agentera 3.0.0-dev.7 → 3.0.0-dev.74

package/dist/state/startupProjection.js

@@ -0,0 +1,499 @@
+import { createHash } from "node:crypto";
+import fs from "node:fs";
+import { StringDecoder } from "node:string_decoder";
+import path from "node:path";
+import YAML from "yaml";
+import { truncateCodePoints } from "../core/text.js";
+import { resolveSourceRoot } from "../core/sourceRoot.js";
+import { fullEntityUpgradePreviewCommand } from "../upgrade/upgradeCommands.js";
+import { canonicalRecordJson, discoverNumberedArchives, numberedArchiveContract, stateCurrentProjectionPath, } from "./archiveDiscovery.js";
+import { legacyIdentity } from "./legacyIdentity.js";
+import { classifyStateRows } from "./listClassification.js";
+const READ_CHUNK_BYTES = 64 * 1024;
+const MAX_CAPTURED_VALUE_CHARS = 512;
+const MAX_LINE_BUFFER_BYTES = 16 * 1024;
+export const STARTUP_ARRAY_LIMIT = 20;
+const ROOT_ITEM = /^  -(?:\s+(.*))?$/;
+const ROOT_FIELD = /^    ([A-Za-z0-9_-]+):(?:\s*(.*))?$/;
+const NESTED_FIELD = /^\s{6,}([A-Za-z0-9_-]+):(?:\s*(.*))?$/;
+const CAPTURED_FIELDS = new Set([
+    "number",
+    "date",
+    "timestamp",
+    "status",
+    "type",
+    "phase",
+    "summary",
+    "what",
+    "question",
+    "choice",
+    "trajectory",
+    "name",
+    "title",
+    "review_needed",
+    "state",
+    "source",
+    "review_date",
+    "review_by",
+    "review_due",
+]);
+const BLOCK_SECTIONS = new Set(["grades", "satisfaction"]);
+const PRESERVED_STARTUP_KEYS = new Set([
+    "number",
+    "entry_number",
+    "stable_id",
+    "artifact_id",
+    "status",
+    "mode",
+    "capability",
+    "command",
+    "get",
+    "list",
+    "fallback_command",
+    "next_cursor",
+    "field",
+]);
+/** Bound optional capability startup detail without touching identity or routes. */
+export function boundStartupValue(value, key = "", depth = 0) {
+    if (typeof value === "string")
+        return PRESERVED_STARTUP_KEYS.has(key) ? value : truncateCodePoints(value, 200, "\u2026");
+    if (Array.isArray(value)) {
+        const bounded = value.slice(0, STARTUP_ARRAY_LIMIT).map((item) => boundStartupValue(item, key, depth + 1));
+        return bounded;
+    }
+    if (!isMapping(value) || depth > 8)
+        return value;
+    const result = {};
+    for (const [childKey, child] of Object.entries(value)) {
+        result[childKey] = boundStartupValue(child, childKey, depth + 1);
+    }
+    return result;
+}
+function isMapping(value) {
+    return value !== null && typeof value === "object" && !Array.isArray(value);
+}
+function capturedString(value) {
+    return truncateCodePoints(value, MAX_CAPTURED_VALUE_CHARS, "\u2026");
+}
+function hashValue(value) {
+    return createHash("sha256").update(canonicalRecordJson(value), "utf8").digest("hex");
+}
+function scalar(value) {
+    if (value === undefined)
+        return undefined;
+    const text = value.trim();
+    if (!text || text === "|" || text === ">" || text === "|-" || text === ">-")
+        return undefined;
+    const bounded = capturedString(text);
+    try {
+        const parsed = YAML.parse(bounded);
+        if (parsed === null || typeof parsed === "string" || typeof parsed === "number" || typeof parsed === "boolean") {
+            return parsed;
+        }
+    }
+    catch {
+        // A bounded prefix can be an incomplete quoted scalar; identity metadata is
+        // still useful as a plain string in that case.
