAAE,MAAM,uBAAuB,CAAC;AACzD,OAAO,EAAE,OAAO,EAAE,MAAM,cAAc,CAAC;AAkCvC,MAAM,cAAc,GAAG,IAAI,CAC1B,OAAO,CAAC,aAAa,CAAC,MAAM,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EACvC,IAAI,EACJ,YAAY,CACZ,CAAC;AAEF,MAAM,CAAC,KAAK,UAAU,MAAM,CAC3B,UAA4B,EAC5B,IAA0B;IAE1B,MAAM,KAAK,GAAG,IAAI,KAAK,CAAC,EAAE,IAAI,EAAE,IAAI,EAAE,MAAM,IAAI,UAAU,EAAE,CAAC,CAAC;IAC9D,MAAM,IAAI,OAAO,CAAC,KAAK,EAAE,cAAc,CAAC,CAAC,GAAG,EAAE,CAAC;IAE/C,oEAAoE;IACpE,oEAAoE;IACpE,KAAK;SACH,OAAO,CACP,2FAA2F,CAC3F;SACA,GAAG,EAAE,CAAC;IACR,KAAK;SACH,OAAO,CACP,6EAA6E,CAC7E;SACA,GAAG,EAAE,CAAC;IACR,yDAAyD;IACzD,uEAAuE;IACvE,uEAAuE;IACvE,kDAAkD;IAClD,KAAK;SACH,OAAO,CACP,8FAA8F,CAC9F;SACA,GAAG,EAAE,CAAC;IACR,KAAK;SACH,OAAO,CACP,gFAAgF,CAChF;SACA,GAAG,EAAE,CAAC;IAER,MAAM,OAAO,GAAG,IAAI,OAAO,CAAC,KAAK,CAAC,CAAC;IACnC,MAAM,aAAa,GAAG,IAAI,aAAa,CAAC,KAAK,EAAE,OAAO,CAAC,CAAC;IACxD,MAAM,QAAQ,GAAG,IAAI,gBAAgB,CAAC,KAAK,EAAE,OAAO,CAAC,CAAC;IACtD,uDAAuD;IACvD,qEAAqE;IACrE,qEAAqE;IACrE,oEAAoE;IACpE,sEAAsE;IACtE,MAAM,aAAa,GAAG,KAAK,CAAC,OAAO,CAClC,0DAA0D,CAC1D,CAAC;IACF,MAAM,SAAS,GAAG,IAAI,SAAS,CAC9B,aAAa,EACb;QACC,UAAU,EAAE,CAAC;QACb,WAAW,EAAE,CAAC,MAAM,EAAE,EAAE,EAAE,EAAE;YAC3B,aAAa,CAAC,GAAG,CAAC,EAAE,EAAE,MAAM,CAAC,CAAC;QAC/B,CAAC;KACD,EACD,OAAO,CACP,CAAC;IACF,MAAM,MAAM,GAAG,IAAI,eAAe,CAAC,KAAK,EAAE,OAAO,CAAC,CAAC;IACnD,MAAM,QAAQ,GAAG,IAAI,eAAe,CAAC,aAAa,EAAE,OAAO,EAAE,MAAM,CAAC,CAAC;IACrE,MAAM,WAAW,GAAG,IAAI,WAAW,CAAC,KAAK,EAAE,aAAa,EAAE,OAAO,CAAC,CAAC;IAEnE,uEAAuE;IACvE,8DAA8D;IAC9D,IAAI,SAAS,GAAG,IAAI,IAAI,CAAC,UAAU,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,IAAI,IAAI,CAAC,GAAG,EAAE,CAAC,CAAC;IACjE,MAAM,QAAQ,GAAG,IAAI,QAAQ,CAAC,KAAK,EAAE,aAAa,EAAE,OAAO,EAAE,GAAG,EAAE,CAAC,SAAS,CAAC,CAAC;IAE9E,MAAM,sBAAsB,GAAG,IAAI,GAAG,EAAkB,CAAC;IACzD,MAAM,SAAS,GAAG,IAAI,GAAG,EAGtB,CAAC;IAEJ,SAAS,YAAY,CAAC,SAAiB;QACtC,OAAO,sBAAsB,CAAC,GAAG,CAAC,SAAS,CAAC,IAAI,EAAE,CAAC;IACpD,CAAC;IAED,KAAK,MAAM,KAAK,IAAI,UAAU,CAAC,MAAM,EAAE,CAAC;QACvC,SAAS,GAAG,IAAI,IAAI,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC;QAC/B,MAAM,WAAW,CAAC,KAAK,CAAC,CAAC;IAC1B,CAAC;IAED,OAAO,CAAC,KAAK,EAAE,CAAC;IAEhB,MAAM,WAAW,GAAG,KAAK;SACvB,OAAO,CACP,sEAAsE,CACtE;SACA,GAAG,EAAsC,CAAC;IAC5C,MAAM,MAAM,GAA2B;QACtC,UAAU,EAAE,CAAC;QACb,SAAS,EAAE,CAAC;QACZ,WAAW,EAAE,CAAC;QACd,YAAY,EAAE,CAAC