refactor: make workspace descriptors schema v3 only

This commit is contained in:
2026-08-10 20:21:20 +02:00
parent 5d016ce635
commit 512b0261b1
12 changed files with 187 additions and 295 deletions
+14 -175
View File
@@ -34,8 +34,8 @@ export interface CanonicalDiagnostics {
embedding?: RestDiagnosticRequest & { response: { model: string; dimensions: string } };
}
interface WorkspaceMetadata<Version extends 1 | 2 | 3> {
schema_version: Version;
interface WorkspaceMetadata {
schema_version: 3;
id: string;
name: string;
description?: string;
@@ -51,50 +51,22 @@ interface WorkspaceDwh {
supported_transports: DwhTransport[];
}
interface VectorStore {
engine: "pgvector";
collection: string;
dimensions: number;
distance: "cosine" | "l2" | "inner_product";
port?: number;
timeout_ms?: number;
supported_transports: VectorTransport[];
}
interface SemanticIndex<TVectorStore extends VectorStore> {
vector_store: TVectorStore;
vector_writer?: Record<string, never>;
embedding: {
provider: "ollama_compatible" | "openai_compatible";
model: string;
dimensions: number;
timeout_ms?: number;
};
}
interface WorkspaceBase<Version extends 1 | 2 | 3, TVectorStore, TDiagnostics = CanonicalDiagnostics> {
workspace: WorkspaceMetadata<Version>;
interface WorkspaceBase<TVectorStore> {
workspace: WorkspaceMetadata;
dwh: WorkspaceDwh;
semantic_index: {
vector_store: TVectorStore;
vector_writer?: Record<string, never>;
embedding: Version extends 3 ? {
embedding: {
provider: "ollama_internal";
model: "qwen3-embedding:0.6b";
dimensions: 1024;
timeout_ms?: number;
} : {
provider: "ollama_compatible" | "openai_compatible";
model: string;
dimensions: number;
timeout_ms?: number;
};
};
llm_policy: {
default?: `${string}/${string}`;
allowed: `${string}/${string}`[];
};
diagnostics?: TDiagnostics;
diagnostics?: Pick<CanonicalDiagnostics, "dwh_rest">;
}
interface QdrantVectorStore {
@@ -147,17 +119,12 @@ export interface WorkspaceEvidence {
policy: EvidencePolicy;
}
export interface WorkspaceV3 extends WorkspaceBase<3, QdrantVectorStore> {
export interface WorkspaceV3 extends WorkspaceBase<QdrantVectorStore> {
evidence?: WorkspaceEvidence;
}
export interface WorkspaceV2 extends WorkspaceBase<2, VectorStore & { database: string; schema: string }> {}
/** A readable, non-operational v1 descriptor. It must be explicitly migrated before use. */
export interface WorkspaceV1 extends WorkspaceBase<1, VectorStore & { database?: string; schema?: string }> {}
export type CanonicalWorkspace = WorkspaceV3;
export type LegacyWorkspace = WorkspaceV1 | WorkspaceV2;
export type WorkspaceDescriptor = WorkspaceV1 | WorkspaceV2 | WorkspaceV3;
export type WorkspaceDescriptor = WorkspaceV3;
const workspaceId = z.string().regex(/^[a-z][a-z0-9-]{2,62}$/, {
message: "workspace id must match ^[a-z][a-z0-9-]{2,62}$",
@@ -165,7 +132,6 @@ const workspaceId = z.string().regex(/^[a-z][a-z0-9-]{2,62}$/, {
const identifier = z.string().regex(/^[A-Za-z_][A-Za-z0-9_]*$/, {
message: "database identifiers must start with a letter or underscore",
});
const dimensions = z.number().int().positive().max(32_768);
const port = z.number().int().min(1).max(65_535);
const timeoutMs = z.number().int().positive();
const modelReference = z.string().regex(/^[^/\s]+\/[^/\s]+$/, {
@@ -190,29 +156,6 @@ const restDiagnosticRequest = z.object({
const dwhRestDiagnostic = restDiagnosticRequest.extend({
response: z.object({ database: responseField, schema: responseField }).strict(),
}).strict();
const vectorMetadataDiagnostic = restDiagnosticRequest.extend({
