​
Overview

The get_prompt function retrieves prompts stored in the Lucidic platform’s prompt database. It supports variable substitution and caching for optimal performance. This enables centralized prompt management and versioning.

​
Syntax

lai.get_prompt(
    prompt_name: str,
    variables: Optional[dict] = None,
    cache_ttl: Optional[int] = 300,
    label: Optional[str] = 'production'
) -> str

​
Parameters

​
prompt_name
string
required
The name/identifier of the prompt to retrieve from the database.
​
variables
dict
Dictionary of variables to substitute in the prompt. Variables in the prompt should be wrapped in double curly braces like {{variable_name}}.
​
cache_ttl
integer
default:"300"
Time-to-live for the cached prompt in seconds. Set to -1 to cache forever, or 0 to disable caching.
​
label
string
default:"production"
Label to retrieve a specific version of the prompt (e.g., “production”, “staging”, “v1.2”).

​
Returns

Returns the prompt as a string with variables substituted.

​
Examples

​
Basic Usage

import lucidicai as lai
from openai import OpenAI

lai.init(session_name="Customer Support", providers=["openai"])

# Get a simple prompt
prompt = lai.get_prompt("greeting_message")

client = OpenAI()
response = client.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "system", "content": prompt}]
)

​
With Variable Substitution

import lucidicai as lai

lai.init(session_name="Dynamic Prompts")

# Get prompt with variables
prompt = lai.get_prompt(
    prompt_name="customer_support",
    variables={
        "issue_type": "billing",
        "customer_name": "John Doe",
        "account_tier": "Premium"
    }
)

# If the prompt contains:
# "Hello {{customer_name}}, I see you have a {{issue_type}} issue with your {{account_tier}} account."
# 
# The result will be:
# "Hello John Doe, I see you have a billing issue with your Premium account."

​
Version Control with Labels

import lucidicai as lai

lai.init(session_name="A/B Testing")

# Get different versions of the same prompt
prompt_v1 = lai.get_prompt(
    prompt_name="product_description",
    label="v1.0"
)

prompt_v2 = lai.get_prompt(
    prompt_name="product_description",
    label="v2.0-beta"
)

# Use production version by default
prompt_prod = lai.get_prompt(
    prompt_name="product_description"  # Uses label="production" by default
)

​
Cache Management

import lucidicai as lai

lai.init(session_name="Performance Optimization")

# Cache for 1 hour (3600 seconds)
prompt_long_cache = lai.get_prompt(
    prompt_name="static_instructions",
    cache_ttl=3600
)

# Disable caching for frequently changing prompts
prompt_no_cache = lai.get_prompt(
    prompt_name="dynamic_news_prompt",
    cache_ttl=0
)

# Cache forever for stable prompts
prompt_permanent = lai.get_prompt(
    prompt_name="company_guidelines",
    cache_ttl=-1
)

​
Error Handling

import lucidicai as lai
from lucidicai.errors import PromptError

lai.init(session_name="Safe Prompt Usage")

try:
    prompt = lai.get_prompt(
        prompt_name="analysis_prompt",
        variables={"data_type": "financial"}
    )
except PromptError as e:
    # Handle missing variables or prompt not found
    print(f"Prompt error: {e}")
    # Use fallback prompt
    prompt = "Please analyze the provided financial data."

​
Complex Workflow Example

import lucidicai as lai
from openai import OpenAI

lai.init(session_name="Multi-Stage Analysis")

# Stage 1: Initial analysis
initial_prompt = lai.get_prompt(
    prompt_name="document_analysis",
    variables={
        "doc_type": "legal_contract",
        "focus_areas": "terms, conditions, obligations"
    }
)

client = OpenAI()
initial_analysis = client.chat.completions.create(
    model="gpt-4",
    messages=[
        {"role": "system", "content": initial_prompt},
        {"role": "user", "content": document_text}
    ]
)

# Stage 2: Detailed review based on initial findings
detail_prompt = lai.get_prompt(
    prompt_name="detailed_review",
    variables={
        "previous_findings": initial_analysis.choices[0].message.content,
        "review_depth": "comprehensive"
    },
    cache_ttl=0  # Don't cache as it includes dynamic content
)

detailed_review = client.chat.completions.create(
    model="gpt-4",
    messages=[
        {"role": "system", "content": detail_prompt},
        {"role": "user", "content": document_text}
    ]
)

​
Notes

  • Prompts must exist in the Lucidic platform before retrieval
  • Variable names in prompts must be wrapped in {{variable_name}}
  • Missing variables in the prompt will raise a PromptError
  • Unreplaced variables will trigger a warning
  • Caching improves performance for frequently used prompts
  • Returns empty string if no active session exists

​
Error Handling

from lucidicai.errors import PromptError

# Missing variable
try:
    prompt = lai.get_prompt(
        prompt_name="template",
        variables={"name": "Alice"}  # But prompt expects {{name}} and {{age}}
    )
except PromptError:
    print("Missing required variables in prompt")

# Prompt not found
try:
    prompt = lai.get_prompt("non_existent_prompt")
except Exception as e:
    print(f"Failed to retrieve prompt: {e}")

​
Best Practices

  1. Centralized Management: Store all prompts in Lucidic for version control
  2. Variable Naming: Use clear, descriptive variable names
  3. Cache Strategy: Use appropriate TTL based on prompt stability
  4. Error Handling: Always handle potential prompt errors
  5. Label Usage: Use labels for A/B testing and versioning

​
See Also

  • init - Required before using get_prompt
  • Prompt Database - Learn about the prompt database feature