Cognee Cloud API - All datasets share same data_id, search returns placeholder data

[LStromann]

=Hi Cognee team! 👋

I'm building an AI app using Cognee Cloud REST API to store user memories. Each user should have isolated, searchable memories.

The Problem:
All users share the same data_id, and search returns placeholder text instead of actual content.

Test Evidence:

User 1 (Alice):
Added: "My name is Alice. I am a software engineer at Google. I specialize in machine learning and Python programming."

API: POST /api/add with dataset_name="user_alice"
Response: dataset_id=b1f17f72-e3c2-5b7f-b789-199e472d0ffa, data_id=c934b0f4-9927-5a75-b128-ab91c89f0973

Searched for: "Alice software engineer Google Python"
Got back: "Warning: long-term memory may contain dad jokes!" ❌

User 2 (Bob):
Added: "My name is Bob. I am a marketing director at Microsoft. I specialize in digital marketing campaigns."

API: POST /api/add with dataset_name="user_bob"
Response: dataset_id=9a00238f-b1eb-50ea-8607-a8005ea1a69f ✅ (different), data_id=c934b0f4-9927-5a75-b128-ab91c89f0973 ❌ (SAME as Alice!)

Searched for: "Bob marketing director Microsoft"
Got back: "Warning: long-term memory may contain dad jokes!" ❌ (same placeholder)

Both datasets:

• Dataset IDs: Different ✅
• Data IDs: Same ❌ (c934b0f4-9927-5a75-b128-ab91c89f0973)
• Files: Same ❌ (text_e6ce77e066ff6a343efd329c665b4392.txt)
• Search results: Same placeholder ❌

Expected: Different users = different data_ids = different search results
Actual: All users = same data_id = same placeholder

This prevents multi-user memory isolation. Is this a bug? Tested multiple times with cleared datasets. I would appreciate any assistance

This discussion was automatically pulled from Discord.

[LStromann]

<@218266741921284096>, could you have a look at this?

[LStromann]

<@876713586972569640> Hi, we don't have any known issues with cognee.add. Could you please provide the piece of code that adds Alice's and Bob's data?

[LStromann]

Another thing, I don't see cognify being called here. Do you have it in your code?

[LStromann]

Sure 🙂 Here's the exact code I used to test - all via curl commands to Cognee Cloud REST API:

Alice's data:

curl -X POST https://api.cognee.ai/api/add \ -H "Content-Type: application/json" \ -H "X-Api-Key: [my-api-key]" \ -d '{ "data": ["My name is Alice. I am a software engineer at Google. I specialize in machine learning and Python programming. My favorite hobby is hiking in the mountains."], "dataset_name": "user_alice" }'

Response:

{ "status": "PipelineRunCompleted", "dataset_id": "b1f17f72-e3c2-5b7f-b789-199e472d0ffa", "data_id": "c934b0f4-9927-5a75-b128-ab91c89f0973" }

Then cognify:

curl -X POST https://api.cognee.ai/api/cognify \ -H "Content-Type: application/json" \ -H "X-Api-Key: [my-api-key]" \ -d '{ "dataset_ids": ["b1f17f72-e3c2-5b7f-b789-199e472d0ffa"] }'

[LStromann]

I will provide for Bob

[LStromann]

Bob's data:

curl -X POST https://api.cognee.ai/api/add \ -H "Content-Type: application/json" \ -H "X-Api-Key: [my-api-key]" \ -d '{ "data": ["My name is Bob. I am a marketing director at Microsoft. I specialize in digital marketing campaigns. My favorite hobby is playing guitar and writing music."], "dataset_name": "user_bob" }'

Response:

{ "status": "PipelineRunCompleted", "dataset_id": "9a00238f-b1eb-50ea-8607-a8005ea1a69f", "data_id": "c934b0f4-9927-5a75-b128-ab91c89f0973" // Same data_id as Alice! }

Check Alice's dataset
curl -L -H "X-Api-Key: [my-api-key]" \ https://api.cognee.ai/api/datasets/b1f17f72-e3c2-5b7f-b789-199e472d0ffa/data

Check Bob's dataset
curl -L -H "X-Api-Key: [my-api-key]" \ https://api.cognee.ai/api/datasets/9a00238f-b1eb-50ea-8607-a8005ea1a69f/data

[LStromann]

I will reset the datasets and will test again, but i am calling cognify for sure

[LStromann]

Thanks for providing the code, will check it to see what can be wrong

[LStromann]

Thank you very much 🙂 I am looking forward to complete the integration. Will wait for your conclusion

[LStromann]

<@876713586972569640> The parameter for add is called text_data. I think that right now the API is using the default value for text_data and ignoring data that you provided.

[LStromann]

Thanks for identifying the issue. Switching to text_data fixed it.

Tested again with the same users:
Alice: got unique data_id 3ba96d21-850b-5fe8-b62d-6ab2f318ef58
Bob: got unique data_id 5ec9d526-32ea-5e30-8bb1-f8c89b2a85f5

Search now returns actual user content instead of placeholder data. I will proceed with the integration on my side and will reach out if any other issues appear. Thank you so much

[LStromann]

Good to hear that 🙂 Glad I could help

[LStromann]

You can see it also documented here:
https://api.cognee.ai/docs#/add/add_text_api_add_post

[LStromann]

Thank you