1Z0-1127 Using OCI Generative AI Service — All Questions With Answers

{
  "model": "cohere.command-light",
  "compartmentId": "ocid1.compartment.oc1..aaaa...",
  "messages": [
    {"role": "user", "content": "Hello"}
  ],
  "parameters": {
    "temperature": 0.5,
    "maxTokens": 100
  }
}

{
  "status": 400,
  "code": "InvalidParameter",
  "message": "The model 'cohere.command-light' is not supported in region 'us-phoenix-1'."
}

{
  "data": [
    {"id": "ocid1.generativeaimodel.oc1..aaaa...", "display-name": "cohere.command", "lifecycle-state": "ACTIVE", "time-created": "2024-01-01T00:00:00Z"},
    {"id": "ocid1.generativeaimodel.oc1..bbbb...", "display-name": "cohere.command-light", "lifecycle-state": "INACTIVE", "time-created": "2024-01-02T00:00:00Z"},
    {"id": "ocid1.generativeaimodel.oc1..cccc...", "display-name": "cohere.embed-english-v3.0", "lifecycle-state": "ACTIVE"}
  ]
}

Policy JSON:
```json
{
  "statements": [
    "Allow group DataScientists to use generative-ai-inference in compartment ABC",
    "Allow group DataScientists to use generative-ai-models in compartment ABC"
  ]
}
```

```python
from langchain_community.llms import OCIGenAI
from langchain_community.embeddings import OCIGenAIEmbeddings
from langchain.vectorstores import FAISS

texts = ["Customer A query", "Customer B query"]
embedding_model = OCIGenAIEmbeddings(model="cohere.embed-v3", ...)
vector_store = FAISS.from_texts(texts, embedding_model)
llm = OCIGenAI(model="cohere.command-r", ...)
```

Refer to the exhibit.

{
  "syntaxVersion": "1.0",
  "statements": [
    {
      "effect": "Allow",
      "action": ["ai:generate-text"],
      "resource": ["ocid1.generativemodel.oc1..aaaaaa..."]
    }
  ]
}

Refer to the exhibit.

Error:
{
  "code": "InvalidParameter",
  "message": "The parameter 'top_p' should be between 0 and 1.",
  "status": 400
}

```json
{
  "statements": [
    "Allow group GenAIUsers to read generative-ai-models in tenancy"
  ]
}
```

Architecture:
- User -> API Gateway -> OCI Generative AI Service (Cohere Command Model)
- API Gateway has a custom authorizer that validates JWT tokens.
- The Generative AI service endpoint is private, accessible only from the VCN.