feat: add AI catalog description generation

This commit is contained in:
Codex
2026-08-29 16:42:56 +02:00
parent b0afba81ca
commit 376dd5a09d
76 changed files with 14860 additions and 102 deletions
@@ -0,0 +1,72 @@
import type { FastifyInstance, FastifyReply, FastifyRequest } from "fastify";
import { z } from "zod";
import { isPrincipalContext, requirePermission } from "../auth/authorization.js";
import type { CatalogOperationCoordinator } from "../catalog/operation-coordinator.js";
import {
CatalogOperationInProgressError,
CatalogUnavailableError,
type CatalogRepository,
} from "../catalog/types.js";
const idSchema = z.uuid();
const consolidationSchema = z.object({
target: z.enum(["tables", "columns"]),
targetIds: z.array(idSchema).min(1).max(10_000),
}).strict();
function manage(request: FastifyRequest, reply: FastifyReply) {
return isPrincipalContext(requirePermission(request, reply, "database.manage"));
}
function safeError(reply: FastifyReply, error: unknown) {
if (error instanceof CatalogUnavailableError) {
return reply.code(503).send({ code: "catalog_unavailable", message: "Database catalog is unavailable." });
}
if (error instanceof CatalogOperationInProgressError) {
return reply.code(409).send({
code: "database_operation_in_progress",
message: "A database operation is already in progress.",
});
}
if (error instanceof z.ZodError) {
return reply.code(400).send({
code: "description_consolidation_invalid",
message: "Description consolidation request is invalid.",
});
}
return reply.code(500).send({
code: "description_consolidation_failed",
message: "Description consolidation failed.",
});
}
export function catalogDescriptionConsolidationRoutes(
app: FastifyInstance,
deps: { repository: CatalogRepository; operations: CatalogOperationCoordinator },
): void {
app.post("/catalog/databases/:databaseId/descriptions/consolidate", async (request, reply) => {
if (!manage(request, reply)) return reply;
try {
const databaseId = idSchema.parse((request.params as { databaseId?: unknown }).databaseId);
const input = consolidationSchema.parse(request.body);
const targetIds = [...new Set(input.targetIds)];
const result = await deps.operations.run(
databaseId,
async () => await deps.repository.consolidateGeneratedDescriptions(
databaseId,
input.target,
targetIds,
),
);
if (!result) {
return reply.code(404).send({
code: "catalog_target_not_found",
message: "The database or one or more selected catalog targets were not found.",
});
}
return result;
} catch (error) {
return safeError(reply, error);
}
});
}
@@ -0,0 +1,347 @@
import type { FastifyInstance, FastifyReply, FastifyRequest } from "fastify";
import { z } from "zod";
import { isPrincipalContext, requirePermission } from "../auth/authorization.js";
import {
DescriptionGenerationDuplicateTargetIdsError,
DescriptionGenerationNoEligibleTargetsError,
DescriptionGenerationRunLiveError,
DescriptionGenerationTargetIdsRequiredError,
DescriptionGenerationTargetNotFoundError,
DescriptionGenerationWorkspaceUnavailableError,
type DescriptionGenerationWorker,
} from "../catalog/description-generation-worker.js";
import { MetadataGenerationModelUnavailableError } from "../catalog/metadata-generation-models.js";
import {
CatalogOperationInProgressError,
CatalogUnavailableError,
DescriptionGenerationRunActiveError,
type CatalogRepository,
type DescriptionGenerationEvent,
type DescriptionGenerationRun,
} from "../catalog/types.js";
const idSchema = z.uuid();
const modelIdSchema = z.string().regex(/^[a-z][a-z0-9._-]{0,63}$/);
const selectedTargetIdsSchema = z.array(idSchema).min(1);
const startSchema = z.discriminatedUnion("scope", [
z.object({
modelId: modelIdSchema,
scope: z.literal("selected_columns"),
targetIds: selectedTargetIdsSchema,
}).strict(),
z.object({
modelId: modelIdSchema,
scope: z.literal("selected_tables"),
targetIds: selectedTargetIdsSchema,
}).strict(),
z.object({ modelId: modelIdSchema, scope: z.literal("all") }).strict(),
z.object({ modelId: modelIdSchema, scope: z.literal("missing") }).strict(),
]);
const eventQuerySchema = z.object({
after: z.coerce.number().int().nonnegative().default(0),
}).strict();
const historyQuerySchema = z.object({
limit: z.coerce.number().int().min(1).max(100).default(50),
}).strict();
const terminalStatuses = new Set<DescriptionGenerationRun["status"]>([
"completed",
"completed_with_errors",
"cancelled",
"failed",
"interrupted",
]);
function manage(request: FastifyRequest, reply: FastifyReply) {
return isPrincipalContext(requirePermission(request, reply, "database.manage"));
}
function publicEvent(event: DescriptionGenerationEvent) {
return {
sequence: event.sequence,
level: event.level,
message: event.message,
createdAt: event.createdAt,
};
}
function publicRun(run: DescriptionGenerationRun) {
return {
id: run.id,
databaseId: run.databaseId,
scope: run.scope,
modelId: run.modelId,