+    }
+    return bounded.replace(/^['"]|['"]$/g, "");
+}
+function setCaptured(fields, key, value) {
+    if (key === "review_needed") {
+        const parsed = scalar(value);
+        if (parsed !== undefined)
+            fields.review_needed = parsed;
+        return;
+    }
+    if (!CAPTURED_FIELDS.has(key))
+        return;
+    const parsed = scalar(value);
+    if (parsed !== undefined)
+        fields[key] = parsed;
+}
+function setNested(fields, section, key, value) {
+    if (section === "grades") {
+        const parsed = scalar(value);
+        if (parsed === undefined)
+            return;
+        const grades = isMapping(fields.grades) ? fields.grades : {};
+        grades[key] = parsed;
+        fields.grades = grades;
+        return;
+    }
+    if (section !== "satisfaction" || !["review_needed", "state", "source", "review_date", "review_by", "review_due"].includes(key))
+        return;
+    const parsed = scalar(value);
+    if (parsed === undefined)
+        return;
+    const satisfaction = isMapping(fields.satisfaction) ? fields.satisfaction : {};
+    satisfaction[key] = parsed;
+    fields.satisfaction = satisfaction;
+}
+function finalizeEntry(entries, fields, artifactId, entryNumberField, filePath, origin) {
+    entries.push({
+        index: entries.length,
+        path: filePath,
+        fields,
+        identity: legacyIdentity(fields, artifactId, entryNumberField),
+        origin,
+        projectionHash: hashValue(fields),
+    });
+}
+/**
+ * Read only the bounded metadata needed by startup consumers. The scanner
+ * never constructs a full YAML record, so a large verification or transcript
+ * field cannot become a retained prime payload or heap-sized object graph.
+ */
+export function scanYamlCollection(filePath, collection, artifactId, entryNumberField, origin = collection === "archive" ? "summary" : "active") {
+    if (!fs.existsSync(filePath))
+        return { path: filePath, exists: false, collection_found: false, bytes_scanned: 0, entries: [] };
+    try {
+        const stat = fs.lstatSync(filePath);
+        if (stat.isSymbolicLink() || !stat.isFile()) {
+            return { path: filePath, exists: true, collection_found: false, bytes_scanned: 0, entries: [] };
+        }
+    }
+    catch {
+        return { path: filePath, exists: true, collection_found: false, bytes_scanned: 0, entries: [] };
+    }
+    let fd;
+    try {
+        fd = fs.openSync(filePath, "r");
+    }
+    catch {
+        return { path: filePath, exists: true, collection_found: false, bytes_scanned: 0, entries: [] };
+    }
+    const decoder = new StringDecoder("utf8");
+    const buffer = Buffer.allocUnsafe(READ_CHUNK_BYTES);
+    const entries = [];
+    let bytesScanned = 0;
+    let line = "";
+    let lineBytes = 0;
+    let droppingLongLine = false;
+    let inCollection = false;
+    let collectionFound = false;
+    let fields = null;
+    let section = null;
+    let blockField = null;
+    const finish = () => {
+        if (fields !== null)
+            finalizeEntry(entries, fields, artifactId, entryNumberField, filePath, origin);
+        fields = null;
+        section = null;
+        blockField = null;
+    };
+    const processLine = (rawLine) => {
+        const text = rawLine.replace(/\r$/, "");
+        if (!inCollection) {
+            if (text === `${collection}:`) {
+                inCollection = true;
+                collectionFound = true;
+            }
+            return;
+        }
+        if (text && !/^\s/.test(text)) {
+            finish();
+            inCollection = text === `${collection}:`;
+            if (inCollection)
+                collectionFound = true;
+            return;
+        }
+        const item = ROOT_ITEM.exec(text);
+        if (item) {
+            finish();
+            fields = {};
+            const inline = item[1] ?? "";
+            const pair = /^([A-Za-z0-9_-]+):(?:\s*(.*))?$/.exec(inline);
+            if (pair) {
+                setCaptured(fields, pair[1], pair[2]);
+                section = BLOCK_SECTIONS.has(pair[1]) && !pair[2] ? pair[1] : null;
+                blockField = pair[1] === "summary" && !pair[2] ? "summary" : null;
+            }
+            else if (inline) {
+                fields.summary = scalar(inline) ?? capturedString(inline);
+            }
+            return;
+        }
+        if (fields === null)
+            return;
+        const root = ROOT_FIELD.exec(text);
+        if (root) {
+            section = BLOCK_SECTIONS.has(root[1]) && !root[2] ? root[1] : null;