response: z.object({
collection: responseField,
dimensions: responseField,
distance: responseField,
}).strict(),
}).strict();
const reversibleVectorProbe = restDiagnosticRequest.extend({
method: z.literal("POST"),
auth: z.enum(["bearer", "x-api-key"]),
response: z.object({ operation: responseField }).strict(),
}).strict();
const embeddingDiagnostic = restDiagnosticRequest.extend({
response: z.object({ model: responseField, dimensions: responseField }).strict(),
}).strict();
const diagnosticsSchema = z.object({
dwh_rest: dwhRestDiagnostic.optional(),
vector_rest: z.object({
metadata: vectorMetadataDiagnostic,
reversible_probe: reversibleVectorProbe.optional(),
}).strict().optional(),
embedding: embeddingDiagnostic.optional(),
}).strict().optional();
const dwhSchema = z.object({
engine: z.literal("postgres"),
@@ -222,36 +165,11 @@ const dwhSchema = z.object({
timeout_ms: timeoutMs.optional(),
supported_transports: z.array(z.enum(DWH_TRANSPORTS)).min(1),
}).strict();
const embeddingSchema = z.object({
provider: z.enum(["ollama_compatible", "openai_compatible"]),
model: z.string().trim().min(1),
dimensions,
timeout_ms: timeoutMs.optional(),
}).strict();
const internalEmbeddingSchema = z.object({
provider: z.literal("ollama_internal"),
model: z.literal("qwen3-embedding:0.6b"),
dimensions: z.literal(1024),
}).strict();
const vectorStoreShape = {
engine: z.literal("pgvector"),
collection: identifier,
dimensions,
distance: z.enum(["cosine", "l2", "inner_product"]),
port: port.optional(),
timeout_ms: timeoutMs.optional(),
supported_transports: z.array(z.enum(VECTOR_TRANSPORTS)).min(1),
};
const legacyVectorStoreSchema = z.object({
...vectorStoreShape,
database: identifier.optional(),
schema: identifier.optional(),
}).strict();
const canonicalVectorStoreSchema = z.object({
...vectorStoreShape,
database: identifier,
schema: identifier,
}).strict();
const qdrantVectorStoreSchema = z.object({
engine: z.literal("qdrant"),
collection: workspaceId,
@@ -420,13 +338,6 @@ function unique<T>(values: readonly T[], context: z.RefinementCtx, path: Propert
function workspaceInvariants(workspace: any, context: z.RefinementCtx): void {
unique(workspace.dwh.supported_transports, context, ["dwh", "supported_transports"]);
if ("supported_transports" in workspace.semantic_index.vector_store) {
unique(
workspace.semantic_index.vector_store.supported_transports,
context,
["semantic_index", "vector_store", "supported_transports"],
);
}
unique(workspace.llm_policy.allowed, context, ["llm_policy", "allowed"]);
if (workspace.evidence?.source.type === "filesystem") {
@@ -461,50 +372,8 @@ function workspaceInvariants(workspace: any, context: z.RefinementCtx): void {
message: "diagnostics.dwh_rest requires dwh rest_api transport support",
});
}
if (
workspace.diagnostics?.vector_rest
&& (
!("supported_transports" in workspace.semantic_index.vector_store)
|| !workspace.semantic_index.vector_store.supported_transports.includes("rest_api")
)
) {
context.addIssue({
code: "custom",
path: ["diagnostics", "vector_rest"],
message: "diagnostics.vector_rest requires vector_store rest_api transport support",
});
}
}
const workspaceShape = {
dwh: dwhSchema,
llm_policy: llmPolicySchema,
diagnostics: diagnosticsSchema,
};
const WorkspaceV1Schema = z.object({
...workspaceShape,
workspace: z.object({
schema_version: z.literal(1), id: workspaceId, name: z.string().trim().min(1),
description: z.string().trim().min(1).optional(), language: z.enum(["en", "it"]),