language: run.language,
status: run.status,
total: run.total,
processed: run.processed,
generated: run.generated,
nonGeneratable: run.nonGeneratable,
failed: run.failed,
createdAt: run.createdAt,
startedAt: run.startedAt,
updatedAt: run.updatedAt,
finishedAt: run.finishedAt,
errorSummary: run.errorSummary,
};
}
function safeError(reply: FastifyReply, error: unknown) {
if (error instanceof CatalogUnavailableError) {
return reply.code(503).send({
code: "catalog_unavailable",
message: "Database catalog is unavailable.",
});
}
if (error instanceof DescriptionGenerationRunActiveError) {
return reply.code(409).send({
code: "description_generation_run_active",
message: "A Description Generation Run is already active.",
});
}
if (error instanceof DescriptionGenerationRunLiveError) {
return reply.code(409).send({
code: "description_generation_run_live",
message: "A local Description Generation worker or helper is still running.",
});
}
if (error instanceof CatalogOperationInProgressError) {
return reply.code(409).send({
code: "database_operation_in_progress",
message: "A database operation is already in progress.",
});
}
if (error instanceof MetadataGenerationModelUnavailableError) {
return reply.code(409).send({
code: "metadata_generation_model_unavailable",
message: "The selected metadata-generation model is unavailable.",
});
}
if (error instanceof DescriptionGenerationDuplicateTargetIdsError) {
return reply.code(400).send({
code: "description_generation_target_ids_duplicate",
message: "Description generation target IDs must be unique.",
});
}
if (error instanceof DescriptionGenerationTargetIdsRequiredError) {
return reply.code(400).send({
code: "description_generation_request_invalid",
message: "At least one description generation target ID is required.",
});
}
if (error instanceof DescriptionGenerationNoEligibleTargetsError) {
return reply.code(409).send({
code: "description_generation_no_eligible_targets",
message: error.scope === "all"
? "No Catalog Tables or Catalog Columns are available for description generation."
: "No Catalog Tables or Catalog Columns have a missing Generated Description.",
});
}
if (error instanceof DescriptionGenerationTargetNotFoundError) {
const code = error.target === "database"
? "database_not_found"
: error.target === "table"
? "catalog_table_not_found"
: "catalog_column_not_found";
const message = error.target === "database"
? "Database configuration was not found."
: error.target === "table"
? "One or more selected Catalog Tables were not found."
: "One or more selected Catalog Columns were not found.";
return reply.code(404).send({
code,
message,
});
}
if (error instanceof DescriptionGenerationWorkspaceUnavailableError) {
return reply.code(409).send({
code: "workspace_configuration_unavailable",
message: "The database workspace configuration is unavailable.",
});
}
if (error instanceof z.ZodError) {
return reply.code(400).send({
code: "description_generation_request_invalid",
message: "Description generation request is invalid.",
});
}
return reply.code(500).send({
code: "description_generation_failed",
message: "Description generation failed.",
});
}
export function catalogDescriptionGenerationRoutes(
app: FastifyInstance,
deps: { repository: CatalogRepository; worker: DescriptionGenerationWorker },
): void {
app.post("/catalog/databases/:databaseId/description-generation-runs", async (request, reply) => {
if (!manage(request, reply)) return reply;
try {
const databaseId = idSchema.parse((request.params as { databaseId?: unknown }).databaseId);
const input = startSchema.parse(request.body);
const targetIds = "targetIds" in input ? input.targetIds : [];
const run = await deps.worker.start(databaseId, input.modelId, input.scope, targetIds);
return reply.code(202).send(publicRun(run));
} catch (error) {
return safeError(reply, error);
}
});
app.get("/catalog/description-generation-runs", async (request, reply) => {
if (!manage(request, reply)) return reply;
try {
const { limit } = historyQuerySchema.parse(request.query);
return (await deps.repository.listDescriptionGenerationRuns(limit)).map(publicRun);
} catch (error) {
return safeError(reply, error);
}
});
app.get("/catalog/description-generation-runs/:runId", async (request, reply) => {
if (!manage(request, reply)) return reply;
try {
const runId = idSchema.parse((request.params as { runId?: unknown }).runId);
const run = await deps.repository.getDescriptionGenerationRun(runId);
return run ? publicRun(run) : reply.code(404).send({
code: "description_generation_run_not_found",
message: "Description Generation Run was not found.",
});
} catch (error) {
return safeError(reply, error);
}
});
app.post("/catalog/description-generation-runs/:runId/cancel", async (request, reply) => {
if (!manage(request, reply)) return reply;
try {
const runId = idSchema.parse((request.params as { runId?: unknown }).runId);