;KACf,CAAC;IACF,KAAK,MAAM,CAAC,IAAI,WAAW,EAAE,CAAC;QAC7B,MAAM,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;IACzB,CAAC;IACD,MAAM,KAAK,GAAG,WAAW,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,CAAC,EAAE,EAAE,CAAC,GAAG,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC;IAE3D,MAAM,UAAU,GAAG,CAAC,GAAW,EAAU,EAAE;QAC1C,MAAM,GAAG,GAAG,KAAK;aACf,OAAO,CAAC,+CAA+C,CAAC;aACxD,GAAG,CAAC,GAAG,CAAkC,CAAC;QAC5C,OAAO,GAAG,EAAE,KAAK,IAAI,CAAC,CAAC;IACxB,CAAC,CAAC;IAEF,OAAO;QACN,UAAU,EAAE,UAAU,CAAC,IAAI;QAC3B,eAAe,EAAE,aAAa,CAAC,KAAK,CAAC;QACrC,UAAU,EAAE;YACX,KAAK;YACL,SAAS,EAAE,MAAM,CAAC,SAAS;YAC3B,WAAW,EAAE,MAAM,CAAC,WAAW;YAC/B,YAAY,EAAE,MAAM,CAAC,YAAY;YACjC,UAAU,EAAE,MAAM,CAAC,UAAU;SAC7B;QACD,aAAa,EAAE,OAAO,CAAC,aAAa,EAAE;QACtC,YAAY,EAAE,OAAO,CAAC,YAAY,EAAE;QACpC,cAAc,EAAE;YACf,SAAS,EAAE,UAAU,CAAC,4BAA4B,CAAC;YACnD,UAAU,EAAE,UAAU,CAAC,4BAA4B,CAAC;SACpD;QACD,OAAO,EAAE,OAAO,CAAC,eAAe,EAAE;KAClC,CAAC;IAEF,KAAK,UAAU,WAAW,CAAC,KAAkB;QAC5C,QAAQ,KAAK,CAAC,IAAI,EAAE,CAAC;YACpB,KAAK,iBAAiB;gBACrB,6DAA6D;gBAC7D,wDAAwD;gBACxD,oDAAoD;gBACpD,wDAAwD;gBACxD,4DAA4D;gBAC5D,QAAQ,CAAC,gBAAgB,CAAC,KAAK,CAAC,SAAS,CAAC,CAAC;gBAC3C,OAAO;YACR,KAAK,cAAc,CAAC,CAAC,CAAC;gBACrB,MAAM,OAAO,GAAG,QAAQ,CAAC,WAAW,CAAC;oBACpC,EAAE,IAAI,EAAE,MAAM,EAAE,OAAO,EAAE,KAAK,CAAC,IAAI,EAAE;iBACrC,CAAC,CAAC;gBACH,IAAI,OAAO,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;oBACxB,sBAAsB,CAAC,GAAG,CAAC,KAAK,CAAC,SAAS,EAAE,OAAO,CAAC,CAAC;gBACtD,CAAC;gBACD,OAAO;YACR,CAAC;YACD,KAAK,aAAa,CAAC,CAAC,CAAC;gBACpB,SAAS,CAAC,GAAG,CAAC,KAAK,CAAC,MAAM,EAAE;oBAC3B,IAAI,EAAE,KAAK,CAAC,IAAI;oBAChB,IAAI,EAAE,KAAK,CAAC,IAAI;iBAChB,CAAC,CAAC;gBACH,QAAQ,CAAC,QAAQ,CAChB;oBACC,IAAI,EAAE,KAAK,CAAC,IAAI;oBAChB,IAAI,EAAE,KAAK,CAAC,IAAI;oBAChB,SAAS,EAAE,KAAK,CAAC,SAAS;oBAC1B,MAAM,EAAE,KAAK,CAAC,MAAM;oBACpB,SAAS,EAAE,IAAI;iBACf,EACD,EAAE,CACF,CAAC;gBACF,OAAO;YACR,CAAC;YACD,KAAK,YAAY,CAAC,CAAC,CAAC;gBACnB,MAAM,MAAM,GAAG,SAAS,CAAC,GAAG,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC;gBAC3C,MAAM,IAAI,GAAG,MAAM,EAAE,IAAI,IAAI,SAAS,CAAC;gBACvC,MAAM,IAAI,GAAG,MAAM,EAAE,IAAI,IAAI,EAAE,CAAC;gBAChC,MAAM,MAAM,GAAG,KAAK,CAAC,MAAM,IAAI,EAAE,CAAC;gBAClC,MAAM,MAAM,GAAG,KAAK,CAAC,MAAM,IAAI,EAAE,CAAC;gBAClC,QAAQ,CAAC,OAAO,CACf;oBACC,IAAI;oBACJ,IAAI;oBACJ,SAAS,EAAE,KAAK,CAAC,SAAS;