akm-cli 0.9.0-rc.1 → 0.9.0-rc.13

package/dist/commands/improve/extract-cli.js

@@ -2,30 +2,30 @@
 // License, v. 2.0. If a copy of the MPL was not distributed with this
 // file, You can obtain one at https://mozilla.org/MPL/2.0/.
 /**
- * CLI surface for `akm extract`.
+ * CLI surface for `akm proposal extract` (formerly the top-level `akm
+ * extract`, moved under the `proposal` group in the 0.9 CLI overhaul, S8 —
+ * see src/commands/proposal/proposal-cli.ts for its registration).
  *
  * Examples:
- *   akm extract --type claude-code --session-id <id>
- *   akm extract --type claude-code --since 24h
- *   akm extract --type opencode --since 7d --dry-run
- *   akm extract --auto                 # iterate all available harnesses
- *   akm extract --type claude-code --location /custom/path --session-id <id>
+ *   akm proposal extract --type claude-code --session-id <id>
+ *   akm proposal extract --type claude-code --since 24h
+ *   akm proposal extract --type opencode --since 7d --dry-run
+ *   akm proposal extract --auto                 # iterate all available harnesses
+ *   akm proposal extract --type claude-code --location /custom/path --session-id <id>
  *
  * Output is the AkmExtractResult JSON envelope (or an aggregated one when
  * `--auto` runs multiple harnesses).
  */
-import fs from "node:fs";
-import path from "node:path";
 import { getStringArg } from "../../cli/parse-args.js";
 import { defineJsonCommand, EXIT_CODES, output } from "../../cli/shared.js";
+import { loadConfig } from "../../core/config/config.js";
 import { UsageError } from "../../core/errors.js";
-import { getAvailableHarnesses, getWatchTargets } from "../../integrations/session-logs/index.js";
-import { akmExtract } from "./extract.js";
-import { akmExtractWatch } from "./extract-watch.js";
+import { getAvailableHarnesses } from "../../integrations/session-logs/index.js";
+import { akmExtract, resolveStandaloneExtractPlan } from "./extract.js";
 export const extractCommand = defineJsonCommand({
     meta: {
         name: "extract",
-        description: "Extract durable insights from native session files (claude-code, opencode) and queue them as proposals. Replaces the legacy session-checkpoint hook.",
+        description: "Extract durable insights from native session files (claude-code, opencode) and queue them as proposals.",
     },
     args: {
         type: {
@@ -61,16 +61,15 @@ export const extractCommand = defineJsonCommand({
         },
         "timeout-ms": {
             type: "string",
-            description: "Per-session LLM timeout in ms (default 60000).",
+            description: "Per-session LLM timeout in ms (default 600000).",
         },
-        watch: {
-            type: "boolean",
-            description: "Watch harness session-log directories and run extract on change (debounced). Stays alive until SIGINT/SIGTERM. Falls back to the */30 cron when off.",
-            default: false,
+        engine: {
+            type: "string",
+            description: "Named LLM engine for this invocation. Mutually exclusive with --strategy.",
         },
-        "debounce-ms": {
+        strategy: {
             type: "string",
-            description: "Debounce window in ms for --watch (default 2000).",
+            description: "Improve strategy supplying extract behavior and engine. Mutually exclusive with --engine.",
         },
     },
     async run({ args }) {
@@ -81,22 +80,16 @@ export const extractCommand = defineJsonCommand({
         const auto = args.auto === true;
         const dryRun = args["dry-run"] === true;
         const force = args.force === true;
+        const engine = getStringArg(args, "engine");
+        const strategy = getStringArg(args, "strategy");
         const timeoutMs = typeof args["timeout-ms"] === "string" && args["timeout-ms"] !== ""
             ? Number.parseInt(args["timeout-ms"], 10)
             : undefined;
         if (timeoutMs !== undefined && (!Number.isFinite(timeoutMs) || timeoutMs <= 0)) {
             throw new UsageError(`--timeout-ms must be a positive integer (got "${args["timeout-ms"]}").`, "INVALID_FLAG_VALUE");
         }
-        const watch = args.watch === true;
-        const debounceMs = typeof args["debounce-ms"] === "string" && args["debounce-ms"] !== ""
-            ? Number.parseInt(args["debounce-ms"], 10)
-            : 2000;
-        if (watch && (!Number.isFinite(debounceMs) || debounceMs <= 0)) {
-            throw new UsageError(`--debounce-ms must be a positive integer (got "${args["debounce-ms"]}").`, "INVALID_FLAG_VALUE");
-        }
-        if (watch) {
-            await runWatchMode({ debounceMs, dryRun, force, ...(since ? { since } : {}) });
-            return;
+        if (engine && strategy) {
+            throw new UsageError("--engine and --strategy are mutually exclusive. Pick one.", "INVALID_FLAG_VALUE");
         }
         if (auto && type) {
             throw new UsageError("--auto and --type are mutually exclusive. Pick one.", "INVALID_FLAG_VALUE");
@@ -104,14 +97,21 @@ export const extractCommand = defineJsonCommand({
         if (!auto && !type) {
             throw new UsageError("--type is required (or pass --auto to try every available harness).", "MISSING_REQUIRED_ARGUMENT");
