Ingest 404 when using ollama

[j-pielen]

trafficstars

Title

Ollama Embedding Provider returns 404 error during ingestion despite successful curl test

Description

When using Ollama as the embedding provider for R2R, the ingestion process fails with a 404 error. However, Ollama is running correctly and responds to curl requests as expected.

Steps to Reproduce

1. Start Ollama locally (Ollama is running on http://localhost:11434)
2. Configure R2R to use Ollama as the embedding provider (see configuration below)
3. Attempt to ingest a document through the R2R interface

Expected Behavior

R2R should successfully connect to Ollama and generate embeddings for the ingested document.

Actual Behavior

R2R fails to connect to Ollama, returning a 404 error:

2024-09-15 13:29:51,205 - INFO - core.main.r2r - Starting R2R with version 3.0.8
**External Ollama instance detected and responsive.**
2024-09-15 13:31:59,086 - ERROR - core.providers.embeddings.ollama - Error getting embeddings: 404 page not found
2024-09-15 13:31:59,086 - WARNING - core.base.providers.embedding - Request failed (attempt 1): Error getting embeddings: 404 page not found
2024-09-15 13:32:00,091 - ERROR - core.providers.embeddings.ollama - Error getting embeddings: 404 page not found
2024-09-15 13:32:00,091 - WARNING - core.base.providers.embedding - Request failed (attempt 2): Error getting embeddings: 404 page not found
2024-09-15 13:32:00,094 - ERROR - core.base.pipeline.base_pipeline - Pipeline failed with error: Error getting embeddings: 404 page not found

Configuration

Here's the relevant part of the TOML configuration used:

[embedding]
provider = "ollama"
base_model = "mxbai-embed-large"
base_dimension = 512
batch_size = 128
add_title_as_prefix = false
rerank_model = "None"
concurrent_request_limit = 256

Environment

• R2R Version: 3.1.0 (using local version with poetry)
• Operating System: Ubuntu

Additional Information

Ollama is running correctly and responds to curl requests. Here's a successful curl test:

curl http://localhost:11434/api/embeddings -d '{
  "model": "mxbai-embed-large",
  "prompt": "Llamas are members of the camelid family"
}'

Response:

{"embedding":[0.5869068503379822,1.1740394830703735,0.6426560282707214,0.8012367486953735,-0.13272863626480103,0.646376371383667,-0.44366782903671265,0.8876819610595703,0.6493498682975769,-0.24569851160049438,0.6341419816017151,0.3442591726779938,-0.6307538151741028,0.1385183334350586,-0.521152138710022,0.6995735168457031,0.21664012968540192,-0.16669663786888123,-0.141....