feat: P3 effective configuration, memory root, and revision-scoped records

This commit is contained in:
2026-08-12 16:01:25 +02:00
parent beaba548c1
commit e23e526966
21 changed files with 1108 additions and 44 deletions
+219
View File
@@ -0,0 +1,219 @@
import { createHash } from "node:crypto";
import { normalize } from "node:path";
export interface CanonicalEffectiveConfig {
schemaVersion: 1;
dwh: CanonicalDwhConfig;
vector: CanonicalVectorConfig;
embedding: CanonicalEmbeddingConfig;
roots: CanonicalRootsConfig;
}
export interface CanonicalDwhConfig {
engine: "postgres";
database: string;
schema: string;
transport: "postgres_direct" | "rest_api" | "ssh_tunnel";
host?: string;
port?: number;
baseUrl?: string;
user?: string;
}
export interface CanonicalVectorConfig {
collection: string;
dimensions: number;
distance: string;
}
export interface CanonicalEmbeddingConfig {
model: string;
dimensions: number;
}
export interface CanonicalRootsConfig {
artifacts: string;
indexes: string;
}
function asRecord(value: unknown): Record<string, unknown> | undefined {
if (typeof value === "object" && value !== null && !Array.isArray(value)) {
return value as Record<string, unknown>;
}
return undefined;
}
function requireString(value: Record<string, unknown>, key: string): string {
const candidate = value[key];
if (typeof candidate !== "string" || candidate.length === 0) {
throw new TypeError(`effective config is missing ${key}`);
}
return candidate;
}
function optionalString(value: Record<string, unknown>, key: string): string | undefined {
const candidate = value[key];
if (candidate === undefined || candidate === null) return undefined;
if (typeof candidate !== "string") return undefined;
return candidate;
}
function optionalNumber(value: Record<string, unknown>, key: string): number | undefined {
const candidate = value[key];
if (candidate === undefined || candidate === null) return undefined;
if (typeof candidate !== "number" || Number.isNaN(candidate)) return undefined;
return candidate;
}
function requireNumber(value: Record<string, unknown>, key: string): number {
const candidate = value[key];
if (typeof candidate !== "number" || Number.isNaN(candidate)) {
throw new TypeError(`effective config is missing numeric ${key}`);
}
return candidate;
}
function normalizeRoot(value: string): string {
return normalize(value);
}
function dwhTransport(rendered: Record<string, unknown>): CanonicalDwhConfig["transport"] {
const dwh = asRecord(rendered.dwh);
const type = dwh ? requireString(dwh, "type") : undefined;
if (type === "postgres_direct") return "postgres_direct";
if (type === "thoth_rest") return "rest_api";
if (type === "ssh_tunnel") return "ssh_tunnel";
throw new TypeError(`effective config has unsupported dwh transport ${type}`);
}
function buildDwhConfig(rendered: Record<string, unknown>): CanonicalDwhConfig {
const transport = dwhTransport(rendered);
const databaseRecord = asRecord(rendered.database) ?? asRecord(asRecord(asRecord(rendered.dwh)?.connection)?.database);
if (!databaseRecord) {
throw new TypeError("effective config is missing database identity");
}
const engine = "postgres";
const database = requireString(databaseRecord, "database");
const schema = requireString(databaseRecord, "schema");
const dwh: Record<string, unknown> = { engine, database, schema, transport };
if (transport === "postgres_direct") {
const host = optionalString(databaseRecord, "host");
const port = optionalNumber(databaseRecord, "port");
const user = optionalString(databaseRecord, "user");
if (host !== undefined) dwh.host = host;
if (port !== undefined) dwh.port = port;
if (user !== undefined) dwh.user = user;
} else if (transport === "rest_api") {
const rest = asRecord(rendered.rest) ?? asRecord(asRecord(asRecord(rendered.dwh)?.endpoint));
const baseUrl = rest ? optionalString(rest, "base_url") : undefined;