         }
-        const commonOptions = {
+        const config = loadConfig();
+        const resolvedPlan = resolveStandaloneExtractPlan(config, {
+            ...(engine ? { engine } : {}),
+            ...(strategy ? { strategy } : {}),
+            ...(timeoutMs !== undefined ? { timeoutMs } : {}),
+        });
+        const commonOptions = Object.freeze({
             ...(sessionId ? { sessionId } : {}),
             ...(location ? { location } : {}),
             ...(since ? { since } : {}),
             dryRun,
             force,
-            ...(timeoutMs !== undefined ? { timeoutMs } : {}),
-        };
+            config,
+            resolvedPlan,
+        });
         if (auto) {
             const harnesses = getAvailableHarnesses();
             if (harnesses.length === 0) {
@@ -140,10 +140,10 @@ export const extractCommand = defineJsonCommand({
                 totalProposals,
                 results,
             });
-            // Signal failure to callers/cron when every harness failed. output()
-            // only renders; without this the */30 cron exits 0 on a total failure
+            // Signal failure to callers/schedulers when every harness failed. output()
+            // only renders; without this a scheduled run exits 0 on a total failure
             // and the breakage is invisible to exit-code monitoring. process.exitCode
-            // (not process.exit) lets stdout flush and the watcher/timers settle.
+            // (not process.exit) lets stdout flush before the process exits.
             if (!ok)
                 process.exitCode = EXIT_CODES.GENERAL;
             return;
@@ -154,94 +154,3 @@ export const extractCommand = defineJsonCommand({
             process.exitCode = EXIT_CODES.GENERAL;
     },
 });
-/**
- * A thin {@link WatchEventSource} over `fs.watch` for each configured root.
- * This adapter is the ONLY place a real `fs.watch` is created (the core stays
- * injectable + fully unit-tested); it is intentionally not unit-covered.
- *
- * `fs.watch(dir, { recursive: true })` is unreliable for recursive mode on some
- * Node/Bun/Linux combinations — that is acceptable here: missed events degrade
- * to the unchanged scheduled cron fallback, never worse than today.
- */
-function createFsWatchEventSource(roots) {
-    return {
-        subscribe(listener) {
-            const watchers = [];
-            for (const root of roots) {
-                try {
-                    const watcher = fs.watch(root, { recursive: true }, (_event, filename) => {
-                        if (!filename)
-                            return;
-                        const filePath = path.isAbsolute(filename.toString())
-                            ? filename.toString()
-                            : path.join(root, filename.toString());
-                        listener({ path: filePath });
-                    });
-                    watchers.push(watcher);
-                }
-                catch {
-                    // A root that can't be watched is skipped; the cron fallback covers it.
-                }
-            }
-            return () => {
-                for (const w of watchers) {
-                    try {
-                        w.close();
-                    }
-                    catch {
-                        // best-effort teardown
-                    }
-                }
-            };
-        },
-    };
-}
-/**
- * Run `akm extract --watch`: watch every available harness's session-log
- * roots and run extract (debounced, per-harness) on change. Stays alive until
- * SIGINT/SIGTERM, then stops cleanly. PROCESS-HYGIENE: stop() removes every
- * watcher + pending timer before the process exits.
- */
-async function runWatchMode(opts) {
-    const targets = getWatchTargets();
-    if (targets.length === 0) {
-        output("extract", {
-            schemaVersion: 1,
-            ok: false,
-            shape: "extract-watch-started",
-            warnings: ["no watchable harness session-log directories found on this machine"],
-            watching: [],
-        });
-        return;
-    }
-    const allRoots = targets.flatMap((t) => t.roots);
-    const eventSource = createFsWatchEventSource(allRoots);
-    const handle = akmExtractWatch({
-        roots: targets,
-        eventSource,
-        debounceMs: opts.debounceMs,
-        onTrigger: async (harnessName) => {
-            await akmExtract({
-                type: harnessName,
-                dryRun: opts.dryRun,
-                force: opts.force,
-                ...(opts.since ? { since: opts.since } : {}),
-            });
-        },
-    });
-    output("extract", {
-        schemaVersion: 1,
-        ok: true,
-        shape: "extract-watch-started",
-        debounceMs: opts.debounceMs,
-        watching: allRoots,
-    });
-    await new Promise((resolve) => {
-        const shutdown = () => {
-            handle.stop();
-            resolve();
-        };
-        process.once("SIGINT", shutdown);
-        process.once("SIGTERM", shutdown);
-    });
-}

package/dist/commands/improve/extract-prompt.js

@@ -15,24 +15,20 @@
  * the model emits get dropped before we parse.
