constants.py

# there is no need for enforcing line length in this file as those are
# mostly special purpose constants
# ruff: noqa: E501
"""Constants used in business logic."""

SUBJECT_VALID = "SUBJECT_VALID"
SUBJECT_INVALID = "SUBJECT_INVALID"
POSSIBLE_QUESTION_VALIDATOR_RESPONSES = (
    SUBJECT_VALID,
    SUBJECT_INVALID,
)

# Default responses
INVALID_QUERY_RESP = (
    "I can only answer questions about OpenShift and Kubernetes. "
    "Please rephrase your question"
)
NO_RAG_CONTENT_RESP = (
    "The following response was generated without access to reference content:\n\n"
)

# templates
SUMMARIZATION_TEMPLATE = """
The following context contains several pieces of documentation. Please answer the user's question based on this context.
Documentation context:
{context_str}

User query:
{query_str}

Summary:

"""

SUMMARY_TASK_BREAKDOWN_TEMPLATE = (
    """
The following documentation contains a task list. Your job is to extract the list of tasks. """
    """If the user-supplied query seems unrelated to the list of tasks, please reply that you do not know what to do with the query and the summary documentation. """
    """Use only the supplied content and extract the task list.

Summary document:
{context_str}

User query:
{query_str}

What are the tasks?
"""
)

TASK_PERFORMER_PROMPT_TEMPLATE = """
Instructions:
- You are a helpful assistant.
- You are an expert in Kubernetes and OpenShift.
- Respond to questions about topics other than Kubernetes and OpenShift with: "I can only answer questions about Kubernetes and OpenShift"
- Refuse to participate in anything that could harm a human.
- Your job is to look at the following description and provide a response.
- Base your answer on the provided task and query and not on prior knowledge.

TASK:
{task}
QUERY:
{query}

Question:
Does the above query contain enough background information to complete the task? Provide a yes or no answer with explanation.

Response:
"""

TASK_REPHRASER_PROMPT_TEMPLATE = """
Instructions:
- You are a helpful assistant.
- Your job is to combine the information from the task and query into a single, new task.
- Base your answer on the provided task and query and not on prior knowledge.

TASK:
{task}
QUERY:
{query}

Please combine the information from the task and query into a single, new task.

Response:
"""

YES_OR_NO_CLASSIFIER_PROMPT_TEMPLATE = """
Instructions:
- determine if a statement is a yes or a no
- return a 1 if the statement is a yes statement
- return a 0 if the statement is a no statement
- return a 9 if you cannot determine if the statement is a yes or no

Examples:
Statement: Yes, that sounds good.
Response: 1

Statement: No, I don't think that is wise.
Response: 0

Statement: Apples are red.
Response: 9

Statement: {statement}
Response:
"""

# {{query}} is escaped because it will be replaced as a parameter at time of use
QUESTION_VALIDATOR_PROMPT_TEMPLATE = f"""
Instructions:
- You are a question classifying tool
- You are an expert in kubernetes and openshift
- Your job is to determine if a question is about kubernetes or openshift and to provide a one-word response
- If a question is not about kubernetes or openshift, answer with only the word {SUBJECT_INVALID}
- If a question is about kubernetes or openshift, answer with the word {SUBJECT_VALID}
- Use a comma to separate the words
- Do not provide explanation, only respond with the chosen words

Example Question:
Can you make me lunch with ham and cheese?
Example Response:
{SUBJECT_INVALID}

Example Question:
Why is the sky blue?
Example Response:
{SUBJECT_INVALID}

Example Question:
Can you help configure my cluster to automatically scale?
Example Response:
{SUBJECT_VALID}

Example Question:
please give me a vertical pod autoscaler configuration to manage my frontend deployment automatically.  Don't update the workload if there are less than 2 pods running.
Example Response:
{SUBJECT_VALID}

Question:
{{query}}
Response:
"""


# providers
PROVIDER_BAM = "bam"
PROVIDER_OPENAI = "openai"
PROVIDER_AZURE_OPENAI = "azure_openai"
PROVIDER_WATSONX = "watsonx"
SUPPORTED_PROVIDER_TYPES = frozenset(
    {PROVIDER_BAM, PROVIDER_OPENAI, PROVIDER_AZURE_OPENAI, PROVIDER_WATSONX}
)

# models
# embedding
TEI_EMBEDDING_MODEL = "BAAI/bge-base-en-v1.5"

# BAM
GRANITE_13B_CHAT_V1 = "ibm/granite-13b-chat-v1"
GRANITE_13B_CHAT_V2 = "ibm/granite-13b-chat-v2"
GRANITE_20B_CODE_INSTRUCT_V1 = "ibm/granite-20b-code-instruct-v1"

# OpenAI & Azure OpenAI
GPT35_TURBO_1106 = "gpt-3.5-turbo-1106"
GPT35_TURBO = "gpt-3.5-turbo"

# indexing constants
PRODUCT_INDEX = "product"
PRODUCT_DOCS_PERSIST_DIR = "./vector-db/ocp-product-docs"

# cache constants
IN_MEMORY_CACHE = "memory"
IN_MEMORY_CACHE_MAX_ENTRIES = 1000
REDIS_CACHE = "redis"
REDIS_CACHE_HOST = "lightspeed-redis-server.openshift-lightspeed.svc"
REDIS_CACHE_PORT = 6379
REDIS_CACHE_MAX_MEMORY = "1024mb"
REDIS_CACHE_MAX_MEMORY_POLICY = "allkeys-lru"
REDIS_CACHE_MAX_MEMORY_POLICIES = frozenset({"allkeys-lru", "volatile-lru"})
REDIS_RETRY_ON_ERROR = True
REDIS_RETRY_ON_TIMEOUT = True
REDIS_NUMBER_OF_RETRIES = 3


# provider and model roles
PROVIDER_MODEL_ROLES = frozenset({"default"})

EMBEDDINGS_ROOT_DIR = "/workspace/source/ocp-product-docs-plaintext"
OCP_DOCS_ROOT_URL = "https://docs.openshift.com/container-platform/"
OCP_DOCS_VERSION = "4.14"

# model configs
DEFAULT_CONTEXT_WINDOW_SIZE = 2000
DEFAULT_RESPONSE_TOKEN_LIMIT = 500

# default indentity for local testing and deployment
DEFAULT_USER_UID = "c1143120-551e-4a47-ad47-2748d6f3c81c"
DEFAULT_USER_NAME = "OLS"