+            blockField = root[1] === "summary" && !root[2] ? "summary" : null;
+            if (root[1] === "verified" && (root[2] === undefined || scalar(root[2]) !== undefined))
+                fields.verified_present = true;
+            setCaptured(fields, root[1], root[2]);
+            return;
+        }
+        const nested = NESTED_FIELD.exec(text);
+        if (nested) {
+            if (section)
+                setNested(fields, section, nested[1], nested[2]);
+            if (blockField && fields[blockField] === undefined && nested[2]) {
+                fields[blockField] = capturedString(nested[2].trim());
+            }
+            return;
+        }
+        if (blockField && text.trim() && fields[blockField] === undefined) {
+            fields[blockField] = capturedString(text.trim());
+        }
+    };
+    try {
+        let bytesRead = fs.readSync(fd, buffer, 0, buffer.length, null);
+        while (bytesRead > 0) {
+            bytesScanned += bytesRead;
+            const chunk = decoder.write(buffer.subarray(0, bytesRead));
+            for (const char of chunk) {
+                if (char === "\n") {
+                    processLine(line);
+                    line = "";
+                    lineBytes = 0;
+                    droppingLongLine = false;
+                }
+                else if (!droppingLongLine) {
+                    line += char;
+                    lineBytes += Buffer.byteLength(char, "utf8");
+                    if (lineBytes > MAX_LINE_BUFFER_BYTES) {
+                        processLine(line);
+                        line = "";
+                        lineBytes = 0;
+                        droppingLongLine = true;
+                    }
+                }
+            }
+            bytesRead = fs.readSync(fd, buffer, 0, buffer.length, null);
+        }
+        const remainder = decoder.end();
+        for (const char of remainder) {
+            if (char === "\n") {
+                processLine(line);
+                line = "";
+                lineBytes = 0;
+                droppingLongLine = false;
+            }
+            else if (!droppingLongLine) {
+                line += char;
+                lineBytes += Buffer.byteLength(char, "utf8");
+            }
+        }
+        if (line)
+            processLine(line);
+        finish();
+    }
+    finally {
+        fs.closeSync(fd);
+    }
+    return { path: filePath, exists: true, collection_found: collectionFound, bytes_scanned: bytesScanned, entries };
+}
+function scanLegacyStartupArtifact(projectRoot, artifactId, sourceRoot = resolveSourceRoot()) {
+    const contract = numberedArchiveContract(artifactId, sourceRoot);
+    const currentPath = stateCurrentProjectionPath(projectRoot, artifactId, sourceRoot);
+    return {
+        contract,
+        active: scanYamlCollection(currentPath, contract.entryCollection, artifactId, contract.entryNumberField, "active"),
+        archive: scanYamlCollection(currentPath, "archive", artifactId, contract.entryNumberField, "summary"),
+    };
+}
+export const scanStartupArtifact = scanLegacyStartupArtifact;
+function archiveNumber(rejection, artifactId) {
+    if (!rejection.path.includes(path.join(`${path.sep}${artifactId}`, "")))
+        return null;
+    const match = /(?:^|[\\/])([1-9][0-9]*)\.yaml$/.exec(rejection.path);
+    return match ? Number(match[1]) : null;
+}
+function classificationRow(entry, artifactId) {
+    if ("path" in entry && "reason" in entry) {
+        const number = archiveNumber(entry, artifactId);
+        return {
+            source: "archive",
+            origin: "rejected_archive",
+            identity: number === null ? "unaddressable" : "canonical_number",
+            entryNumber: number,
+            representation: "unavailable",
+            projectionHash: hashValue({ path: entry.path, reason: entry.reason, message: entry.message }),
+            rejection: { reason: entry.reason },
+        };
+    }
+    if ("entryNumber" in entry) {
+        return {
+            source: "archive",
+            origin: "numbered_archive",
+            identity: "canonical_number",
+            entryNumber: entry.entryNumber,
+            representation: "summary",
+            projectionHash: entry.recordSha256,
+        };
+    }
+    return {
+        source: "current_projection",
+        origin: entry.origin,
+        identity: entry.identity.kind,
+        entryNumber: entry.identity.number,
+        // The bounded scanner retains a startup summary, not the full record. Keep
+        // it comparable for duplicate/conflict detection without retaining YAML.