  */
 import promptTemplate from "../../assets/prompts/extract-session.md" with { type: "text" };
+const EXTRACT_CANDIDATE_NAME_PATTERN = "^[a-z0-9](?:[a-z0-9-]*[a-z0-9])?(?:/[a-z0-9](?:[a-z0-9-]*[a-z0-9])?)?$";
+const EXTRACT_CANDIDATE_NAME_RE = new RegExp(EXTRACT_CANDIDATE_NAME_PATTERN);
 /**
  * JSON Schema for the structured extract output. Passed to `chatCompletion`
  * when the configured LLM connection has `supportsJsonSchema: true`.
  *
  * Shape:
  *   {
- *     "candidates": [{type, name, description, when_to_use?, body, confidence, evidence,
- *                     orderedActions?, outcomeData?}, ...],
+ *     "candidates": [{type, name, description, when_to_use?, body, confidence, evidence}, ...],
  *     "rationale_if_empty"?: string
  *   }
  *
  * `additionalProperties: false` at each level so any hallucinated keys are
  * dropped before parsing.
- *
- * `orderedActions` and `outcomeData` are additive fields for #615 procedural
- * compilation (WS-0 data-capture hook). Source transcripts are external logs
- * not guaranteed re-extractable; we capture the ordered-action sequence and
- * outcome now even though the detection/compilation feature is deferred to 0.10+.
  */
 export const EXTRACT_JSON_SCHEMA = {
     type: "object",
@@ -54,8 +50,8 @@ export const EXTRACT_JSON_SCHEMA = {
                     },
                     name: {
                         type: "string",
-                        description: "Kebab-case slug for the new asset.",
-                        pattern: "^[a-z0-9][a-z0-9-]*[a-z0-9]$",
+                        description: "Kebab-case slug, optionally under one stable scope/domain segment.",
+                        pattern: EXTRACT_CANDIDATE_NAME_PATTERN,
                     },
                     description: {
                         type: "string",
@@ -85,17 +81,6 @@ export const EXTRACT_JSON_SCHEMA = {
                         minLength: 5,
                         description: "One-line pointer to the moment in the session that supports this candidate.",
                     },
-                    orderedActions: {
-                        type: "array",
-                        description: "OPTIONAL. Ordered list of discrete actions taken during this episode, as brief imperative phrases (e.g. 'run deploy.sh', 'check VPN status', 'retry with --force'). Capture only when the candidate represents a recurring action sequence that could become a skill or workflow. Omit when the candidate is a standalone fact or observation.",
-                        items: { type: "string", minLength: 3 },
-                        maxItems: 20,
-                    },
-                    outcomeData: {
-                        type: "string",
-                        description: "OPTIONAL. One-sentence description of the outcome when the ordered action sequence completed (e.g. 'deploy succeeded after VPN reconnect', 'build failed with error X'). Required when orderedActions is present; omit otherwise.",
-                        maxLength: 400,
-                    },
                 },
             },
         },
@@ -219,7 +204,7 @@ export function parseExtractPayload(stdout) {
         const type = c.type;
         if (type !== "memory" && type !== "lesson" && type !== "knowledge")
             continue;
-        if (typeof c.name !== "string" || !/^[a-z0-9][a-z0-9-]*[a-z0-9]$/.test(c.name))
+        if (typeof c.name !== "string" || !EXTRACT_CANDIDATE_NAME_RE.test(c.name))
             continue;
         if (typeof c.description !== "string" || c.description.trim().length < 20)
             continue;
@@ -244,20 +229,6 @@ export function parseExtractPayload(stdout) {
         };
         if (typeof c.when_to_use === "string")
             candidate.when_to_use = c.when_to_use.trim();
-        // #615 WS-0: parse optional ordered-action + outcome fields.
-        // Defensive: silently drop malformed entries rather than rejecting the whole candidate.
-        if (Array.isArray(c.orderedActions) && c.orderedActions.length > 0) {
-            const actions = c.orderedActions
-                .filter((a) => typeof a === "string" && a.trim().length >= 3)
-                .map((a) => a.trim())
-                .slice(0, 20);
-            if (actions.length > 0) {
-                candidate.orderedActions = actions;
-                if (typeof c.outcomeData === "string" && c.outcomeData.trim().length > 0) {
-                    candidate.outcomeData = c.outcomeData.trim().slice(0, 400);
-                }
-            }
-        }
         candidates.push(candidate);
     }
     const result = { candidates };