const run = await deps.worker.cancel(runId);
return run ? publicRun(run) : reply.code(404).send({
code: "description_generation_run_not_found",
message: "Description Generation Run was not found.",
});
} catch (error) {
return safeError(reply, error);
}
});
app.post("/catalog/description-generation-runs/unlock", async (request, reply) => {
if (!manage(request, reply)) return reply;
try {
const run = await deps.worker.unlock();
return run ? publicRun(run) : reply.code(404).send({
code: "description_generation_run_not_found",
message: "No stale active Description Generation Run was found.",
});
} catch (error) {
return safeError(reply, error);
}
});
app.get("/catalog/description-generation-runs/:runId/events", async (request, reply) => {
if (!manage(request, reply)) return reply;
let unsubscribe: (() => void) | undefined;
let hijacked = false;
try {
const runId = idSchema.parse((request.params as { runId?: unknown }).runId);
let after = eventQuerySchema.parse(request.query).after;
const headerCursor = Number(request.headers["last-event-id"]);
if (Number.isInteger(headerCursor) && headerCursor >= 0) after = Math.max(after, headerCursor);
if (!(await deps.repository.getDescriptionGenerationRun(runId))) {
return reply.code(404).send({
code: "description_generation_run_not_found",
message: "Description Generation Run was not found.",
});
}
const buffered: DescriptionGenerationEvent[] = [];
let ready = false;
let closed = false;
let lastRunSnapshot = "";
let delivery = Promise.resolve();
const close = () => {
if (closed) return;
closed = true;
unsubscribe?.();
if (!reply.raw.destroyed) reply.raw.end();
};
const writeRun = (run: DescriptionGenerationRun) => {
if (closed) return;
const snapshot = JSON.stringify(publicRun(run));
if (snapshot === lastRunSnapshot) return;
lastRunSnapshot = snapshot;
reply.raw.write(`event: run\ndata: ${snapshot}\n\n`);
if (terminalStatuses.has(run.status)) close();
};
const enqueue = (event: DescriptionGenerationEvent) => {
delivery = delivery.then(async () => {
if (closed || event.sequence <= after) return;
after = event.sequence;
reply.raw.write(
`id: ${event.sequence}\nevent: log\ndata: ${JSON.stringify(publicEvent(event))}\n\n`,
);
const run = await deps.repository.getDescriptionGenerationRun(runId);
if (run) writeRun(run);
}).catch(close);
};
unsubscribe = deps.worker.subscribeEvents(runId, (event) => {
if (ready) enqueue(event);
else buffered.push(event);
});
const persisted = await deps.repository.listDescriptionGenerationEvents(runId, after);
reply.hijack();
hijacked = true;
reply.raw.writeHead(200, {
"content-type": "text/event-stream; charset=utf-8",
"cache-control": "no-cache, no-transform",
connection: "keep-alive",
"x-accel-buffering": "no",
});
reply.raw.once("close", close);
request.raw.once("aborted", close);
for (const event of persisted) {
if (event.sequence <= after) continue;
after = event.sequence;
reply.raw.write(
`id: ${event.sequence}\nevent: log\ndata: ${JSON.stringify(publicEvent(event))}\n\n`,
);
}
ready = true;
buffered.sort((a, b) => a.sequence - b.sequence).forEach(enqueue);
let pending = delivery;
await pending;
while (pending !== delivery) {
pending = delivery;
await pending;
}
const current = await deps.repository.getDescriptionGenerationRun(runId);
if (current) writeRun(current);
return reply;
} catch (error) {
unsubscribe?.();
if (hijacked) {
if (!reply.raw.destroyed) reply.raw.end();
return reply;
}
return safeError(reply, error);
}
});
app.get("/catalog/description-generation-runs/:runId/events-list", async (request, reply) => {
if (!manage(request, reply)) return reply;
try {
const runId = idSchema.parse((request.params as { runId?: unknown }).runId);
const { after } = eventQuerySchema.parse(request.query);
if (!(await deps.repository.getDescriptionGenerationRun(runId))) {
return reply.code(404).send({
code: "description_generation_run_not_found",
message: "Description Generation Run was not found.",
});
}
return (await deps.repository.listDescriptionGenerationEvents(runId, after)).map(publicEvent);
} catch (error) {
return safeError(reply, error);
}
});
}
+25 -16
View File
@@ -43,7 +43,13 @@ function safeError(reply: FastifyReply, error: unknown) {
if (error instanceof CatalogUnavailableError) {
return reply.code(503).send({ code: "catalog_unavailable", message: "Database catalog is unavailable." });
}
if (error instanceof CatalogConflictError || error instanceof CatalogOperationInProgressError) {
if (error instanceof CatalogOperationInProgressError) {
return reply.code(409).send({
code: "database_operation_in_progress",
message: "A database operation is already in progress.",
});
}
if (error instanceof CatalogConflictError) {
return reply.code(409).send({ code: "schema_sync_conflict", message: error.message });
}
if (error instanceof CatalogConnectorError) {