+        representation: "full",
+        projectionHash: entry.projectionHash,
+    };
+}
+function classifyNumber(artifactId, number, current, archives, rejected) {
+    const rows = [
+        ...current.filter((entry) => entry.identity.number === number).map((entry) => classificationRow(entry, artifactId)),
+        ...archives.filter((entry) => entry.entryNumber === number).map((entry) => classificationRow(entry, artifactId)),
+        ...rejected.filter((entry) => archiveNumber(entry, artifactId) === number).map((entry) => classificationRow(entry, artifactId)),
+    ];
+    return classifyStateRows(rows);
+}
+function classifyUnaddressable(entry, artifactId) {
+    return classifyStateRows([classificationRow(entry, artifactId)]);
+}
+function entrySummary(artifactId, entry, classification) {
+    const number = "entryNumber" in entry ? entry.entryNumber : entry.identity.number;
+    const fields = "fields" in entry ? entry.fields : {};
+    const summary = {};
+    for (const key of ["summary", "status", "date", "timestamp", "trajectory", "type", "phase", "review_needed"]) {
+        if (key in fields) {
+            const value = fields[key];
+            summary[key] = typeof value === "string" ? truncateCodePoints(value, 160, "\u2026") : value;
+        }
+    }
+    return {
+        stable_id: number === null ? null : `${artifactId}:${number}`,
+        artifact_id: artifactId,
+        entry_number: number,
+        addressable: number !== null,
+        classification,
+        detail_availability: "entryNumber" in entry ? "full" : "summary",
+        source: "entryNumber" in entry ? "archive" : "current_projection",
+        provenance: {
+            source: "entryNumber" in entry ? "archive" : "current_projection",
+            origin: "entryNumber" in entry ? "numbered_archive" : entry.origin,
+            path: "entryNumber" in entry ? entry.path : entry.path,
+        },
+        ...(Object.keys(summary).length > 0 ? { summary } : {}),
+    };
+}
+function countsFor(artifactId, current, archives, rejected, returned) {
+    const currentByNumber = new Map();
+    const archiveByNumber = new Map();
+    let unaddressable = 0;
+    let ambiguous = 0;
+    for (const entry of current) {
+        if (entry.identity.kind === "ambiguous")
+            ambiguous += 1;
+        else if (entry.identity.number === null)
+            unaddressable += 1;
+        else
+            currentByNumber.set(entry.identity.number, (currentByNumber.get(entry.identity.number) ?? 0) + 1);
+    }
+    for (const entry of archives)
+        archiveByNumber.set(entry.entryNumber, (archiveByNumber.get(entry.entryNumber) ?? 0) + 1);
+    for (const item of rejected) {
+        if (!item.path.includes(path.join(`${path.sep}${artifactId}`, "")))
+            continue;
+        const number = archiveNumber(item, artifactId);
+        if (number === null)
+            unaddressable += 1;
+        else
+            archiveByNumber.set(number, (archiveByNumber.get(number) ?? 0) + 1);
+    }
+    const ids = new Set([...currentByNumber.keys(), ...archiveByNumber.keys()]);
+    let mirrored = 0;
+    let duplicate = 0;
+    let conflict = 0;
+    for (const number of ids) {
+        const classification = classifyNumber(artifactId, number, current, archives, rejected);
+        if (classification === "mirrored")
+            mirrored += 1;
+        if (classification === "duplicate")
+            duplicate += 1;
+        if (classification === "conflict")
+            conflict += 1;
+    }
+    const physical = current.length + archives.length + rejected.filter((item) => item.path.includes(path.join(`${path.sep}${artifactId}`, ""))).length;
+    return {
+        physical,
+        addressable: physical - unaddressable - ambiguous,
