orxtra v0.13.0 /scheduler.src.orxtra.scheduler._prompt_providers
Updated
On this page

Fragment providers assembling the agent prompt: task preamble, task prompt, constraints, notepad entries, verified/stale lessons, and prior-failure context, each at a fixed priority.

#scheduler.src.orxtra.scheduler._prompt_providers

#scheduler.src.orxtra.scheduler._prompt_providers

Fragment providers for scheduler prompt assembly.

Each provider generates Fragment(s) for one layer of the assembled agent prompt. Providers read from scheduler state and produce content matching the exact format of the old inline string construction, preserving golden-output equivalence.

Priority ordering (ascending = earlier in output): 10 - task preamble (task ID + start_task instruction) 20 - task prompt (the user's prompt text, after variable substitution) 30 - constraints 40 - notepad 50 - lessons (verified) 55 - lessons (stale) 60 - prior failure context

#_load_template

python
def _load_template(name: str) -> str

Load a .md template from the prompts/ directory.

#TaskPreambleProvider

Produces the task-ID preamble fragment.

Context keys: task_id (str)

#fragments

python
def fragments(self, context: dict[str, Any]) -> list[Fragment]

#TaskPromptProvider

Produces the task prompt fragment (after variable substitution).

The task prompt is passed in via context["task_prompt"] already resolved (variables substituted by the caller before composition).

#fragments

python
def fragments(self, context: dict[str, Any]) -> list[Fragment]

#ConstraintsProvider

Produces the active constraints fragment.

Context keys: constraints (list of (text, tier) tuples)

#fragments

python
def fragments(self, context: dict[str, Any]) -> list[Fragment]

#NotepadProvider

Produces the notepad section fragment.

Renders notepad entries grouped by type. The notepad module provides data-only APIs; all rendering lives here.

Context keys: notepad_entries (list[NotepadEntry])

#fragments

python
def fragments(self, context: dict[str, Any]) -> list[Fragment]

#_render_notepad

python
def _render_notepad(entries: list[NotepadEntry]) -> str

Render notepad entries to markdown grouped by type.

#LessonsProvider

Produces lesson fragments (verified and/or stale).

Context keys: lessons (list of dicts with 'text' and 'stale' keys)

#fragments

python
def fragments(self, context: dict[str, Any]) -> list[Fragment]

#FailureContextProvider

Produces the prior failure context fragment.

Context keys: attempt (int) - current attempt number retry_inject_failure (bool) - whether to inject failure context prior_attempts (list of dicts with 'attempt' and 'error' keys)

#fragments

python
def fragments(self, context: dict[str, Any]) -> list[Fragment]
Search