In ODC, every prompt tweak requires a full republish which takes time. This "Compilation Tax" makes tuning AI in production nearly impossible.
Prompt Assembler treats prompts as dynamic assets in a central registry. You can update AI behavior instantly without a single redeploy.
How it works:
Store prompt templates in the Templates UI with unique codes.
Use tags like <<>> for dynamic values.
Call the Assemble_Prompt action in your logic.
The engine returns a finalized string for your LLM.
By separating instructions from code, you can fix AI issues in seconds and prevent prompt sprawl across your apps. It is a lightweight way to bring "PromptOps" to your ODC environment.
This initial release focuses on the core assembly engine. While we are continuously refining this pattern for our internal projects at DB Results Lab, please note the following boundaries:
No Version History: Current updates overwrite existing templates. We recommend keeping backups of critical prompt versions.
Basic Audit Trail: There is no built-in log linking specific requests to the exact prompt version used at that time.
Manual Data Sanitization: Developers are responsible for scrubbing PII or sensitive data before passing it into the context list.
BSD-3 license (https://opensource.org/licenses/BSD-3-Clause)