+        addressable_ids: ids.size,
+        unaddressable,
+        ambiguous,
+        mirrored,
+        duplicate,
+        conflict,
+        omitted: Math.max(0, physical - returned),
+    };
+}
+function sourceRows(current, archives, rejected, artifactId) {
+    const rows = [];
+    for (const entry of current) {
+        const number = entry.identity.number;
+        const classification = number === null
+            ? classifyUnaddressable(entry, artifactId)
+            : classifyNumber(artifactId, number, current, archives, rejected);
+        rows.push(entrySummary(artifactId, entry, classification));
+    }
+    for (const entry of archives) {
+        rows.push(entrySummary(artifactId, entry, classifyNumber(artifactId, entry.entryNumber, current, archives, rejected)));
+    }
+    for (const item of rejected.filter((candidate) => candidate.path.includes(path.join(`${path.sep}${artifactId}`, "")))) {
+        const number = archiveNumber(item, artifactId);
+        const classification = number === null
+            ? classifyUnaddressable(item, artifactId)
+            : classifyNumber(artifactId, number, current, archives, rejected);
+        rows.push({
+            stable_id: number === null ? null : `${artifactId}:${number}`,
+            artifact_id: artifactId,
+            entry_number: number,
+            addressable: number !== null,
+            classification,
+            detail_availability: "unavailable",
+            source: "archive",
+            provenance: { source: "archive", origin: "rejected_archive", path: item.path },
+            compatibility: "blocked",
+            rejection: item.reason,
+        });
+    }
+    return rows.sort((left, right) => Number(right.entry_number ?? 0) - Number(left.entry_number ?? 0)).slice(0, 2);
+}
+export function startupHistorySummary(projectRoot, artifactId, sourceRoot = resolveSourceRoot()) {
+    const scans = scanLegacyStartupArtifact(projectRoot, artifactId, sourceRoot);
+    const currentPath = scans.active.path;
+    const scan = {
+        ...scans.active,
+        bytes_scanned: scans.active.bytes_scanned + scans.archive.bytes_scanned,
+        entries: [...scans.active.entries, ...scans.archive.entries],
+    };
+    const discovery = discoverNumberedArchives(projectRoot, { sourceRoot, artifactId, retainRecords: false });
+    const rejected = discovery.rejected;
+    const entries = sourceRows(scan.entries, discovery.entries, rejected, artifactId);
+    const counts = countsFor(artifactId, scan.entries, discovery.entries, rejected, entries.length);
+    const retrieval = {
+        upgrade_preview: fullEntityUpgradePreviewCommand(path.resolve(projectRoot)),
+    };
+    const status = !scan.exists && discovery.entries.length === 0 && rejected.length === 0
+        ? "absent"
+        : rejected.length > 0 || (scan.exists && !scans.active.collection_found)
+            ? "degraded"
+            : "available";
+    return {
+        artifact: artifactId,
+        status,
+        counts,
+        entries,
+        source: {
+            current_projection: { path: currentPath, present: scan.exists, entries: scan.entries.length },
+            archive: { root: discovery.archiveRoot, validated_entries: discovery.entries.length, rejected_count: rejected.length },
+            input_scan: {
+                current_entries: scan.entries.length,
+                current_bytes: scan.bytes_scanned,
+                archive_files: discovery.entries.length + rejected.length,
+            },
+        },
+        retrieval,
+        omission: {
+            omitted: counts.omitted > 0,
+            omitted_count: counts.omitted,
+            omission_reason: counts.omitted > 0 ? "startup_detail_capacity" : "none",
+            retrieval,
+        },
+    };
+}