AI_APICallError: Not Found

[wangscn]

The ollama local 127.0.0.1:11434/v1 test has passed, the model has been added, but an AI_APICallError: Not Found error occurs. How should this be ha

ndled?

[wangscn]

The container has already been restarted

[bbird81]

I have the same problem, but only when the "knowledge base search" button is enabled:

I do have though an embedding model selected (EmbeddingGemma 300B).

I tried to tcpdump the traffic and I see that the embed api call seems to be implemented wrong as it tries to POST on /embed (instead of the /api/embed).
Here's what I captured from tcpdump:

POST /embed HTTP/1.1
host: x.x.x.x:11434
connection: keep-alive
Content-Type: application/json
accept: /
accept-language: *
sec-fetch-mode: cors
user-agent: node
accept-encoding: gzip, deflate
content-length: 58

{"input":["My input"],"model":"embeddinggemma:300m"}

If I disable knowledge base search, the chat works correctly and I see in tcpdump that the API endpoint called is indeed correct /api/chat:
POST /api/chat HTTP/1.1
host: x.x.x.x:11434
connection: keep-alive
Content-Type: application/json
accept: /
accept-language: *
sec-fetch-mode: cors
user-agent: node
accept-encoding: gzip, deflate
content-length: 504

{"model":"llama3.2:3b","options":{"temperature":0},"messages":[{"content":"You are a conversation title summarizer. Rules:\n 1. Summarize the content \n 2. Return the title only\n 3. Generate titles based on the user's language\n 4. Do not return any punctuation marks in the result\n 5. Keep it short and concise","role":"system"},{"content":"[\n {\n "role": "user",\n "content": "MY INPUT"\n }\n]","role":"user"}],"stream":false}

[bbird81]

I tried to change the base url for the Ollama provider from http://x.x.x.x:11434/ to http://x.x.x.x:11434/api
Doing that, chat still works and from the traffic capture I see the emdedding call now is correctly sent to /api/embed; I receive vectors back so it seems that UI should make clear where to point the url (also the check routine should be corrected so that one is not erroneously brought to think that the settings are correct).

Unfortunately, while the API seems to work ok with this, I still got an error from the UI:

{"message":"SQLITE_ERROR: no such table: blinko","details":{"message":"SQLITE_ERROR: no such table: blinko","domain":"STORAGE","category":"THIRD_PARTY","details":{}},"code":"LIBSQL_VECTOR_QUERY_FAILED"}

So I tried to force rebuild the index and that seems to have corrected the issue (I don't get any errors).

I do get an error during rebuilding though, on the images:

I do have a vision model set (llama3.2-vision:11b).

Is this normal?

Also @wangscn let me know if anything I wrote helps you :)