oBAC1B,MAAM,EAAE,KAAK,CAAC,MAAM;oBACpB,SAAS,EAAE,IAAI;iBACf,EACD,EAAE,OAAO,EAAE,KAAK,CAAC,OAAO,EAAE,MAAM,EAAE,MAAM,EAAE,QAAQ,EAAE,KAAK,CAAC,QAAQ,EAAE,CACpE,CAAC;gBACF,IAAI,CAAC,KAAK,CAAC,OAAO,EAAE,CAAC;oBACpB,MAAM,SAAS,CAAC,MAAM,CAAC;wBACtB,QAAQ,EAAE,IAAI;wBACd,WAAW,EAAE,QAAQ,CAAC,aAAa,CAAC,IAAI,CAAC;wBACzC,MAAM;wBACN,MAAM;wBACN,QAAQ,EAAE,KAAK,CAAC,QAAQ;wBACxB,SAAS,EAAE,QAAQ,CAAC,cAAc,CAAC,MAAM,EAAE,MAAM,EAAE,KAAK,CAAC,QAAQ,CAAC;wBAClE,SAAS,EAAE,KAAK,CAAC,SAAS;wBAC1B,MAAM,EAAE,KAAK,CAAC,MAAM;wBACpB,SAAS,EAAE,IAAI;qBACf,CAAC,CAAC;gBACJ,CAAC;qBAAM,CAAC;oBACP,WAAW,CAAC,SAAS,CAAC,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,KAAK,CAAC,SAAS,CAAC,CAAC;gBAC1D,CAAC;gBACD,OAAO;YACR,CAAC;YACD,KAAK,kBAAkB,CAAC,CAAC,CAAC;gBACzB,MAAM,KAAK,GAAG,YAAY,CAAC,KAAK,CAAC,SAAS,CAAC,CAAC;gBAC5C,IAAI,CAAC,KAAK;oBAAE,OAAO;gBACnB,MAAM,MAAM,GAAG,EAAE,MAAM,EAAE,EAAc,EAAE,CAAC;gBAC1C,QAAQ,CAAC,iBAAiB,CACzB;oBACC,SAAS,EAAE,KAAK,CAAC,SAAS;oBAC1B,QAAQ,EAAE,CAAC,EAAE,IAAI,EAAE,MAAM,EAAE,OAAO,EAAE,KAAK,EAAE,CAAC;iBAC5C,EACD,MAAM,CACN,CAAC;gBACF,OAAO;YACR,CAAC;YACD,KAAK,YAAY,CAAC,CAAC,CAAC;gBACnB,MAAM,KAAK,GAAG,YAAY,CAAC,KAAK,CAAC,SAAS,CAAC,CAAC;gBAC5C,IAAI,CAAC,KAAK;oBAAE,OAAO;gBACnB,MAAM,MAAM,GAAG,EAAE,OAAO,EAAE,EAAc,EAAE,CAAC;gBAC3C,QAAQ,CAAC,YAAY,CACpB;oBACC,SAAS,EAAE,KAAK,CAAC,SAAS;oBAC1B,QAAQ,EAAE,CAAC,EAAE,IAAI,EAAE,MAAM,EAAE,OAAO,EAAE,KAAK,EAAE,CAAC;iBAC5C,EACD,MAAM,CACN,CAAC;gBACF,OAAO;YACR,CAAC;YACD,KAAK,cAAc,CAAC,CAAC,CAAC;gBACrB,MAAM,CAAC,MAAM,CAAC,KAAK,CAAC,SAAS,CAAC,CAAC;gBAC/B,QAAQ,CAAC,GAAG,EAAE,CAAC;gBACf,OAAO,CAAC,KAAK,EAAE,CAAC;gBAChB,OAAO;YACR,CAAC;QACF,CAAC;IACF,CAAC;AACF,CAAC;AAED,SAAS,aAAa,CAAC,KAAY;IAClC,MAAM,GAAG,GAAG,KAAK,CAAC,OAAO,CAAC,oCAAoC,CAAC,CAAC,GAAG,EAElE,CAAC;IACF,OAAO,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC;AACnB,CAAC"}

package/migrations/006_v05_glassbox.sql

@@ -0,0 +1,118 @@
+-- ============================================================================
+-- 006_v05_glassbox.sql — v0.5.0 "Glass Box"
+--
+-- Honest measurement, human feedback, lifecycle completion.
+--
+-- Section 1: rebuild kevin_injections to admit a fourth outcome.
+-- Section 2: human feedback storage.
+-- Section 3: memory lifecycle columns.
+-- Section 4: metric seeds.
+-- Section 5: setting seeds.
