Add notebook for semantic reranking with HF Eland model

[leemthompo]

Notebook for semantic reranking using retriever + model uploaded via Eland

Test it in Colab

[gitnotebooks]

Found 1 changed notebook. Review the changes at https://gitnotebooks.com/elastic/elasticsearch-labs/pull/314

[jeffvestal]

Enabling telemetry step throws an error, referencing es_client before it is created. Should you be passing client ?

[<ipython-input-5-738307b21c1d>](https://localhost:8080/#) in <cell line: 4>()
      2 from telemetry import enable_telemetry
      3 
----> 4 es_client = enable_telemetry(es_client, "11-semantic-reranking-hugging-face")

NameError: name 'es_client' is not defined```

[jeffvestal]

eland_import_hub_model: error: argument --cloud-id: expected one argument
also an error with ES_API_KEY

should be

!eland_import_hub_model \
  --cloud-id $ELASTIC_CLOUD_ID \
  --es-api-key $ELASTIC_API_KEY \
  --hub-model-id cross-encoder/ms-marco-MiniLM-L-6-v2 \
  --task-type text_similarity \
  --clear-previous \
  --start

[jeffvestal]

I think you need to create an Inference endpoint after uploading the reranking model.
The last query in the notebook throws

  "error": {
    "root_cause": [],
    "type": "search_phase_execution_exception",
    "reason": "Computing updated ranks for results failed",
    "phase": "rank-feature",
    "grouped": true,
    "failed_shards": [],
    "caused_by": {
      "type": "resource_not_found_exception",
      "reason": "Inference endpoint not found [my-msmarco-minilm-model]"
    }
  },
  "status": 404
}```

[demjened]

Please add the missing step and check the other comments.

[demjened]

From what I've seen in other notebooks, in general we omit unnecessary output and only display the output of stages that showcase the current feature (in this case the before/after search results). This makes the notebook cleaner and easier to follow.
You can do this by clearing all cell outputs in your notebook tool (in VSCode there is a button on top), and replaying only the steps you want to see outputs of.

[leemthompo]

100% I deleted what I think is noise, bu it can be helpful to have some outputs (like the eland step) for error prone steps so users can see what 'success' looks like

[demjened]

I think we're missing a step in the notebook, the one that creates the rerank inference endpoint pointing to the uploaded MSMarco model.

[demjened]

LGTM. Not sure why the CI is failing, looks unrelated to this notebook.