@@ -98,21 +104,24 @@ export function catalogSchemaRoutes(
const tableId = idSchema.parse(params.tableId);
const columnId = idSchema.parse(params.columnId);
const input = metadataSchema.parse(request.body);
const current = await deps.repository.getColumn(databaseId, tableId, columnId);
if (!current) return reply.code(404).send({ code: "column_not_found", message: "Catalog column was not found." });
if (current.version !== input.version) {
return reply.code(409).send({ code: "column_stale", message: "Column metadata changed. Reload and try again." });
}
const updated = await deps.repository.updateColumnMetadata(
databaseId,
tableId,
columnId,
input.version,
normalized(input.description),
normalized(input.generatedDescription),
);
if (!updated) return reply.code(409).send({ code: "column_stale", message: "Column metadata changed. Reload and try again." });
return updated;
const operation = async () => {
const current = await deps.repository.getColumn(databaseId, tableId, columnId);
if (!current) return reply.code(404).send({ code: "column_not_found", message: "Catalog column was not found." });
if (current.version !== input.version) {
return reply.code(409).send({ code: "column_stale", message: "Column metadata changed. Reload and try again." });
}
const updated = await deps.repository.updateColumnMetadata(
databaseId,
tableId,
columnId,
input.version,
normalized(input.description),
normalized(input.generatedDescription),
);
if (!updated) return reply.code(409).send({ code: "column_stale", message: "Column metadata changed. Reload and try again." });
return updated;
};
return deps.operations ? await deps.operations.run(databaseId, operation) : await operation();
} catch (error) { return safeError(reply, error); }
});
+22 -15
View File
@@ -2,6 +2,7 @@ import type { FastifyInstance, FastifyReply, FastifyRequest } from "fastify";
import { z } from "zod";
import { isPrincipalContext, requirePermission } from "../auth/authorization.js";
import type { CatalogTableService } from "../catalog/table-service.js";
import type { CatalogOperationCoordinator } from "../catalog/operation-coordinator.js";
import {
CatalogConnectorError,
CatalogOperationInProgressError,
@@ -38,7 +39,11 @@ function safeError(reply: FastifyReply, error: unknown) {
export function catalogTableRoutes(
app: FastifyInstance,
deps: { repository: CatalogRepository; service: CatalogTableService },
deps: {
repository: CatalogRepository;
service: CatalogTableService;
operations: CatalogOperationCoordinator;
},
): void {
app.get("/catalog/databases/:databaseId/tables", async (request, reply) => {
if (!manage(request, reply)) return reply;
@@ -59,20 +64,22 @@ export function catalogTableRoutes(
const tableId = idSchema.parse(params.tableId);
const input = updateSchema.parse(request.body);
const { version, description } = input;
const current = await deps.repository.getTable(databaseId, tableId);
if (!current) return reply.code(404).send({ code: "table_not_found", message: "Catalog table was not found." });
if (current.version !== version) {
return reply.code(409).send({ code: "table_stale", message: "Table description changed. Reload and try again." });
}
const updated = await deps.service.updateMetadata(
databaseId,
tableId,
version,
description,
input.generatedDescription === undefined ? current.generatedDescription : input.generatedDescription,
);
if (!updated) return reply.code(409).send({ code: "table_stale", message: "Table description changed. Reload and try again." });
return updated;
return await deps.operations.run(databaseId, async () => {
const current = await deps.repository.getTable(databaseId, tableId);
if (!current) return reply.code(404).send({ code: "table_not_found", message: "Catalog table was not found." });
if (current.version !== version) {
return reply.code(409).send({ code: "table_stale", message: "Table description changed. Reload and try again." });
}
const updated = await deps.service.updateMetadata(
databaseId,
tableId,
version,
description,
input.generatedDescription === undefined ? current.generatedDescription : input.generatedDescription,
);
if (!updated) return reply.code(409).send({ code: "table_stale", message: "Table description changed. Reload and try again." });
return updated;
});
} catch (error) { return safeError(reply, error); }
});
@@ -0,0 +1,13 @@
import type { FastifyInstance } from "fastify";
import { isPrincipalContext, requirePermission } from "../auth/authorization.js";
import type { MetadataGenerationModels } from "../catalog/metadata-generation-models.js";
export function metadataGenerationModelRoutes(
app: FastifyInstance,
models: MetadataGenerationModels,
): void {
app.get("/catalog/metadata-generation/models", async (request, reply) => {
if (!isPrincipalContext(requirePermission(request, reply, "database.manage"))) return reply;
return models.catalog();
});
}