+-- Section 6: schema_version.
+-- ============================================================================
+
+-- ---------------------------------------------------------------------------
+-- 1. kevin_injections: add 'inconclusive'.
+--
+-- SQLite cannot ALTER a CHECK constraint, so the table must be rebuilt.
+-- Migration 004 set this precedent. kevin_injections has no FTS5 triggers,
+-- so unlike 004 this is a straight four-step rebuild.
+--
+-- Existing rows with outcome='effective' are remapped to 'inconclusive'.
+-- This is not data loss: v0.4's 'effective' meant "the error did not recur",
+-- which is the exact definition of the new 'inconclusive' bucket. Rows that
+-- genuinely earned the new 'effective' will be re-settled naturally, and the
+-- post-apply hook re-derives the counters from the table.
+-- ---------------------------------------------------------------------------
+CREATE TABLE IF NOT EXISTS kevin_injections_new (
+  id          TEXT PRIMARY KEY,
+  memory_id   TEXT NOT NULL,
+  fingerprint TEXT NOT NULL,
+  session_id  TEXT NOT NULL,
+  hook        TEXT NOT NULL CHECK (hook IN ('pre_prompt','compacting')),
+  tokens      INTEGER NOT NULL,
+  injected_at TEXT NOT NULL DEFAULT (datetime('now')),
+  outcome     TEXT NOT NULL DEFAULT 'unmeasured'
+              CHECK (outcome IN ('unmeasured','effective','ineffective','inconclusive'))
+);
+
+INSERT INTO kevin_injections_new
+  (id, memory_id, fingerprint, session_id, hook, tokens, injected_at, outcome)
+SELECT
+  id, memory_id, fingerprint, session_id, hook, tokens, injected_at,
+  CASE WHEN outcome = 'effective' THEN 'inconclusive' ELSE outcome END
+FROM kevin_injections;
+
+DROP TABLE kevin_injections;
+ALTER TABLE kevin_injections_new RENAME TO kevin_injections;
+
+CREATE INDEX IF NOT EXISTS idx_injections_fp      ON kevin_injections(fingerprint);
+CREATE INDEX IF NOT EXISTS idx_injections_session ON kevin_injections(session_id);
+CREATE INDEX IF NOT EXISTS idx_injections_outcome ON kevin_injections(outcome);
+
+-- ---------------------------------------------------------------------------
+-- 2. Human feedback. Append-only audit trail; the hot path reads the
+--    denormalized counters on `memories` (section 3), never this table.
