fix: field normalization, AI connection validation, autofix filter (#581) (#638)

- Normalize name→nodeName and id→nodeId for node-targeting operations in
  the Zod schema transform, so LLMs using natural field names no longer
  get "Node not found" errors
- Replace hardcoded ALL_CONNECTION_TYPES with dynamic iteration so AI
  sub-nodes (ai_outputParser, ai_document, ai_textSplitter, etc.) are
  not flagged as disconnected during save
- Add .catchall() to workflowConnectionSchema and extend connection
  reference validation to cover all connection types, not just main
- Fix filterOperationsByFixes ID-vs-name mismatch: typeversion-upgrade
  operations now include nodeName alongside nodeId, and the filter checks
  both fields

Conceived by Romuald Członkowski - https://www.aiadvisors.pl/en

Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
Romuald Członkowski
2026-03-15 14:32:14 +01:00
committed by GitHub
parent 65ab94deb2
commit f7a1cfe8bf
41 changed files with 657 additions and 154 deletions

View File

@@ -47,7 +47,7 @@ exports.workflowConnectionSchema = zod_1.z.record(zod_1.z.object({
ai_memory: connectionArraySchema.optional(),
ai_embedding: connectionArraySchema.optional(),
ai_vectorStore: connectionArraySchema.optional(),
}));
}).catchall(connectionArraySchema));
exports.workflowSettingsSchema = zod_1.z.object({
executionOrder: zod_1.z.enum(['v0', 'v1']).default('v1'),
timezone: zod_1.z.string().optional(),
@@ -152,11 +152,10 @@ function validateWorkflowStructure(workflow) {
}
else if (connectionCount > 0 || executableNodes.length > 1) {
const connectedNodes = new Set();
const ALL_CONNECTION_TYPES = ['main', 'error', 'ai_tool', 'ai_languageModel', 'ai_memory', 'ai_embedding', 'ai_vectorStore'];
Object.entries(workflow.connections).forEach(([sourceName, connection]) => {
connectedNodes.add(sourceName);
ALL_CONNECTION_TYPES.forEach(connType => {
const connData = connection[connType];
const connectionRecord = connection;
Object.values(connectionRecord).forEach((connData) => {
if (connData && Array.isArray(connData)) {
connData.forEach((outputs) => {
if (Array.isArray(outputs)) {
@@ -282,23 +281,28 @@ function validateWorkflowStructure(workflow) {
errors.push(`Connection references non-existent node: ${sourceName}`);
}
}
if (connection.main && Array.isArray(connection.main)) {
connection.main.forEach((outputs, outputIndex) => {
if (Array.isArray(outputs)) {
outputs.forEach((target, targetIndex) => {
if (!nodeNames.has(target.node)) {
if (nodeIds.has(target.node)) {
const correctName = nodeIdToName.get(target.node);
errors.push(`Connection target uses node ID '${target.node}' but must use node name '${correctName}' (from ${sourceName}[${outputIndex}][${targetIndex}])`);
const connectionRecord = connection;
Object.values(connectionRecord).forEach((connData) => {
if (connData && Array.isArray(connData)) {
connData.forEach((outputs, outputIndex) => {
if (Array.isArray(outputs)) {
outputs.forEach((target, targetIndex) => {
if (!target?.node)
return;
if (!nodeNames.has(target.node)) {
if (nodeIds.has(target.node)) {
const correctName = nodeIdToName.get(target.node);
errors.push(`Connection target uses node ID '${target.node}' but must use node name '${correctName}' (from ${sourceName}[${outputIndex}][${targetIndex}])`);
}
else {
errors.push(`Connection references non-existent target node: ${target.node} (from ${sourceName}[${outputIndex}][${targetIndex}])`);
}
}
else {
errors.push(`Connection references non-existent target node: ${target.node} (from ${sourceName}[${outputIndex}][${targetIndex}])`);
}
}
});
}
});
}
});
}
});
}
});
});
}
return errors;