if (baseUrl !== undefined) dwh.baseUrl = baseUrl;
}
return dwh as unknown as CanonicalDwhConfig;
}
function buildVectorConfig(rendered: Record<string, unknown>): CanonicalVectorConfig {
const resources = asRecord(rendered.resources);
const vector = resources ? asRecord(resources.vector) : undefined;
if (!vector) {
throw new TypeError("effective config is missing vector resources");
}
const collection = requireString(vector, "collection");
const semanticIndex = asRecord(rendered.semantic_index);
const vectorStore = semanticIndex ? asRecord(semanticIndex.vector_store) : undefined;
const dimensions = vectorStore
? requireNumber(vectorStore, "dimensions")
: requireNumber(vector, "dimensions");
const distance = vectorStore
? requireString(vectorStore, "distance")
: (optionalString(vector, "distance") ?? "cosine");
return { collection, dimensions, distance };
}
function buildEmbeddingConfig(rendered: Record<string, unknown>): CanonicalEmbeddingConfig {
const resources = asRecord(rendered.resources);
const embeddings = resources ? asRecord(resources.embeddings) : undefined;
if (!embeddings) {
throw new TypeError("effective config is missing embedding resources");
}
return {
model: requireString(embeddings, "model"),
dimensions: requireNumber(embeddings, "dimensions"),
};
}
function buildRootsConfig(rendered: Record<string, unknown>): CanonicalRootsConfig {
const roots = asRecord(rendered.roots) ?? asRecord(rendered.paths);
if (!roots) {
throw new TypeError("effective config is missing roots");
}
return {
artifacts: normalizeRoot(requireString(roots, "artifacts")),
indexes: normalizeRoot(requireString(roots, "indexes")),
};
}
/**
* Build the versioned, non-secret effective DWH/preprocessing configuration from a
* rendered runtime configuration object. The result contains only the fields that
* affect DWH generation identity; credentials, runtime identity, session storage,
* evidence, and service endpoints are excluded.
*/
export function buildCanonicalEffectiveConfig(renderedConfig: unknown): CanonicalEffectiveConfig {
const rendered = asRecord(renderedConfig);
if (!rendered) {
throw new TypeError("effective config requires a rendered configuration object");
}
return {
schemaVersion: 1,
dwh: buildDwhConfig(rendered),
vector: buildVectorConfig(rendered),
embedding: buildEmbeddingConfig(rendered),
roots: buildRootsConfig(rendered),
};
}
function sha256(value: string | Buffer): string {
return `sha256:${createHash("sha256").update(value).digest("hex")}`;
}
/**
* Serialize the canonical effective config to a deterministic JSON string with the
* fixed key order defined by the shared contract. No whitespace is included.
*/
export function canonicalEffectiveConfigJson(config: CanonicalEffectiveConfig): string {
const ordered: Record<string, unknown> = { schemaVersion: config.schemaVersion };
ordered.dwh = { ...config.dwh };
ordered.vector = { ...config.vector };
ordered.embedding = { ...config.embedding };
ordered.roots = { ...config.roots };
return JSON.stringify(ordered);
}
/**
* Return the stable logical config-source identity for a workspace revision.
* This is `workspace://<workspaceId>@v1:<sha256-hex-of-canonical-json>`.
*/
export function effectiveConfigIdentity(workspaceId: string, renderedConfig: unknown): string {
const canonical = buildCanonicalEffectiveConfig(renderedConfig);
const digest = createHash("sha256").update(canonicalEffectiveConfigJson(canonical)).digest("hex");
return `workspace://${workspaceId}@v1:${digest}`;
}
/**
* Return the config fingerprint: `sha256:` + the SHA-256 of the canonical effective
* config JSON bytes.
*/
export function configFingerprint(renderedConfig: unknown): string {
const canonical = buildCanonicalEffectiveConfig(renderedConfig);
return sha256(canonicalEffectiveConfigJson(canonical));
}
/**
* Return the input fingerprint: `sha256:` + the SHA-256 of the logical config-source
* identity string.
