Describe the problem/error/question
I’m using an AI Agent node that executes a subworkflow.
Some of the fields of the subworkflow should be filled by the AI Node. When I fill only one field with the AI Node it works. As soon as the AI Agent should fill two fields it doesn’t work anymore.
I have tried searching a lot and I cannot find why it will fail. Probably somewhere some minor tweak I need to adjust I just don’t happen to find it.
Here is my very simple Agent:
This execution will fail:
The error looks like this:
This is the input into the subworkflow from the agent.
[
{
"query": {
"vdb_query": "Purora emotional social media post examples",
"filter": "{\"must\":[{\"key\":\"type\",\"match\":{\"value\":\"social_media_post\"}},{\"key\":\"tone\",\"match\":{\"any\":[\"emotional\",\"story\",\"heartfelt\",\"inspirational\"]}}]}"
}
}
]
This is all the same but now I only fill one component with AI. Now all works.
This is the input from the AI Agent
[
{
"query": {
"vdb_query": "Purora Social-Media-Stil Beispiel: emotionaler Post über einen Smoothie, achtsamer Lebensstil, Gemeinschaft, Gesundheit, einfache Sprache, strukturierte Aufzählung mit Emojis"
}
}
]
[
{
"result": {
"points": [
{
"id": "e1b4b10c-41bd-5ae4-9e3d-d66141b71e10",
"version": 46,
"score": 0.6839099,
"payload": {
"ID": "23",
"Topic": "Smoothie am See",
"Type": "Community",
"Aktion": "Bewusstseinbildend",
"Style": "Inspirierend",
"TargetGroup": "Alle",
"Platform": "Instagram, Facebook",
"Hashtags": "#mypuroramoment #purdu #changeforthebetter #lake #healthybreak ....
"status": "ok",
"time": 0.001376942
}
]
What is the error message (if any)?
This is the output of the subworkflow unfortunately it doesn t contain an Error Message.
{
"errorMessage": "",
"errorDetails": {},
"n8nDetails": {
"n8nVersion": "1.112.4 (Self Hosted)",
"binaryDataMode": "default",
"cause": {
"with_vector": false,
"with_payload": true,
"isTest": false,
"collection": "purora",
"limit": 5,
"offsett": 0,
"Timeout": 100,
"embedding": "text-embedding-3-small",
"encoding_format": "float",
"score_threshhold": 0,
"vdb_query": "={{ /*n8n-auto-generated-fromAI-override*/ $fromAI('vdb_query', `Write here the text you want to query the Qdrant Database. Just a string asking it. `, 'string') }}",
"filter": "={{ /*n8n-auto-generated-fromAI-override*/ $fromAI('filter', `A Qdrant filter object, formatted as a valid, minified JSON string. \\\\n\\\\n**CRITICAL:** The syntax MUST be 100% perfect Qdrant filter format or the entire operation will fail. Do not include any explanations or markdown formatting. \\\\n\\\\n**STRUCTURE:** \\\\nThe top-level object must have a key of \\\\must\\\\, \\\\should\\\\, or \\\\must_not\\\\. The value is an array of condition objects. \\\\n\\\\n**CONDITION TEMPLATES:** \\\\n**For exact text or keyword matches**, use the match template: \\\\n {\"key\": \"field_name\", \"match\": {\"value\": \"string_to_match\"}} For multiple text or keyword matches**, use the match and any template: \\\\n {\"key\": \"field_name\", \"match\": {\"any\": [\"string_to_match\", \"string_to_match2\"]}}`, 'string') }}"
}
}
}
Please share your workflow
This tool doesn’t work:
This tool works:
This is the subworkflow
Share the output returned by the last node
Information on your n8n setup
- n8n version: 1.112.4 (Self Hosted)
- Database (default: SQLite): pqsql
- n8n EXECUTIONS_PROCESS setting (default: own, main): own
- Running n8n via (Docker, npm, n8n cloud, desktop app): Docker
- Operating system: QNAP NAS