}).strict(),
semantic_index: z.object({
vector_store: legacyVectorStoreSchema,
vector_writer: z.object({}).strict().optional(),
embedding: embeddingSchema,
}).strict(),
}).strict().superRefine(workspaceInvariants);
const WorkspaceV2Schema = z.object({
...workspaceShape,
workspace: z.object({
schema_version: z.literal(2), id: workspaceId, name: z.string().trim().min(1),
description: z.string().trim().min(1).optional(), language: z.enum(["en", "it"]),
}).strict(),
semantic_index: z.object({
vector_store: canonicalVectorStoreSchema,
vector_writer: z.object({}).strict().optional(),
embedding: embeddingSchema,
}).strict(),
}).strict().superRefine(workspaceInvariants);
const WorkspaceV3Schema = z.object({
dwh: dwhSchema,
llm_policy: llmPolicySchema,
@@ -521,7 +390,7 @@ const WorkspaceV3Schema = z.object({
embedding: internalEmbeddingSchema,
}).strict(),
}).strict().superRefine(workspaceInvariants);
const WorkspaceDescriptorSchema = z.union([WorkspaceV3Schema, WorkspaceV2Schema, WorkspaceV1Schema]);
const WorkspaceDescriptorSchema = WorkspaceV3Schema;
export function parseWorkspaceYaml(source: string): WorkspaceDescriptor {
const documents = parseAllDocuments(source, { uniqueKeys: true });
@@ -538,46 +407,16 @@ export function validateWorkspaceDescriptor(workspace: unknown): WorkspaceDescri
return WorkspaceDescriptorSchema.parse(workspace) as WorkspaceDescriptor;
}
export function isCanonicalWorkspace(workspace: WorkspaceDescriptor): workspace is CanonicalWorkspace {
return workspace.workspace.schema_version === 3;
export function isCanonicalWorkspace(workspace: unknown): workspace is CanonicalWorkspace {
return WorkspaceDescriptorSchema.safeParse(workspace).success;
}
export function isOperationalWorkspace(workspace: WorkspaceDescriptor): workspace is WorkspaceV3 {
return workspace.workspace.schema_version === 3;
export function isOperationalWorkspace(workspace: unknown): workspace is WorkspaceV3 {
return WorkspaceDescriptorSchema.safeParse(workspace).success;
}
/** Rejects readable v1 descriptors at every operational boundary until a caller migrates them. */
export function validateCanonicalWorkspace(workspace: unknown): CanonicalWorkspace {
return validateOperationalWorkspace(workspace);
}
/** Rejects readable v1/v2 descriptors at every operational boundary until a caller migrates them. */
export function validateOperationalWorkspace(workspace: unknown): WorkspaceV3 {
const descriptor = validateWorkspaceDescriptor(workspace);
if (!isOperationalWorkspace(descriptor)) {
throw new Error("Workspace descriptor requires explicit migration to schema version 3");
}
return descriptor;
}
/**
* Explicitly upgrades a readable v1 descriptor. The caller must supply vector identity; the
* transformer never derives it from DWH identity, even where both services share a database.
*/
export function migrateWorkspaceV1ToV2(
workspace: WorkspaceV1,
vectorIdentity: { database: string; schema: string },
): WorkspaceV2 {
const legacy = WorkspaceV1Schema.parse(workspace) as WorkspaceV1;
const identity = z.object({ database: identifier, schema: identifier }).strict().parse(vectorIdentity);
return WorkspaceV2Schema.parse({
...legacy,
workspace: { ...legacy.workspace, schema_version: 2 },
semantic_index: {
...legacy.semantic_index,
vector_store: { ...legacy.semantic_index.vector_store, ...identity },
},
}) as WorkspaceV2;
return validateWorkspaceDescriptor(workspace);
}
/** Builds a request URL only after rejecting values that can leave the declared service origin. */