*/
export function inputFingerprint(workspaceId: string, renderedConfig: unknown): string {
return sha256(effectiveConfigIdentity(workspaceId, renderedConfig));
}
@@ -30,6 +30,9 @@ export interface WorkspaceOperationResult {
completedStages: string[];
counts?: Record<string, number>;
artifactIdentities?: Array<{ kind: string; digest: string }>;
effectiveConfigIdentity?: string;
configFingerprint?: string;
inputFingerprint?: string;
/** Suggested FK annotations YAML for the operator to write to --output (schema suggest-fks). */
suggestedFksYaml?: string;
warnings?: string[];
@@ -92,6 +95,9 @@ function baseResult(
descriptorBlob: runtime.descriptorBlob,
operation,
completedStages: [],
effectiveConfigIdentity: runtime.configLease.effectiveConfigIdentity,
configFingerprint: runtime.configLease.configFingerprint,
inputFingerprint: runtime.configLease.inputFingerprint,
...extra,
};
}
+45 -4
View File
@@ -18,6 +18,14 @@ import {
import { dirname, isAbsolute, join, relative, resolve } from "node:path";
import { readFile as readFileAsync } from "node:fs/promises";
import { parse, parseAllDocuments, stringify } from "yaml";
import {
buildCanonicalEffectiveConfig,
canonicalEffectiveConfigJson,
configFingerprint,
effectiveConfigIdentity,
inputFingerprint,
type CanonicalEffectiveConfig,
} from "./effective-config.js";
import { resolveRuntimeBindings, type RuntimeBindings } from "./bindings.js";
import { GitWorkspaceRepository } from "./git-repository.js";
import { WorkspaceRegistry } from "./registry.js";
@@ -63,6 +71,10 @@ export interface DeterministicRuntimeConfigLease extends RuntimeConfigLease {
catalogBlob: string;
configDigest: string;
bindingDigest: string;
effectiveConfig: CanonicalEffectiveConfig;
effectiveConfigIdentity: string;
configFingerprint: string;
inputFingerprint: string;
}
export class RuntimeConfigLeaseError extends Error {
@@ -84,6 +96,9 @@ interface PublishedRuntimeConfigManifest {
catalogBlob: string;
configDigest: string;
bindingDigest: string;
effectiveConfigIdentity: string;
configFingerprint: string;
inputFingerprint: string;
path: string;
file: {
dev: number;
@@ -230,6 +245,7 @@ function runtimePaths(dataRoot: string, workspaceId: string): RuntimePaths {
sessions: join(root, "sessions"),
artifacts: join(root, "artifacts"),
indexes: join(root, "indexes"),
memory: join(root, "memory"),
};
}
@@ -388,6 +404,9 @@ function decodePublishedRuntimeConfigManifest(value: unknown): PublishedRuntimeC
|| typeof manifest.catalogBlob !== "string"
|| typeof manifest.configDigest !== "string"
|| typeof manifest.bindingDigest !== "string"
|| typeof manifest.effectiveConfigIdentity !== "string"
|| typeof manifest.configFingerprint !== "string"
|| typeof manifest.inputFingerprint !== "string"
|| typeof manifest.path !== "string"
|| !file
|| typeof file.dev !== "number"
@@ -413,6 +432,13 @@ export async function publishDeterministicRuntimeConfigLease(options: {
}): Promise<DeterministicRuntimeConfigLease> {
const rendered = await renderActiveWorkspaceRuntime(options);
const publishedConfig = applyCollectionLifecycle(rendered.renderedConfig, "require_existing");
const renderedConfigObject = parse(rendered.renderedConfig) as Record<string, unknown>;
const effectiveConfig = buildCanonicalEffectiveConfig(renderedConfigObject);
const effectiveConfigIdentityValue = effectiveConfigIdentity(rendered.workspaceId, renderedConfigObject);
const configFingerprintValue = configFingerprint(renderedConfigObject);