+-- ---------------------------------------------------------------------------
+CREATE TABLE IF NOT EXISTS memory_feedback (
+  id         TEXT PRIMARY KEY,
+  memory_id  TEXT NOT NULL,
+  verdict    TEXT NOT NULL CHECK (verdict IN ('useful','wrong','outdated','ignore')),
+  session_id TEXT,
+  note       TEXT,
+  created_at TEXT NOT NULL DEFAULT (datetime('now'))
+);
+
+CREATE INDEX IF NOT EXISTS idx_feedback_memory  ON memory_feedback(memory_id);
+CREATE INDEX IF NOT EXISTS idx_feedback_created ON memory_feedback(created_at);
+
+-- ---------------------------------------------------------------------------
+-- 3. Memory lifecycle and feedback columns.
+--
+--    feedback_positive / feedback_negative are SEPARATE from evidence_count
+--    and recurrence_count by design: human judgement is evidence about the
+--    memory, causal counters are evidence about the world. Mixing them was
+--    the confidence-poisoning defect closed in v0.4.0.
+--
+--    superseded_by has no REFERENCES clause on purpose. Store enables
+--    PRAGMA foreign_keys=ON, and a hard FK would block deletion of a memory
+--    that superseded another.
+-- ---------------------------------------------------------------------------
+ALTER TABLE memories ADD COLUMN feedback_positive INTEGER NOT NULL DEFAULT 0;
+ALTER TABLE memories ADD COLUMN feedback_negative INTEGER NOT NULL DEFAULT 0;
+ALTER TABLE memories ADD COLUMN ignored           INTEGER NOT NULL DEFAULT 0;
+ALTER TABLE memories ADD COLUMN superseded_by     TEXT;
+ALTER TABLE memories ADD COLUMN archived_at       TEXT;
+
+CREATE INDEX IF NOT EXISTS idx_memories_ignored  ON memories(ignored);
+CREATE INDEX IF NOT EXISTS idx_memories_archived ON memories(archived_at);
+
+-- ---------------------------------------------------------------------------
+-- 4. Metric seeds. Order matches the additions to METRIC_KEYS in metrics.ts.
+-- ---------------------------------------------------------------------------
+INSERT OR IGNORE INTO kevin_metrics (key, value) VALUES
+  ('injections_inconclusive',      0),
+  ('injections_blocked_seen',      0),
+  ('injections_blocked_weak',      0),
+  ('injections_blocked_recurrence',0),
+  ('injections_blocked_stale',     0),
+  ('injections_blocked_ignored',   0),
+  ('feedback_positive_total',      0),
+  ('feedback_negative_total',      0),
+  ('memories_archived',            0);
+
+-- ---------------------------------------------------------------------------
+-- 5. Setting seeds. Values are TEXT, always. Read them with an explicit
+--    string comparison or an explicit Number() parse — never `=== 1`.
+--    (That exact mistake kept cross_project_enabled unreachable for the
+--    whole of v0.3.0.)
+-- ---------------------------------------------------------------------------
+INSERT OR IGNORE INTO kevin_settings (key, value) VALUES
+  ('deterministic_retrieval',    '0'),
+  ('pre_prompt_budget_tokens', '900'),
+  ('archive_after_days',        '30');
+
+-- ---------------------------------------------------------------------------
+-- 6. Version marker.
