Give a search agent several prompts and score a role by its best one

Migration 14 moves each non-empty agents.intent into agent_prompts as one
prompt named after the agent, with its cached vector, and drops the three
intent columns. A role's fit is the maximum over the agent's prompts of the
calibrated cosine, and scores.prompt names the prompt that won. The agents
API takes prompts and still takes intent; the role endpoint serves the
agent's score and prompt.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
WolfandClaude Opus 5.5 committed 2026-09-29 16:19:57 +02:00
1 parent b6afe24eaa
commit 4e3c3f130e
9 files changed
+552 -127

No files matched your search

+5 -2
View File
@@ -22,7 +22,7 @@ from typing import Any, Literal
import yaml
from pydantic import Field, ValidationError, field_validator, model_validator
from app.models import Gate, Limits, Weights
from app.models import Gate, Limits, Prompts, Weights
from auth import AuthConfig
from desk.config import AgentConfig, ModelTaskConfig, Section, WorkersConfig
@@ -129,9 +129,12 @@ class EmbedConfig(Section):
class AgentSeed(Section):
"""A search agent as `config.yaml` lists it. Imported into an empty
`agents` table once; after that the table is the answer."""
`agents` table once; after that the table is the answer. `prompts` is a
list of `{name, text}`; without it, `intent` is the one prompt, named
after the agent."""
name: str
prompts: Prompts | None = None
intent: str = ""
gates: list[Gate] = Field(default_factory=list)
weights: Weights = Field(default_factory=Weights)