Struggling with pipeline - qdrant, chromadb and ollama

[damienmc3]

Hello all, I am struggling setting up a custom pipeline with qdrant, chromadb and ollama.

Could someone please have a look? 🙂
Attached is the pipeline, Dockerfile, requirements.txt and docker-compose.yml

Any hint is really appreciated.

`2025-05-20 17:09:45 | WARNING | hayhooks.settings:check_cors_settings:74 | Using default CORS settings - All origins, methods, and headers are allowed.

INFO: Started server process [1]

INFO: Waiting for application startup.

2025-05-20 17:09:45 | INFO | hayhooks.server.app:deploy_pipelines:103 | Pipelines dir set to: /pipelines

INFO: Application startup complete.

INFO: Uvicorn running on http://0.0.0.0:1416/ (Press CTRL+C to quit)

2025-05-20 17:09:57 | ERROR | hayhooks.server.utils.deploy_utils:load_pipeline_module:176 | Error loading pipeline module: Failed to load 'chat_pipeline' pipeline module spec

INFO: 127.0.0.1:46674 - "POST /deploy_files HTTP/1.1" 422 Unprocessable Entity`

requirements.txt
[Dockerfile.txt](https://github.com/user-attachments/files/20
docker-compose.txt
353221/Dockerfile.txt)

pipeline_wrapper.txt

[damienmc3]

pipeline_wrapper.txt
Dockerfile.txt
requirements.txt
docker-compose.txt

[onestardao]

Hey! Looked over the logs — seems like it trips on:

Error loading pipeline module: Failed to load ‘chat pipeline’ pipeline module spec

Two possible rabbit holes:

1. Wrong file name or module import:
   Make sure the pipeline_wrapper.txt you’re referring to is actually getting interpreted as a proper .py or YAML pipeline — and that your Dockerfile includes it in the right WORKDIR. Haystack wants clean file names, and custom Python modules need to be inside your PYTHONPATH or explicitly registered.

2. Malformed pipeline definition:
   If you’re using YAML to define the pipeline, even a tiny formatting issue (like indentation or a wrong type field) could silently crash. Try pasting your YAML into yamllint.com and validating it.

Bonus tip: Temporarily comment out Ollama or Qdrant and try booting a minimal pipeline with just ChromaDB or Retriever to isolate which one’s being picky.

Keep poking it — half of RAG is just debugging YAML with faith and caffeine.