Expand description
The LLM prompts indexing uses: built in (v0) or an activated version, stored with the index
(<index>/_team, table prompts), for what the next index writes; nothing is redone for them.
Structs§
- Library
- Every saved version and which one each task uses; a task with none active uses v0.
- Prompts
- The instructions each task runs with, fixed for one index run.
- Version
- One saved version of a task’s instructions.
Enums§
Functions§
- builtin_
hash - Changes whenever a release changes this task’s built-in prompt.
- fixed
- What is always added after the instructions: the format and its JSON schema, so a server that ignores structured output still gets the shape.
- load
- save_as
- Save
libraryasbyif nobody saved sinceexpected.