How to use Natural Language Categorization & Search with Ollama as backend?

[romelsalwi]

I added the following variables to use Ollama as backend, but nothing happened.

Note that Ollama doesn't use API keys for using local models.

OPENAI_BASE_URL="http://localhost:11434"
OPENAI_API_KEY="ollama"
EMBEDDING_MODEL="embeddinggemma:latest"
AI_MODEL="qwen2.5:3b"
IIB_PROMPT_NORMALIZE=1
IIB_PROMPT_NORMALIZE_MODE="balanced"

[zanllp]

First, check what errors are recorded in the log.log file in the root directory.

[romelsalwi]

2026-01-30 06:09:25,731 - scripts.iib.logger - DEBUG - Version probe failed for imageio-ffmpeg/imageio_ffmpeg: No module named 'imageio_ffmpeg'
2026-01-30 06:09:25,733 - scripts.iib.logger - INFO - Version info requested: {'python_version': '3.13.11 (main, Jan 13 2026, 14:34:13) [GCC 14.3.1 20250801]', 'platform': 'Linux-6.12.63-gentoo-dist-x86_64-AMD_Ryzen_5_3500_6-Core_Processor-with-glibc2.41', 'hash': '59283c3f8a6ac567a5484693d92e2f81b108931b', 'tag': 'v1.5.1', 'av': '14.4.0', 'imageio': '2.37.2', 'pillow': '12.1.0', 'pillow_avif_plugin': '1.5.2', 'requests': '2.32.5', 'numpy': '2.4.1', 'hnswlib': '0.8.0'}

[zanllp]

Maybe you can try entering any random value into the API key field first to allow some of the validation logic to run properly.

[romelsalwi]

That's what did by:
OPENAI_API_KEY="ollama"

[zanllp]

I've added a lot of logging. Please try it again now.

[romelsalwi]

2026-01-31 02:16:56,015 - scripts.iib.logger - INFO - [cluster_job] === _run_cluster_job START ===
2026-01-31 02:16:56,015 - scripts.iib.logger - INFO - [cluster_job] job_id=38f7546cb2d94dbdb5e18f749273625f
2026-01-31 02:16:56,015 - scripts.iib.logger - INFO - [cluster_job] Initializing job
2026-01-31 02:16:56,015 - scripts.iib.logger - INFO - [cluster_job] Job initialized successfully
2026-01-31 02:16:56,015 - scripts.iib.logger - INFO - [cluster_job] Extracting and validating folders
2026-01-31 02:16:56,015 - scripts.iib.logger - INFO - [cluster_job] Folders validated: ['/home/l3m0r/Pictures/sdxl']
2026-01-31 02:16:56,015 - scripts.iib.logger - INFO - [build_embeddings] === _build_embeddings_one_folder START ===
2026-01-31 02:16:56,016 - scripts.iib.logger - INFO - [build_embeddings] folder=/home/l3m0r/Pictures/sdxl model=embeddinggemma:latest force=False batch_size=64 max_chars=4000
2026-01-31 02:16:56,016 - scripts.iib.logger - INFO - [build_embeddings] Configuration check passed
2026-01-31 02:16:56,016 - scripts.iib.logger - INFO - [build_embeddings] API URL: http://localhost:11434
2026-01-31 02:16:56,016 - scripts.iib.logger - INFO - [build_embeddings] Normalized folder path: /home/l3m0r/Pictures/sdxl
2026-01-31 02:16:56,111 - scripts.iib.logger - INFO - [build_embeddings] Calling embedding API for batch 1/1, size=5
2026-01-31 02:16:56,112 - scripts.iib.logger - INFO - [embeddings] === _call_embeddings_sync START ===
2026-01-31 02:16:56,112 - scripts.iib.logger - INFO - [embeddings] base_url=http://localhost:11434 model=embeddinggemma:latest n_inputs=5
2026-01-31 02:16:56,112 - scripts.iib.logger - INFO - [embeddings] API Key configured (length=6)
2026-01-31 02:16:56,113 - scripts.iib.logger - INFO - [embeddings] Request URL: http://localhost:11434/embeddings
2026-01-31 02:16:56,113 - scripts.iib.logger - INFO - [embeddings] Request model: embeddinggemma:latest
2026-01-31 02:16:56,113 - scripts.iib.logger - INFO - [embeddings] Request n_inputs: 5
2026-01-31 02:16:56,113 - scripts.iib.logger - INFO - [embeddings] Detected local API request (Ollama/LocalAI): http://localhost:11434
2026-01-31 02:16:56,113 - scripts.iib.logger - DEBUG - [embeddings] Request payload: {"model": "embeddinggemma:latest", "input": ["(truncated)", "(truncated)", "(truncated)", "(truncated)", "(truncated)"]}
2026-01-31 02:16:56,113 - scripts.iib.logger - INFO - [embeddings] Initiating HTTP request, timeout=120s
2026-01-31 02:16:56,119 - scripts.iib.logger - INFO - [embeddings] HTTP request completed
2026-01-31 02:16:56,119 - scripts.iib.logger - INFO - [embeddings] Response: status=404 content_length=18
2026-01-31 02:16:56,119 - scripts.iib.logger - ERROR - [embeddings] Non-200 response: status=404 body=404 page not found
2026-01-31 02:16:56,119 - scripts.iib.logger - ERROR - [embeddings] 404 error, please check API address and model name
2026-01-31 02:16:56,119 - scripts.iib.logger - ERROR - [build_embeddings] Embedding API failed for batch 1/1: 404 page not found
2026-01-31 02:16:56,119 - scripts.iib.logger - ERROR - [build_embeddings] Local API request failed, please check if Ollama is running: ollama serve

[romelsalwi]

If I click on "AI Analyze Tags"
INFO: 127.0.0.1:34176 - "POST /infinite_image_browsing/ai-chat HTTP/1.1" 404 Not Found

[zanllp]

The OPENAI_BASE_URL you provided is incorrect. If it were correct, you should be able to call the endpoint {base_url}/chat/completions (the API needs to be OpenAI-compatible). For example, with OpenRouter, it would be https://openrouter.ai/api/v1.

[romelsalwi]

You are right!
OPENAI_BASE_URL was incorrect. It's http://localhost:11434/v1/