+-- ---------------------------------------------------------------------------
+INSERT OR IGNORE INTO schema_version (version) VALUES ('006');

package/migrations/007_v06_pull.sql

@@ -0,0 +1,145 @@
+-- ============================================================================
+-- 007_v06_pull.sql — v0.6.0 "Pull"
+--
+-- Distribution: curated artifacts instead of a per-prompt token tax.
+--
+-- Section 1: curation_proposals — the persisted human decision record.
+-- Section 2: artifact_writes — the disk audit trail, including refusals.
+-- Section 3: memories curation + inferability columns.
+-- Section 4: metric seeds.
+-- Section 5: setting seeds.
+-- Section 6: conditional push-budget demotion.
+-- Section 7: schema_version.
+--
+-- Additive only. Every CHECK constraint introduced here is on a NEW table;
+-- no existing constraint is widened, so unlike migration 006 there is no
+-- table rebuild in this file.
+-- ============================================================================
+
+-- ---------------------------------------------------------------------------
+-- 1. curation_proposals — one row per proposed artifact change.
+--
+--    This table is the "explicit human decision between generation and
+--    application" of Principle 22. `kevin_propose` writes 'pending' rows and
+--    nothing else; `kevin_approve` moves them to 'approved' → 'applied' or to
+--    'rejected'. Rows are never deleted: rejection history is the evidence
+--    base for roadmap kill criterion K4 ("proposals rejected more often than
+--    approved"), which is uncheckable if rejections are discarded.
+--
+--    memory_id has no REFERENCES clause. Store sets PRAGMA foreign_keys = ON,
+--    and a hard FK would block deleting a memory that a historical proposal
+--    once mentioned — the same reasoning as superseded_by in migration 006.
+-- ---------------------------------------------------------------------------
+CREATE TABLE IF NOT EXISTS curation_proposals (
+  id            TEXT PRIMARY KEY,
+  project_id    TEXT NOT NULL,
+  memory_id     TEXT NOT NULL,
+  kind          TEXT NOT NULL CHECK (kind IN ('agents_md','skill','reference')),
+  target_path   TEXT NOT NULL,
+  proposed_text TEXT NOT NULL,
+  diff          TEXT NOT NULL,
+  status        TEXT NOT NULL DEFAULT 'pending'
+                CHECK (status IN ('pending','approved','rejected','applied','superseded')),
+  created_at    TEXT NOT NULL DEFAULT (datetime('now')),
+  decided_at    TEXT,
+  applied_at    TEXT
+);
+CREATE INDEX IF NOT EXISTS idx_proposals_status  ON curation_proposals(status);
+CREATE INDEX IF NOT EXISTS idx_proposals_memory  ON curation_proposals(memory_id);
+CREATE INDEX IF NOT EXISTS idx_proposals_project ON curation_proposals(project_id);
+
+-- ---------------------------------------------------------------------------
+-- 2. artifact_writes — the append-only audit trail for every disk operation.
+--
+--    A row is written for EVERY ArtifactWriter.apply() call, including
+--    outcome='noop' and outcome='refused'. A refusal that leaves no trace is
+--    indistinguishable from a write that never happened, and the whole point
+--    of §5.1 rule 3 is that a refusal is a reportable event.
+--
+--    hash_before / hash_after are SHA-256 of the full file contents, not of
+--    the marker block. That is what makes rule 4 (bytes outside the markers
+--    are byte-identical) auditable after the fact rather than only at test
+--    time.
+-- ---------------------------------------------------------------------------
+CREATE TABLE IF NOT EXISTS artifact_writes (
+  id           TEXT PRIMARY KEY,
+  proposal_id  TEXT,
+  project_id   TEXT NOT NULL,
+  path         TEXT NOT NULL,
+  bytes_before INTEGER,
+  bytes_after  INTEGER,
+  hash_before  TEXT,
+  hash_after   TEXT,
+  outcome      TEXT NOT NULL CHECK (outcome IN ('written','noop','refused')),
+  reason       TEXT,
+  wrote_at     TEXT NOT NULL DEFAULT (datetime('now'))
+);
+CREATE INDEX IF NOT EXISTS idx_artifact_writes_path ON artifact_writes(path);
+
+-- ---------------------------------------------------------------------------
+-- 3. memories: curation state and inferability.