const inputFingerprintValue = inputFingerprint(rendered.workspaceId, renderedConfigObject);
const identitySuffix = inputFingerprintValue.slice(7, 23);
const preprocessingRoot = ensureTrustedDirectory(join(
options.dataRoot,
"sessions",
@@ -421,8 +447,8 @@ export async function publishDeterministicRuntimeConfigLease(options: {
));
const configDirectory = ensureTrustedDirectory(join(preprocessingRoot, "runtime-config"));
const manifestDirectory = ensureTrustedDirectory(join(preprocessingRoot, "runtime-config-manifests"));
const path = join(configDirectory, `${rendered.workspaceRevision}.yaml`);
const manifestPath = join(manifestDirectory, `${rendered.workspaceRevision}.json`);
const path = join(configDirectory, `${rendered.workspaceRevision}-${identitySuffix}.yaml`);
const manifestPath = join(manifestDirectory, `${rendered.workspaceRevision}-${identitySuffix}.json`);
const configDigest = sha256(publishedConfig);
const verifyPublished = (): PublishedRuntimeConfigManifest | undefined => {
@@ -456,6 +482,9 @@ export async function publishDeterministicRuntimeConfigLease(options: {
|| manifest.catalogBlob !== rendered.catalogBlob
|| manifest.configDigest !== configDigest
|| manifest.bindingDigest !== rendered.bindingDigest
|| manifest.effectiveConfigIdentity !== effectiveConfigIdentityValue
|| manifest.configFingerprint !== configFingerprintValue
|| manifest.inputFingerprint !== inputFingerprintValue
|| manifest.path !== path
|| manifest.file.dev !== configStat!.dev
|| manifest.file.ino !== configStat!.ino
@@ -479,6 +508,10 @@ export async function publishDeterministicRuntimeConfigLease(options: {
catalogBlob: rendered.catalogBlob,
configDigest,
bindingDigest: rendered.bindingDigest,
effectiveConfig,
effectiveConfigIdentity: effectiveConfigIdentityValue,
configFingerprint: configFingerprintValue,
inputFingerprint: inputFingerprintValue,
release: () => undefined,
};
}
@@ -505,6 +538,9 @@ export async function publishDeterministicRuntimeConfigLease(options: {
catalogBlob: rendered.catalogBlob,
configDigest,
bindingDigest: rendered.bindingDigest,
effectiveConfigIdentity: effectiveConfigIdentityValue,
configFingerprint: configFingerprintValue,
inputFingerprint: inputFingerprintValue,
path,
file: {
dev: Number(publishedStat.dev),
@@ -518,7 +554,8 @@ export async function publishDeterministicRuntimeConfigLease(options: {
readStrictJson(manifestPath);
} catch (error) {
if ((error as NodeJS.ErrnoException).code === "ENOENT") {
writeAtomicFile(manifestPath, `${JSON.stringify(manifest)}\n`, 0o600);
writeAtomicFile(manifestPath, `${JSON.stringify(manifest)}
`, 0o600);
} else {
throw error;
}
@@ -533,6 +570,10 @@ export async function publishDeterministicRuntimeConfigLease(options: {
catalogBlob: rendered.catalogBlob,
configDigest,
bindingDigest: rendered.bindingDigest,
effectiveConfig,
effectiveConfigIdentity: effectiveConfigIdentityValue,
configFingerprint: configFingerprintValue,
inputFingerprint: inputFingerprintValue,
release: () => undefined,
};
}
}
@@ -9,6 +9,7 @@ export interface RuntimePaths {
sessions: string;
artifacts: string;
indexes: string;
memory: string;
}
export interface RuntimeIdentity {
@@ -255,6 +256,7 @@ export function renderRuntimeConfig(
...(installation.profile === undefined ? {} : { profile: installation.profile }),
language: descriptor.workspace.language,
database,
semantic_index: descriptor.semantic_index,
resources: {
vector: {
engine: "qdrant",