+--
+--    curated / curated_at record that a memory has already been published,
+--    so the Curator does not re-propose it on every session idle.
+--
+--    inferable is deliberately NULLABLE with no default. Three states are
+--    needed, not two: 1 = inferable (a self-describing diagnostic the model
+--    resolves for free), 0 = non-inferable (project truth), NULL = unknown
+--    (not yet classified, or classified as 'unknown'). The Curator predicate
+--    is `inferable != 1`, so NULL rows stay eligible — an unclassified memory
+--    must not be silently withheld from curation.
+-- ---------------------------------------------------------------------------
+ALTER TABLE memories ADD COLUMN curated    INTEGER NOT NULL DEFAULT 0;
+ALTER TABLE memories ADD COLUMN curated_at TEXT;
+ALTER TABLE memories ADD COLUMN inferable  INTEGER;
+
+CREATE INDEX IF NOT EXISTS idx_memories_curated   ON memories(curated);
+CREATE INDEX IF NOT EXISTS idx_memories_inferable ON memories(inferable);
+
+-- ---------------------------------------------------------------------------
+-- 4. Metric seeds. Order matches the additions to METRIC_KEYS in metrics.ts.
+--    injections_blocked_confidence is the sixth member of the v0.5 blocked
+--    family and MUST be counted like the other five (Principle 16).
+-- ---------------------------------------------------------------------------
+INSERT OR IGNORE INTO kevin_metrics (key, value) VALUES
+  ('proposals_created',             0),
+  ('proposals_approved',            0),
+  ('proposals_rejected',            0),
+  ('artifact_writes_total',         0),
+  ('artifact_writes_noop',          0),
+  ('injections_blocked_confidence', 0);
+
+-- ---------------------------------------------------------------------------
+-- 5. Setting seeds. Values are TEXT, always. Read them with an explicit
+--    string comparison or an explicit Number() parse — never `=== 1`.
+--    (That exact mistake kept cross_project_enabled unreachable for the
+--    whole of v0.3.0.)
+--
+--    skill_emission_enabled and reference_emission_enabled default to '0':
+--    the pull channels ship OFF and are opted into, because they depend on a
+--    v2 domain Kevin does not pin.
+-- ---------------------------------------------------------------------------
+INSERT OR IGNORE INTO kevin_settings (key, value) VALUES
+  ('curation_enabled',           '1'),
+  ('agents_md_path',             'AGENTS.md'),
+  ('skill_emission_enabled',     '0'),
+  ('reference_emission_enabled', '0'),
+  ('injection_confidence_floor', '0.6');
+
+-- ---------------------------------------------------------------------------
+-- 6. Push-budget demotion.
+--
+--    Lower the default push budget only where the user has not overridden it.
+--    A user who deliberately set 1200 (or 1500, or 200) keeps their value;
+--    only an installation still sitting on the v0.5 default of '900' is
+--    moved to '400'. An unconditional UPDATE here would silently discard a
+--    deliberate configuration choice, which is a worse defect than the token
+--    cost it would save.
+-- ---------------------------------------------------------------------------
+UPDATE kevin_settings SET value = '400'
+ WHERE key = 'pre_prompt_budget_tokens' AND value = '900';
+
+-- ---------------------------------------------------------------------------
+-- 7. Version marker.
+-- ---------------------------------------------------------------------------
+INSERT OR IGNORE INTO schema_version (version) VALUES ('007');

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@jmtrin/opencode-kevin",
-	"version": "0.4.0",
+	"version": "0.6.0",
 	"description": "Kevin — Observa y aprende: capa de aprendizaje para OpenCode",
 	"type": "module",
 	"main": "dist/plugin/index.js",
@@ -21,7 +21,8 @@
 		"typecheck": "tsc --noEmit",
 		"lint": "biome check .",
 		"format": "biome format --write .",
-		"verify": "node --import tsx scripts/verify-install.ts"
+		"verify": "node --import tsx scripts/verify-install.ts",
+		"replay": "node --import tsx scripts/replay.ts"
 	},
 	"dependencies": {
 		"@opencode-ai/plugin": "^1.17.6",